Merge branch 'develop' into backtest_fitlivepredictions

.github/workflows/ci.yml

@@ -66,12 +66,6 @@ jobs:
     - name: Tests
       run: |
         pytest --random-order --cov=freqtrade --cov-config=.coveragerc
-      if: matrix.python-version != '3.9' || matrix.os != 'ubuntu-22.04'
-
-    - name: Tests incl. ccxt compatibility tests
-      run: |
-        pytest --random-order --cov=freqtrade --cov-config=.coveragerc --longrun
-      if: matrix.python-version == '3.9' && matrix.os == 'ubuntu-22.04'
 
     - name: Coveralls
       if: (runner.os == 'Linux' && matrix.python-version == '3.10' && matrix.os == 'ubuntu-22.04')
@@ -310,9 +304,64 @@ jobs:
           details: Freqtrade doc test failed!
           webhookUrl: ${{ secrets.DISCORD_WEBHOOK }}
 
+
+  build_linux_online:
+    # Run pytest with "live" checks
+    runs-on: ubuntu-22.04
+    # permissions:
+    steps:
+    - uses: actions/checkout@v3
+
+    - name: Set up Python
+      uses: actions/setup-python@v4
+      with:
+        python-version: "3.9"
+
+    - name: Cache_dependencies
+      uses: actions/cache@v3
+      id: cache
+      with:
+        path: ~/dependencies/
+        key: ${{ runner.os }}-dependencies
+
+    - name: pip cache (linux)
+      uses: actions/cache@v3
+      if: runner.os == 'Linux'
+      with:
+        path: ~/.cache/pip
+        key: test-${{ matrix.os }}-${{ matrix.python-version }}-pip
+
+    - name: TA binary *nix
+      if: steps.cache.outputs.cache-hit != 'true'
+      run: |
+        cd build_helpers && ./install_ta-lib.sh ${HOME}/dependencies/; cd ..
+
+    - name: Installation - *nix
+      if: runner.os == 'Linux'
+      run: |
+        python -m pip install --upgrade pip wheel
+        export LD_LIBRARY_PATH=${HOME}/dependencies/lib:$LD_LIBRARY_PATH
+        export TA_LIBRARY_PATH=${HOME}/dependencies/lib
+        export TA_INCLUDE_PATH=${HOME}/dependencies/include
+        pip install -r requirements-dev.txt
+        pip install -e .
+
+    - name: Tests incl. ccxt compatibility tests
+      run: |
+        pytest --random-order --cov=freqtrade --cov-config=.coveragerc --longrun
+
+
   # Notify only once - when CI completes (and after deploy) in case it's successfull
   notify-complete:
-    needs: [ build_linux, build_macos, build_windows, docs_check, mypy_version_check, pre-commit ]
+    needs: [
+      build_linux,
+      build_macos,
+      build_windows,
+      docs_check,
+      mypy_version_check,
+      pre-commit,
+      build_linux_online
+    ]
     runs-on: ubuntu-22.04
     # Discord notification can't handle schedule events
     if: (github.event_name != 'schedule')

build_helpers/publish_docker_arm64.sh

@@ -7,11 +7,13 @@ export DOCKER_BUILDKIT=1
 TAG=$(echo "${BRANCH_NAME}" | sed -e "s/\//_/g")
 TAG_PLOT=${TAG}_plot
 TAG_FREQAI=${TAG}_freqai
+TAG_FREQAI_RL=${TAG_FREQAI}rl
 TAG_PI="${TAG}_pi"
 
 TAG_ARM=${TAG}_arm
 TAG_PLOT_ARM=${TAG_PLOT}_arm
 TAG_FREQAI_ARM=${TAG_FREQAI}_arm
+TAG_FREQAI_RL_ARM=${TAG_FREQAI_RL}_arm
 CACHE_IMAGE=freqtradeorg/freqtrade_cache
 
 echo "Running for ${TAG}"
@@ -41,9 +43,11 @@ docker tag freqtrade:$TAG_ARM ${CACHE_IMAGE}:$TAG_ARM
 
 docker build --cache-from freqtrade:${TAG_ARM} --build-arg sourceimage=${CACHE_IMAGE} --build-arg sourcetag=${TAG_ARM} -t freqtrade:${TAG_PLOT_ARM} -f docker/Dockerfile.plot .
 docker build --cache-from freqtrade:${TAG_ARM} --build-arg sourceimage=${CACHE_IMAGE} --build-arg sourcetag=${TAG_ARM} -t freqtrade:${TAG_FREQAI_ARM} -f docker/Dockerfile.freqai .
+docker build --cache-from freqtrade:${TAG_ARM} --build-arg sourceimage=${CACHE_IMAGE} --build-arg sourcetag=${TAG_ARM} -t freqtrade:${TAG_FREQAI_RL_ARM} -f docker/Dockerfile.freqai_rl .
 
 docker tag freqtrade:$TAG_PLOT_ARM ${CACHE_IMAGE}:$TAG_PLOT_ARM
 docker tag freqtrade:$TAG_FREQAI_ARM ${CACHE_IMAGE}:$TAG_FREQAI_ARM
+docker tag freqtrade:$TAG_FREQAI_RL_ARM ${CACHE_IMAGE}:$TAG_FREQAI_RL_ARM
 
 # Run backtest
 docker run --rm -v $(pwd)/config_examples/config_bittrex.example.json:/freqtrade/config.json:ro -v $(pwd)/tests:/tests freqtrade:${TAG_ARM} backtesting --datadir /tests/testdata --strategy-path /tests/strategy/strats/ --strategy StrategyTestV3
@@ -58,6 +62,7 @@ docker images
 # docker push ${IMAGE_NAME}
 docker push ${CACHE_IMAGE}:$TAG_PLOT_ARM
 docker push ${CACHE_IMAGE}:$TAG_FREQAI_ARM
+docker push ${CACHE_IMAGE}:$TAG_FREQAI_RL_ARM
 docker push ${CACHE_IMAGE}:$TAG_ARM
 
 # Create multi-arch image
@@ -74,6 +79,9 @@ docker manifest push -p ${IMAGE_NAME}:${TAG_PLOT}
 docker manifest create ${IMAGE_NAME}:${TAG_FREQAI} ${CACHE_IMAGE}:${TAG_FREQAI_ARM} ${CACHE_IMAGE}:${TAG_FREQAI}
 docker manifest push -p ${IMAGE_NAME}:${TAG_FREQAI}
 
+docker manifest create ${IMAGE_NAME}:${TAG_FREQAI_RL} ${CACHE_IMAGE}:${TAG_FREQAI_RL_ARM} ${CACHE_IMAGE}:${TAG_FREQAI_RL}
+docker manifest push -p ${IMAGE_NAME}:${TAG_FREQAI_RL}
+
 # Tag as latest for develop builds
 if [ "${TAG}" = "develop" ]; then
     docker manifest create ${IMAGE_NAME}:latest ${CACHE_IMAGE}:${TAG_ARM} ${IMAGE_NAME}:${TAG_PI} ${CACHE_IMAGE}:${TAG}

build_helpers/publish_docker_multi.sh

@@ -6,6 +6,7 @@
 TAG=$(echo "${BRANCH_NAME}" | sed -e "s/\//_/g")
 TAG_PLOT=${TAG}_plot
 TAG_FREQAI=${TAG}_freqai
+TAG_FREQAI_RL=${TAG_FREQAI}rl
 TAG_PI="${TAG}_pi"
 
 PI_PLATFORM="linux/arm/v7"
@@ -51,9 +52,11 @@ docker tag freqtrade:$TAG ${CACHE_IMAGE}:$TAG
 
 docker build --cache-from freqtrade:${TAG} --build-arg sourceimage=${CACHE_IMAGE} --build-arg sourcetag=${TAG} -t freqtrade:${TAG_PLOT} -f docker/Dockerfile.plot .
 docker build --cache-from freqtrade:${TAG} --build-arg sourceimage=${CACHE_IMAGE} --build-arg sourcetag=${TAG} -t freqtrade:${TAG_FREQAI} -f docker/Dockerfile.freqai .
+docker build --cache-from freqtrade:${TAG_FREQAI} --build-arg sourceimage=${CACHE_IMAGE} --build-arg sourcetag=${TAG_FREQAI} -t freqtrade:${TAG_FREQAI_RL} -f docker/Dockerfile.freqai_rl .
 
 docker tag freqtrade:$TAG_PLOT ${CACHE_IMAGE}:$TAG_PLOT
 docker tag freqtrade:$TAG_FREQAI ${CACHE_IMAGE}:$TAG_FREQAI
+docker tag freqtrade:$TAG_FREQAI_RL ${CACHE_IMAGE}:$TAG_FREQAI_RL
 
 # Run backtest
 docker run --rm -v $(pwd)/config_examples/config_bittrex.example.json:/freqtrade/config.json:ro -v $(pwd)/tests:/tests freqtrade:${TAG} backtesting --datadir /tests/testdata --strategy-path /tests/strategy/strats/ --strategy StrategyTestV3
@@ -68,6 +71,7 @@ docker images
 docker push ${CACHE_IMAGE}
 docker push ${CACHE_IMAGE}:$TAG_PLOT
 docker push ${CACHE_IMAGE}:$TAG_FREQAI
+docker push ${CACHE_IMAGE}:$TAG_FREQAI_RL
 docker push ${CACHE_IMAGE}:$TAG
 
 

docker/Dockerfile.freqai_rl

@@ -0,0 +1,8 @@
+ARG sourceimage=freqtradeorg/freqtrade
+ARG sourcetag=develop_freqai
+FROM ${sourceimage}:${sourcetag}
+
+# Install dependencies
+COPY requirements-freqai.txt requirements-freqai-rl.txt /freqtrade/
+
+RUN pip install -r requirements-freqai-rl.txt --user --no-cache-dir

docs/backtesting.md

@@ -583,7 +583,8 @@ To utilize this, you can append `--timeframe-detail 5m` to your regular backtest
 freqtrade backtesting --strategy AwesomeStrategy --timeframe 1h --timeframe-detail 5m
 ```
 
-This will load 1h data as well as 5m data for the timeframe. The strategy will be analyzed with the 1h timeframe - and for every "open trade candle" (candles where a trade is open) the 5m data will be used to simulate intra-candle movements.
+This will load 1h data as well as 5m data for the timeframe. The strategy will be analyzed with the 1h timeframe, and Entry orders will only be placed at the main timeframe, however Order fills and exit signals will be evaluated at the 5m candle, simulating intra-candle movements.
+
 All callback functions (`custom_exit()`, `custom_stoploss()`, ... ) will be running for each 5m candle once the trade is opened (so 12 times in the above example of 1h timeframe, and 5m detailed timeframe).
 
 `--timeframe-detail` must be smaller than the original timeframe, otherwise backtesting will fail to start.

docs/developer.md

@@ -49,6 +49,13 @@ For more information about the [Remote container extension](https://code.visuals
 New code should be covered by basic unittests. Depending on the complexity of the feature, Reviewers may request more in-depth unittests.
 If necessary, the Freqtrade team can assist and give guidance with writing good tests (however please don't expect anyone to write the tests for you).
 
+#### How to run tests
+
+Use `pytest` in root folder to run all available testcases and confirm your local environment is setup correctly
+
+!!! Note "feature branches"
+    Tests are expected to pass on the `develop` and `stable` branches. Other branches may be work in progress with tests not working yet.
+
 #### Checking log content in tests
 
 Freqtrade uses 2 main methods to check log content in tests, `log_has()` and `log_has_re()` (to check using regex, in case of dynamic log-messages).
@@ -434,6 +441,11 @@ To keep the release-log short, best wrap the full git changelog into a collapsib
 </details>
 ```
 
+### FreqUI release
+
+If FreqUI has been updated substantially, make sure to create a release before merging the release branch.
+Make sure that freqUI CI on the release is finished and passed before merging the release.
+
 ### Create github release / tag
 
 Once the PR against stable is merged (best right after merging):

docs/freqai-parameter-table.md

@@ -4,9 +4,11 @@ The table below will list all configuration parameters available for FreqAI. Som
 
 Mandatory parameters are marked as **Required** and have to be set in one of the suggested ways.
 
+### General configuration parameters
+
 |  Parameter | Description |
 |------------|-------------|
-|  |  **General configuration parameters**
+|  |  **General configuration parameters within the `config.freqai` tree**
 | `freqai` | **Required.** <br> The parent dictionary containing all the parameters for controlling FreqAI. <br> **Datatype:** Dictionary.
 | `train_period_days` | **Required.** <br> Number of days to use for the training data (width of the sliding window). <br> **Datatype:** Positive integer.
 | `backtest_period_days` | **Required.** <br> Number of days to inference from the trained model before sliding the `train_period_days` window defined above, and retraining the model during backtesting (more info [here](freqai-running.md#backtesting)). This can be fractional days, but beware that the provided `timerange` will be divided by this number to yield the number of trainings necessary to complete the backtest. <br> **Datatype:** Float.
@@ -19,7 +21,13 @@ Mandatory parameters are marked as **Required** and have to be set in one of the
 | `follow_mode` | Use a `follower` that will look for models associated with a specific `identifier` and load those for inferencing. A `follower` will **not** train new models. <br> **Datatype:** Boolean. <br> Default: `False`.
 | `continual_learning` | Use the final state of the most recently trained model as starting point for the new model, allowing for incremental learning (more information can be found [here](freqai-running.md#continual-learning)). <br> **Datatype:** Boolean. <br> Default: `False`.
 | `write_metrics_to_disk` | Collect train timings, inference timings and cpu usage in json file. <br> **Datatype:** Boolean. <br> Default: `False`
-|  |  **Feature parameters**
+| `data_kitchen_thread_count` | <br> Designate the number of threads you want to use for data processing (outlier methods, normalization, etc.). This has no impact on the number of threads used for training. If user does not set it (default), FreqAI will use max number of threads - 2 (leaving 1 physical core available for Freqtrade bot and FreqUI) <br> **Datatype:** Positive integer.
+
+### Feature parameters
+
+|  Parameter | Description |
+|------------|-------------|
+|  |  **Feature parameters within the `freqai.feature_parameters` sub dictionary**
 | `feature_parameters` | A dictionary containing the parameters used to engineer the feature set. Details and examples are shown [here](freqai-feature-engineering.md). <br> **Datatype:** Dictionary.
 | `include_timeframes` | A list of timeframes that all indicators in `populate_any_indicators` will be created for. The list is added as features to the base indicators dataset. <br> **Datatype:** List of timeframes (strings).
 | `include_corr_pairlist` | A list of correlated coins that FreqAI will add as additional features to all `pair_whitelist` coins. All indicators set in `populate_any_indicators` during feature engineering (see details [here](freqai-feature-engineering.md)) will be created for each correlated coin. The correlated coins features are added to the base indicators dataset. <br> **Datatype:** List of assets (strings).
@@ -38,16 +46,49 @@ Mandatory parameters are marked as **Required** and have to be set in one of the
 | `noise_standard_deviation` | If set, FreqAI adds noise to the training features with the aim of preventing overfitting. FreqAI generates random deviates from a gaussian distribution with a standard deviation of `noise_standard_deviation` and adds them to all data points. `noise_standard_deviation` should be kept relative to the normalized space, i.e., between -1 and 1. In other words, since data in FreqAI is always normalized to be between -1 and 1, `noise_standard_deviation: 0.05` would result in 32% of the data being randomly increased/decreased by more than 2.5% (i.e., the percent of data falling within the first standard deviation). <br> **Datatype:** Integer. <br> Default: `0`.
 | `outlier_protection_percentage` | Enable to prevent outlier detection methods from discarding too much data. If more than `outlier_protection_percentage` % of points are detected as outliers by the SVM or DBSCAN, FreqAI will log a warning message and ignore outlier detection, i.e., the original dataset will be kept intact. If the outlier protection is triggered, no predictions will be made based on the training dataset. <br> **Datatype:** Float. <br> Default: `30`.
 | `reverse_train_test_order` | Split the feature dataset (see below) and use the latest data split for training and test on historical split of the data. This allows the model to be trained up to the most recent data point, while avoiding overfitting. However, you should be careful to understand the unorthodox nature of this parameter before employing it. <br> **Datatype:** Boolean. <br> Default: `False` (no reversal).
-|  |  **Data split parameters**
+
+### Data split parameters
+
+|  Parameter | Description |
+|------------|-------------|
+|  |  **Data split parameters within the `freqai.data_split_parameters` sub dictionary**
 | `data_split_parameters` | Include any additional parameters available from Scikit-learn `test_train_split()`, which are shown [here](https://scikit-learn.org/stable/modules/generated/sklearn.model_selection.train_test_split.html) (external website). <br> **Datatype:** Dictionary.
 | `test_size` | The fraction of data that should be used for testing instead of training. <br> **Datatype:** Positive float < 1.
 | `shuffle` | Shuffle the training data points during training. Typically, to not remove the chronological order of data in time-series forecasting, this is set to `False`. <br> **Datatype:** Boolean. <br> Defaut: `False`.
-|  |  **Model training parameters**
+
+### Model training parameters
+
+|  Parameter | Description |
+|------------|-------------|
+|  |  **Model training parameters within the `freqai.model_training_parameters` sub dictionary**
 | `model_training_parameters` | A flexible dictionary that includes all parameters available by the selected model library. For example, if you use `LightGBMRegressor`, this dictionary can contain any parameter available by the `LightGBMRegressor` [here](https://lightgbm.readthedocs.io/en/latest/pythonapi/lightgbm.LGBMRegressor.html) (external website). If you select a different model, this dictionary can contain any parameter from that model. A list of the currently available models can be found [here](freqai-configuration.md#using-different-prediction-models).  <br> **Datatype:** Dictionary.
 | `n_estimators` | The number of boosted trees to fit in the training of the model. <br> **Datatype:** Integer.
 | `learning_rate` | Boosting learning rate during training of the model. <br> **Datatype:** Float.
 | `n_jobs`, `thread_count`, `task_type` | Set the number of threads for parallel processing and the `task_type` (`gpu` or `cpu`). Different model libraries use different parameter names. <br> **Datatype:** Float.
+
+### Reinforcement Learning parameters
+
+|  Parameter | Description |
+|------------|-------------|
+|  |  **Reinforcement Learning Parameters within the `freqai.rl_config` sub dictionary**
+| `rl_config` | A dictionary containing the control parameters for a Reinforcement Learning model. <br> **Datatype:** Dictionary.
+| `train_cycles` | Training time steps will be set based on the `train_cycles * number of training data points. <br> **Datatype:** Integer.
+| `cpu_count` | Number of processors to dedicate to the Reinforcement Learning training process. <br> **Datatype:** int.
+| `max_trade_duration_candles`| Guides the agent training to keep trades below desired length. Example usage shown in `prediction_models/ReinforcementLearner.py` within the customizable `calculate_reward()` function. <br> **Datatype:** int.
+| `model_type` | Model string from stable_baselines3 or SBcontrib. Available strings include: `'TRPO', 'ARS', 'RecurrentPPO', 'MaskablePPO', 'PPO', 'A2C', 'DQN'`. User should ensure that `model_training_parameters` match those available to the corresponding stable_baselines3 model by visiting their documentaiton. [PPO doc](https://stable-baselines3.readthedocs.io/en/master/modules/ppo.html) (external website) <br> **Datatype:** string.
+| `policy_type` | One of the available policy types from stable_baselines3 <br> **Datatype:** string.
+| `max_training_drawdown_pct` | The maximum drawdown that the agent is allowed to experience during training. <br> **Datatype:** float. <br> Default: 0.8
+| `cpu_count` | Number of threads/cpus to dedicate to the Reinforcement Learning training process (depending on if `ReinforcementLearning_multiproc` is selected or not). Recommended to leave this untouched, by default, this value is set to the total number of physical cores minus 1. <br> **Datatype:** int. 
+| `model_reward_parameters` | Parameters used inside the customizable `calculate_reward()` function in `ReinforcementLearner.py` <br> **Datatype:** int.
+| `add_state_info` | Tell FreqAI to include state information in the feature set for training and inferencing. The current state variables include trade duration, current profit, trade position. This is only available in dry/live runs, and is automatically switched to false for backtesting. <br> **Datatype:** bool. <br> Default: `False`.
+| `net_arch` | Network architecture which is well described in [`stable_baselines3` doc](https://stable-baselines3.readthedocs.io/en/master/guide/custom_policy.html#examples). In summary: `[<shared layers>, dict(vf=[<non-shared value network layers>], pi=[<non-shared policy network layers>])]`. By default this is set to `[128, 128]`, which defines 2 shared hidden layers with 128 units each.
+| `randomize_starting_position` | Randomize the starting point of each episode to avoid overfitting. <br> **Datatype:** bool. <br> Default: `False`.
+
+### Additional parameters
+
+|  Parameter | Description |
+|------------|-------------|
 |  |  **Extraneous parameters**
-| `keras` | If the selected model makes use of Keras (typical for Tensorflow-based prediction models), this flag needs to be activated so that the model save/loading follows Keras standards. <br> **Datatype:** Boolean. <br> Default: `False`.
+| `freqai.keras` | If the selected model makes use of Keras (typical for Tensorflow-based prediction models), this flag needs to be activated so that the model save/loading follows Keras standards. <br> **Datatype:** Boolean. <br> Default: `False`.
-| `conv_width` | The width of a convolutional neural network input tensor. This replaces the need for shifting candles (`include_shifted_candles`) by feeding in historical data points as the second dimension of the tensor. Technically, this parameter can also be used for regressors, but it only adds computational overhead and does not change the model training/prediction. <br> **Datatype:** Integer. <br> Default: `2`.
+| `freqai.conv_width` | The width of a convolutional neural network input tensor. This replaces the need for shifting candles (`include_shifted_candles`) by feeding in historical data points as the second dimension of the tensor. Technically, this parameter can also be used for regressors, but it only adds computational overhead and does not change the model training/prediction. <br> **Datatype:** Integer. <br> Default: `2`.
-| `reduce_df_footprint` | Recast all numeric columns to float32/int32, with the objective of reducing ram/disk usage and decreasing train/inference timing. This parameter is set in the main level of the Freqtrade configuration file (not inside FreqAI). <br> **Datatype:** Boolean. <br> Default: `False`.
+| `freqai.reduce_df_footprint` | Recast all numeric columns to float32/int32, with the objective of reducing ram/disk usage and decreasing train/inference timing. This parameter is set in the main level of the Freqtrade configuration file (not inside FreqAI). <br> **Datatype:** Boolean. <br> Default: `False`.

docs/freqai-reinforcement-learning.md

@@ -0,0 +1,260 @@
+# Reinforcement Learning
+
+!!! Note "Installation size"
+    Reinforcement learning dependencies include large packages such as `torch`, which should be explicitly requested during `./setup.sh -i` by answering "y" to the question "Do you also want dependencies for freqai-rl (~700mb additional space required) [y/N]?".
+    Users who prefer docker should ensure they use the docker image appended with `_freqairl`.
+
+## Background and terminology
+
+### What is RL and why does FreqAI need it?
+
+Reinforcement learning involves two important components, the *agent* and the training *environment*. During agent training, the agent moves through historical data candle by candle, always making 1 of a set of actions: Long entry, long exit, short entry, short exit, neutral). During this training process, the environment tracks the performance of these actions and rewards the agent according to a custom user made `calculate_reward()` (here we offer a default reward for users to build on if they wish [details here](#creating-a-custom-reward-function)). The reward is used to train weights in a neural network.
+
+A second important component of the FreqAI RL implementation is the use of *state* information. State information is fed into the network at each step, including current profit, current position, and current trade duration. These are used to train the agent in the training environment, and to reinforce the agent in dry/live (this functionality is not available in backtesting). *FreqAI + Freqtrade is a perfect match for this reinforcing mechanism since this information is readily available in live deployments.*
+
+Reinforcement learning is a natural progression for FreqAI, since it adds a new layer of adaptivity and market reactivity that Classifiers and Regressors cannot match. However, Classifiers and Regressors have strengths that RL does not have such as robust predictions. Improperly trained RL agents may find "cheats" and "tricks" to maximize reward without actually winning any trades. For this reason, RL is more complex and demands a higher level of understanding than typical Classifiers and Regressors.
+
+### The RL interface
+
+With the current framework, we aim to expose the training environment via the common "prediction model" file, which is a user inherited `BaseReinforcementLearner` object (e.g. `freqai/prediction_models/ReinforcementLearner`). Inside this user class, the RL environment is available and customized via `MyRLEnv` as [shown below](#creating-a-custom-reward-function).
+
+We envision the majority of users focusing their effort on creative design of the `calculate_reward()` function [details here](#creating-a-custom-reward-function), while leaving the rest of the environment untouched. Other users may not touch the environment at all, and they will only play with the configuration settings and the powerful feature engineering that already exists in FreqAI. Meanwhile, we enable advanced users to create their own model classes entirely.
+
+The framework is built on stable_baselines3 (torch) and OpenAI gym for the base environment class. But generally speaking, the model class is well isolated. Thus, the addition of competing libraries can be easily integrated into the existing framework. For the environment, it is inheriting from `gym.env` which means that it is necessary to write an entirely new environment in order to switch to a different library.
+
+### Important considerations
+
+As explained above, the agent is "trained" in an artificial trading "environment". In our case, that environment may seem quite similar to a real Freqtrade backtesting environment, but it is *NOT*. In fact, the RL training environment is much more simplified. It does not incorporate any of the complicated strategy logic, such as callbacks like `custom_exit`, `custom_stoploss`, leverage controls, etc. The RL environment is instead a very "raw" representation of the true market, where the agent has free-will to learn the policy (read: stoploss, take profit, etc.) which is enforced by the `calculate_reward()`. Thus, it is important to consider that the agent training environment is not identical to the real world.
+
+## Running Reinforcement Learning
+
+Setting up and running a Reinforcement Learning model is the same as running a Regressor or Classifier. The same two flags, `--freqaimodel` and `--strategy`, must be defined on the command line:
+
+```bash
+freqtrade trade --freqaimodel ReinforcementLearner --strategy MyRLStrategy --config config.json
+```
+
+where `ReinforcementLearner` will use the templated `ReinforcementLearner` from `freqai/prediction_models/ReinforcementLearner` (or a custom user defined one located in `user_data/freqaimodels`). The strategy, on the other hand, follows the same base [feature engineering](freqai-feature-engineering.md) with `populate_any_indicators` as a typical Regressor:
+
+```python
+    def populate_any_indicators(
+        self, pair, df, tf, informative=None, set_generalized_indicators=False
+    ):
+
+        if informative is None:
+            informative = self.dp.get_pair_dataframe(pair, tf)
+
+        # first loop is automatically duplicating indicators for time periods
+        for t in self.freqai_info["feature_parameters"]["indicator_periods_candles"]:
+
+            t = int(t)
+            informative[f"%-{pair}rsi-period_{t}"] = ta.RSI(informative, timeperiod=t)
+            informative[f"%-{pair}mfi-period_{t}"] = ta.MFI(informative, timeperiod=t)
+            informative[f"%-{pair}adx-period_{t}"] = ta.ADX(informative, window=t)
+
+        # The following raw price values are necessary for RL models
+        informative[f"%-{pair}raw_close"] = informative["close"]
+        informative[f"%-{pair}raw_open"] = informative["open"]
+        informative[f"%-{pair}raw_high"] = informative["high"]
+        informative[f"%-{pair}raw_low"] = informative["low"]
+
+        indicators = [col for col in informative if col.startswith("%")]
+        # This loop duplicates and shifts all indicators to add a sense of recency to data
+        for n in range(self.freqai_info["feature_parameters"]["include_shifted_candles"] + 1):
+            if n == 0:
+                continue
+            informative_shift = informative[indicators].shift(n)
+            informative_shift = informative_shift.add_suffix("_shift-" + str(n))
+            informative = pd.concat((informative, informative_shift), axis=1)
+
+        df = merge_informative_pair(df, informative, self.config["timeframe"], tf, ffill=True)
+        skip_columns = [
+            (s + "_" + tf) for s in ["date", "open", "high", "low", "close", "volume"]
+        ]
+        df = df.drop(columns=skip_columns)
+
+        # Add generalized indicators here (because in live, it will call this
+        # function to populate indicators during training). Notice how we ensure not to
+        # add them multiple times
+        if set_generalized_indicators:
+
+            # For RL, there are no direct targets to set. This is filler (neutral)
+            # until the agent sends an action.
+            df["&-action"] = 0
+
+        return df
+```
+
+Most of the function remains the same as for typical Regressors, however, the function above shows how the strategy must pass the raw price data to the agent so that it has access to raw OHLCV in the training environment:
+
+```python
+        # The following features are necessary for RL models
+        informative[f"%-{pair}raw_close"] = informative["close"]
+        informative[f"%-{pair}raw_open"] = informative["open"]
+        informative[f"%-{pair}raw_high"] = informative["high"]
+        informative[f"%-{pair}raw_low"] = informative["low"]
+```
+
+Finally, there is no explicit "label" to make - instead it is necessary to assign the `&-action` column which will contain the agent's actions when accessed in `populate_entry/exit_trends()`. In the present example, the neutral action to 0. This value should align with the environment used. FreqAI provides two environments, both use 0 as the neutral action.
+
+After users realize there are no labels to set, they will soon understand that the agent is making its "own" entry and exit decisions. This makes strategy construction rather simple. The entry and exit signals come from the agent in the form of an integer - which are used directly to decide entries and exits in the strategy:
+
+```python
+    def populate_entry_trend(self, df: DataFrame, metadata: dict) -> DataFrame:
+
+        enter_long_conditions = [df["do_predict"] == 1, df["&-action"] == 1]
+
+        if enter_long_conditions:
+            df.loc[
+                reduce(lambda x, y: x & y, enter_long_conditions), ["enter_long", "enter_tag"]
+            ] = (1, "long")
+
+        enter_short_conditions = [df["do_predict"] == 1, df["&-action"] == 3]
+
+        if enter_short_conditions:
+            df.loc[
+                reduce(lambda x, y: x & y, enter_short_conditions), ["enter_short", "enter_tag"]
+            ] = (1, "short")
+
+        return df
+
+    def populate_exit_trend(self, df: DataFrame, metadata: dict) -> DataFrame:
+        exit_long_conditions = [df["do_predict"] == 1, df["&-action"] == 2]
+        if exit_long_conditions:
+            df.loc[reduce(lambda x, y: x & y, exit_long_conditions), "exit_long"] = 1
+
+        exit_short_conditions = [df["do_predict"] == 1, df["&-action"] == 4]
+        if exit_short_conditions:
+            df.loc[reduce(lambda x, y: x & y, exit_short_conditions), "exit_short"] = 1
+
+        return df
+```
+
+It is important to consider that `&-action` depends on which environment they choose to use. The example above shows 5 actions, where 0 is neutral, 1 is enter long, 2 is exit long, 3 is enter short and 4 is exit short.
+
+## Configuring the Reinforcement Learner
+
+In order to configure the `Reinforcement Learner` the following dictionary must exist in the `freqai` config:
+
+```json
+        "rl_config": {
+            "train_cycles": 25,
+            "add_state_info": true,
+            "max_trade_duration_candles": 300,
+            "max_training_drawdown_pct": 0.02,
+            "cpu_count": 8,
+            "model_type": "PPO",
+            "policy_type": "MlpPolicy",
+            "model_reward_parameters": {
+                "rr": 1,
+                "profit_aim": 0.025
+            }
+        }
+```
+
+Parameter details can be found [here](freqai-parameter-table.md), but in general the `train_cycles` decides how many times the agent should cycle through the candle data in its artificial environment to train weights in the model. `model_type` is a string which selects one of the available models in [stable_baselines](https://stable-baselines3.readthedocs.io/en/master/)(external link).
+
+!!! Note
+    If you would like to experiment with `continual_learning`, then you should set that value to `true` in the main `freqai` configuration dictionary. This will tell the Reinforcement Learning library to continue training new models from the final state of previous models, instead of retraining new models from scratch each time a retrain is initiated.
+
+!!! Note
+    Remember that the general `model_training_parameters` dictionary should contain all the model hyperparameter customizations for the particular `model_type`. For example, `PPO` parameters can be found [here](https://stable-baselines3.readthedocs.io/en/master/modules/ppo.html).
+
+## Creating a custom reward function
+
+As you begin to modify the strategy and the prediction model, you will quickly realize some important differences between the Reinforcement Learner and the Regressors/Classifiers. Firstly, the strategy does not set a target value (no labels!). Instead, you set the `calculate_reward()` function inside the `MyRLEnv` class (see below). A default `calculate_reward()` is provided inside `prediction_models/ReinforcementLearner.py` to demonstrate the necessary building blocks for creating rewards, but users are encouraged to create their own custom reinforcement learning model class (see below) and save it to `user_data/freqaimodels`. It is inside the `calculate_reward()` where creative theories about the market can be expressed. For example, you can reward your agent when it makes a winning trade, and penalize the agent when it makes a losing trade. Or perhaps, you wish to reward the agent for entering trades, and penalize the agent for sitting in trades too long. Below we show examples of how these rewards are all calculated:
+
+```python
+    from freqtrade.freqai.prediction_models.ReinforcementLearner import ReinforcementLearner
+    from freqtrade.freqai.RL.Base5ActionRLEnv import Actions, Base5ActionRLEnv, Positions
+
+
+    class MyCoolRLModel(ReinforcementLearner):
+        """
+        User created RL prediction model.
+
+        Save this file to `freqtrade/user_data/freqaimodels`
+
+        then use it with:
+
+        freqtrade trade --freqaimodel MyCoolRLModel --config config.json --strategy SomeCoolStrat
+
+        Here the users can override any of the functions
+        available in the `IFreqaiModel` inheritance tree. Most importantly for RL, this
+        is where the user overrides `MyRLEnv` (see below), to define custom
+        `calculate_reward()` function, or to override any other parts of the environment.
+
+        This class also allows users to override any other part of the IFreqaiModel tree.
+        For example, the user can override `def fit()` or `def train()` or `def predict()`
+        to take fine-tuned control over these processes.
+
+        Another common override may be `def data_cleaning_predict()` where the user can
+        take fine-tuned control over the data handling pipeline.
+        """
+        class MyRLEnv(Base5ActionRLEnv):
+            """
+            User made custom environment. This class inherits from BaseEnvironment and gym.env.
+            Users can override any functions from those parent classes. Here is an example
+            of a user customized `calculate_reward()` function.
+            """
+            def calculate_reward(self, action: int) -> float:
+                # first, penalize if the action is not valid
+                if not self._is_valid(action):
+                    return -2
+                pnl = self.get_unrealized_profit()
+
+                factor = 100
+                # reward agent for entering trades
+                if action in (Actions.Long_enter.value, Actions.Short_enter.value) \
+                        and self._position == Positions.Neutral:
+                    return 25
+                # discourage agent from not entering trades
+                if action == Actions.Neutral.value and self._position == Positions.Neutral:
+                    return -1
+                max_trade_duration = self.rl_config.get('max_trade_duration_candles', 300)
+                trade_duration = self._current_tick - self._last_trade_tick
+                if trade_duration <= max_trade_duration:
+                    factor *= 1.5
+                elif trade_duration > max_trade_duration:
+                    factor *= 0.5
+                # discourage sitting in position
+                if self._position in (Positions.Short, Positions.Long) and \
+                action == Actions.Neutral.value:
+                    return -1 * trade_duration / max_trade_duration
+                # close long
+                if action == Actions.Long_exit.value and self._position == Positions.Long:
+                    if pnl > self.profit_aim * self.rr:
+                        factor *= self.rl_config['model_reward_parameters'].get('win_reward_factor', 2)
+                    return float(pnl * factor)
+                # close short
+                if action == Actions.Short_exit.value and self._position == Positions.Short:
+                    if pnl > self.profit_aim * self.rr:
+                        factor *= self.rl_config['model_reward_parameters'].get('win_reward_factor', 2)
+                    return float(pnl * factor)
+                return 0.
+```
+
+### Using Tensorboard
+
+Reinforcement Learning models benefit from tracking training metrics. FreqAI has integrated Tensorboard to allow users to track training and evaluation performance across all coins and across all retrainings. Tensorboard is activated via the following command:
+
+```bash
+cd freqtrade
+tensorboard --logdir user_data/models/unique-id
+```
+
+where `unique-id` is the `identifier` set in the `freqai` configuration file. This command must be run in a separate shell to view the output in their browser at 127.0.0.1:6060 (6060 is the default port used by Tensorboard).
+
+![tensorboard](assets/tensorboard.jpg)
+
+### Choosing a base environment
+
+FreqAI provides two base environments, `Base4ActionEnvironment` and `Base5ActionEnvironment`. As the names imply, the environments are customized for agents that can select from 4 or 5 actions. In the `Base4ActionEnvironment`, the agent can enter long, enter short, hold neutral, or exit position. Meanwhile, in the `Base5ActionEnvironment`, the agent has the same actions as Base4, but instead of a single exit action, it separates exit long and exit short. The main changes stemming from the environment selection include:
+
+* the actions available in the `calculate_reward`
+* the actions consumed by the user strategy
+
+Both of the FreqAI provided environments inherit from an action/position agnostic environment object called the `BaseEnvironment`, which contains all shared logic. The architecture is designed to be easily customized. The simplest customization is the `calculate_reward()` (see details [here](#creating-a-custom-reward-function)). However, the customizations can be further extended into any of the functions inside the environment. You can do this by simply overriding those functions inside your `MyRLEnv` in the prediction model file. Or for more advanced customizations, it is encouraged to create an entirely new environment inherited from `BaseEnvironment`.
+
+!!! Note
+    FreqAI does not provide by default, a long-only training environment. However, creating one should be as simple as copy-pasting one of the built in environments and removing the `short` actions (and all associated references to those).

docs/producer-consumer.md

@@ -21,6 +21,7 @@ Enable subscribing to an instance by adding the `external_message_consumer` sect
                 "name": "default", // This can be any name you'd like, default is "default"
                 "host": "127.0.0.1", // The host from your producer's api_server config
                 "port": 8080, // The port from your producer's api_server config
+                "secure": false, // Use a secure websockets connection, default false
                 "ws_token": "sercet_Ws_t0ken" // The ws_token from your producer's api_server config
             }
         ],
@@ -42,6 +43,7 @@ Enable subscribing to an instance by adding the `external_message_consumer` sect
 | `producers.name` | **Required.** Name of this producer. This name must be used in calls to `get_producer_pairs()` and `get_producer_df()` if more than one producer is used.<br> **Datatype:** string
 | `producers.host` | **Required.** The hostname or IP address from your producer.<br> **Datatype:** string
 | `producers.port` | **Required.** The port matching the above host.<br> **Datatype:** string
+| `producers.secure` | **Optional.**  Use ssl in websockets connection. Default False.<br> **Datatype:** string
 | `producers.ws_token` | **Required.**  `ws_token` as configured on the producer.<br> **Datatype:** string
 | | **Optional settings**
 | `wait_timeout` | Timeout until we ping again if no message is received. <br>*Defaults to `300`.*<br> **Datatype:** Integer - in seconds.

docs/rest-api.md

@@ -389,6 +389,44 @@ Now anytime those types of RPC messages are sent in the bot, you will receive th
 }
 ```
 
+#### Reverse Proxy setup
+
+When using [Nginx](https://nginx.org/en/docs/), the following configuration is required for WebSockets to work (Note this configuration is incomplete, it's missing some information and can not be used as is):
+
+Please make sure to replace `<freqtrade_listen_ip>` (and the subsequent port) with the IP and Port matching your configuration/setup.
+
+```
+http {
+    map $http_upgrade $connection_upgrade {
+        default upgrade;
+        '' close;
+    }
+
+    #...
+
+    server {
+        #...
+
+        location / {
+            proxy_http_version 1.1;
+            proxy_pass http://<freqtrade_listen_ip>:8080;
+            proxy_set_header Upgrade $http_upgrade;
+            proxy_set_header Connection $connection_upgrade;
+            proxy_set_header Host $host;
+        }
+    }
+}
+```
+
+To properly configure your reverse proxy (securely), please consult it's documentation for proxying websockets.
+
+- **Traefik**: Traefik supports websockets out of the box, see the [documentation](https://doc.traefik.io/traefik/)
+- **Caddy**: Caddy v2 supports websockets out of the box, see the [documentation](https://caddyserver.com/docs/v2-upgrade#proxy)
+
+!!! Tip "SSL certificates"
+    You can use tools like certbot to setup ssl certificates to access your bot's UI through encrypted connection by using any fo the above reverse proxies.
+    While this will protect your data in transit, we do not recommend to run the freqtrade API outside of your private network (VPN, SSH tunnel).
+
 ### OpenAPI interface
 
 To enable the builtin openAPI interface (Swagger UI), specify `"enable_openapi": true` in the api_server configuration.

docs/strategy-customization.md

@@ -446,15 +446,17 @@ A full sample can be found [in the DataProvider section](#complete-data-provider
 
 ??? Note "Alternative candle types"
     Informative_pairs can also provide a 3rd tuple element defining the candle type explicitly.
-    Availability of alternative candle-types will depend on the trading-mode and the exchange. Details about this can be found in the exchange documentation.
+    Availability of alternative candle-types will depend on the trading-mode and the exchange. 
+    In general, spot pairs cannot be used in futures markets, and futures candles can't be used as informative pairs for spot bots.
+    Details about this may vary, if they do, this can be found in the exchange documentation.
 
     ``` python
     def informative_pairs(self):
         return [
-            ("ETH/USDT", "5m", ""),   # Uses default candletype, depends on trading_mode 
+            ("ETH/USDT", "5m", ""),   # Uses default candletype, depends on trading_mode (recommended)
-            ("ETH/USDT", "5m", "spot"),   # Forces usage of spot candles
+            ("ETH/USDT", "5m", "spot"),   # Forces usage of spot candles (only valid for bots running on spot markets).
-            ("BTC/TUSD", "15m", "futures"),  # Uses futures candles
+            ("BTC/TUSD", "15m", "futures"),  # Uses futures candles (only bots with `trading_mode=futures`)
-            ("BTC/TUSD", "15m", "mark"),  # Uses mark candles
+            ("BTC/TUSD", "15m", "mark"),  # Uses mark candles (only bots with `trading_mode=futures`)
         ]
     ```
 ***

docs/strategy_analysis_example.md

@@ -232,7 +232,7 @@ graph = generate_candlestick_graph(pair=pair,
 # Show graph inline
 # graph.show()
 
-# Render graph in a seperate window
+# Render graph in a separate window
 graph.show(renderer="browser")
 
 ```

freqtrade/__init__.py

@@ -1,5 +1,5 @@
 """ Freqtrade bot """
-__version__ = '2022.11.dev'
+__version__ = '2022.12.dev'
 
 if 'dev' in __version__:
     try:

freqtrade/constants.py

@@ -512,6 +512,7 @@ CONF_SCHEMA = {
                                 'minimum': 0,
                                 'maximum': 65535
                             },
+                            'secure': {'type': 'boolean', 'default': False},
                             'ws_token': {'type': 'string'},
                         },
                         'required': ['name', 'host', 'ws_token']
@@ -577,9 +578,27 @@ CONF_SCHEMA = {
                     },
                 },
                 "model_training_parameters": {
+                    "type": "object"
+                },
+                "rl_config": {
                     "type": "object",
                     "properties": {
-                        "n_estimators": {"type": "integer", "default": 1000}
+                        "train_cycles": {"type": "integer"},
+                        "max_trade_duration_candles": {"type": "integer"},
+                        "add_state_info": {"type": "boolean", "default": False},
+                        "max_training_drawdown_pct": {"type": "number", "default": 0.02},
+                        "cpu_count": {"type": "integer", "default": 1},
+                        "model_type": {"type": "string", "default": "PPO"},
+                        "policy_type": {"type": "string", "default": "MlpPolicy"},
+                        "net_arch": {"type": "array", "default": [128, 128]},
+                        "randomize_startinng_position": {"type": "boolean", "default": False},
+                        "model_reward_parameters": {
+                            "type": "object",
+                            "properties": {
+                                "rr": {"type": "number", "default": 1},
+                                "profit_aim": {"type": "number", "default": 0.025}
+                            }
+                        }
                     },
                 },
             },

freqtrade/exchange/bybit.py

@@ -20,7 +20,7 @@ class Bybit(Exchange):
     """
 
     _ft_has: Dict = {
-        "ohlcv_candle_limit": 200,
+        "ohlcv_candle_limit": 1000,
         "ccxt_futures_name": "linear",
         "ohlcv_has_history": False,
     }

freqtrade/exchange/kraken.py

@@ -218,3 +218,19 @@ class Kraken(Exchange):
             fees = sum(df['open_fund'] * df['open_mark'] * amount * time_in_ratio)
 
         return fees if is_short else -fees
+
+    def _trades_contracts_to_amount(self, trades: List) -> List:
+        """
+        Fix "last" id issue for kraken data downloads
+        This whole override can probably be removed once the following
+        issue is closed in ccxt: https://github.com/ccxt/ccxt/issues/15827
+        """
+        super()._trades_contracts_to_amount(trades)
+        if (
+            len(trades) > 0
+            and isinstance(trades[-1].get('info'), list)
+            and len(trades[-1].get('info', [])) > 7
+        ):
+
+            trades[-1]['id'] = trades[-1].get('info', [])[-1]
+        return trades

freqtrade/freqai/RL/Base4ActionRLEnv.py

@@ -0,0 +1,135 @@
+import logging
+from enum import Enum
+
+from gym import spaces
+
+from freqtrade.freqai.RL.BaseEnvironment import BaseEnvironment, Positions
+
+
+logger = logging.getLogger(__name__)
+
+
+class Actions(Enum):
+    Neutral = 0
+    Exit = 1
+    Long_enter = 2
+    Short_enter = 3
+
+
+class Base4ActionRLEnv(BaseEnvironment):
+    """
+    Base class for a 4 action environment
+    """
+
+    def set_action_space(self):
+        self.action_space = spaces.Discrete(len(Actions))
+
+    def step(self, action: int):
+        """
+        Logic for a single step (incrementing one candle in time)
+        by the agent
+        :param: action: int = the action type that the agent plans
+            to take for the current step.
+        :returns:
+            observation = current state of environment
+            step_reward = the reward from `calculate_reward()`
+            _done = if the agent "died" or if the candles finished
+            info = dict passed back to openai gym lib
+        """
+        self._done = False
+        self._current_tick += 1
+
+        if self._current_tick == self._end_tick:
+            self._done = True
+
+        self._update_unrealized_total_profit()
+
+        step_reward = self.calculate_reward(action)
+        self.total_reward += step_reward
+
+        trade_type = None
+        if self.is_tradesignal(action):
+            """
+            Action: Neutral, position: Long ->  Close Long
+            Action: Neutral, position: Short -> Close Short
+
+            Action: Long, position: Neutral -> Open Long
+            Action: Long, position: Short -> Close Short and Open Long
+
+            Action: Short, position: Neutral -> Open Short
+            Action: Short, position: Long -> Close Long and Open Short
+            """
+
+            if action == Actions.Neutral.value:
+                self._position = Positions.Neutral
+                trade_type = "neutral"
+                self._last_trade_tick = None
+            elif action == Actions.Long_enter.value:
+                self._position = Positions.Long
+                trade_type = "long"
+                self._last_trade_tick = self._current_tick
+            elif action == Actions.Short_enter.value:
+                self._position = Positions.Short
+                trade_type = "short"
+                self._last_trade_tick = self._current_tick
+            elif action == Actions.Exit.value:
+                self._update_total_profit()
+                self._position = Positions.Neutral
+                trade_type = "neutral"
+                self._last_trade_tick = None
+            else:
+                print("case not defined")
+
+            if trade_type is not None:
+                self.trade_history.append(
+                    {'price': self.current_price(), 'index': self._current_tick,
+                     'type': trade_type})
+
+        if self._total_profit < 1 - self.rl_config.get('max_training_drawdown_pct', 0.8):
+            self._done = True
+
+        self._position_history.append(self._position)
+
+        info = dict(
+            tick=self._current_tick,
+            total_reward=self.total_reward,
+            total_profit=self._total_profit,
+            position=self._position.value
+        )
+
+        observation = self._get_observation()
+
+        self._update_history(info)
+
+        return observation, step_reward, self._done, info
+
+    def is_tradesignal(self, action: int) -> bool:
+        """
+        Determine if the signal is a trade signal
+        e.g.: agent wants a Actions.Long_exit while it is in a Positions.short
+        """
+        return not ((action == Actions.Neutral.value and self._position == Positions.Neutral) or
+                    (action == Actions.Neutral.value and self._position == Positions.Short) or
+                    (action == Actions.Neutral.value and self._position == Positions.Long) or
+                    (action == Actions.Short_enter.value and self._position == Positions.Short) or
+                    (action == Actions.Short_enter.value and self._position == Positions.Long) or
+                    (action == Actions.Exit.value and self._position == Positions.Neutral) or
+                    (action == Actions.Long_enter.value and self._position == Positions.Long) or
+                    (action == Actions.Long_enter.value and self._position == Positions.Short))
+
+    def _is_valid(self, action: int) -> bool:
+        """
+        Determine if the signal is valid.
+        e.g.: agent wants a Actions.Long_exit while it is in a Positions.short
+        """
+        # Agent should only try to exit if it is in position
+        if action == Actions.Exit.value:
+            if self._position not in (Positions.Short, Positions.Long):
+                return False
+
+        # Agent should only try to enter if it is not in position
+        if action in (Actions.Short_enter.value, Actions.Long_enter.value):
+            if self._position != Positions.Neutral:
+                return False
+
+        return True

freqtrade/freqai/RL/Base5ActionRLEnv.py

@@ -0,0 +1,145 @@
+import logging
+from enum import Enum
+
+from gym import spaces
+
+from freqtrade.freqai.RL.BaseEnvironment import BaseEnvironment, Positions
+
+
+logger = logging.getLogger(__name__)
+
+
+class Actions(Enum):
+    Neutral = 0
+    Long_enter = 1
+    Long_exit = 2
+    Short_enter = 3
+    Short_exit = 4
+
+
+class Base5ActionRLEnv(BaseEnvironment):
+    """
+    Base class for a 5 action environment
+    """
+
+    def set_action_space(self):
+        self.action_space = spaces.Discrete(len(Actions))
+
+    def step(self, action: int):
+        """
+        Logic for a single step (incrementing one candle in time)
+        by the agent
+        :param: action: int = the action type that the agent plans
+            to take for the current step.
+        :returns:
+            observation = current state of environment
+            step_reward = the reward from `calculate_reward()`
+            _done = if the agent "died" or if the candles finished
+            info = dict passed back to openai gym lib
+        """
+        self._done = False
+        self._current_tick += 1
+
+        if self._current_tick == self._end_tick:
+            self._done = True
+
+        self._update_unrealized_total_profit()
+        step_reward = self.calculate_reward(action)
+        self.total_reward += step_reward
+
+        trade_type = None
+        if self.is_tradesignal(action):
+            """
+            Action: Neutral, position: Long ->  Close Long
+            Action: Neutral, position: Short -> Close Short
+
+            Action: Long, position: Neutral -> Open Long
+            Action: Long, position: Short -> Close Short and Open Long
+
+            Action: Short, position: Neutral -> Open Short
+            Action: Short, position: Long -> Close Long and Open Short
+            """
+
+            if action == Actions.Neutral.value:
+                self._position = Positions.Neutral
+                trade_type = "neutral"
+                self._last_trade_tick = None
+            elif action == Actions.Long_enter.value:
+                self._position = Positions.Long
+                trade_type = "long"
+                self._last_trade_tick = self._current_tick
+            elif action == Actions.Short_enter.value:
+                self._position = Positions.Short
+                trade_type = "short"
+                self._last_trade_tick = self._current_tick
+            elif action == Actions.Long_exit.value:
+                self._update_total_profit()
+                self._position = Positions.Neutral
+                trade_type = "neutral"
+                self._last_trade_tick = None
+            elif action == Actions.Short_exit.value:
+                self._update_total_profit()
+                self._position = Positions.Neutral
+                trade_type = "neutral"
+                self._last_trade_tick = None
+            else:
+                print("case not defined")
+
+            if trade_type is not None:
+                self.trade_history.append(
+                    {'price': self.current_price(), 'index': self._current_tick,
+                     'type': trade_type})
+
+        if (self._total_profit < self.max_drawdown or
+                self._total_unrealized_profit < self.max_drawdown):
+            self._done = True
+
+        self._position_history.append(self._position)
+
+        info = dict(
+            tick=self._current_tick,
+            total_reward=self.total_reward,
+            total_profit=self._total_profit,
+            position=self._position.value
+        )
+
+        observation = self._get_observation()
+
+        self._update_history(info)
+
+        return observation, step_reward, self._done, info
+
+    def is_tradesignal(self, action: int) -> bool:
+        """
+        Determine if the signal is a trade signal
+        e.g.: agent wants a Actions.Long_exit while it is in a Positions.short
+        """
+        return not ((action == Actions.Neutral.value and self._position == Positions.Neutral) or
+                    (action == Actions.Neutral.value and self._position == Positions.Short) or
+                    (action == Actions.Neutral.value and self._position == Positions.Long) or
+                    (action == Actions.Short_enter.value and self._position == Positions.Short) or
+                    (action == Actions.Short_enter.value and self._position == Positions.Long) or
+                    (action == Actions.Short_exit.value and self._position == Positions.Long) or
+                    (action == Actions.Short_exit.value and self._position == Positions.Neutral) or
+                    (action == Actions.Long_enter.value and self._position == Positions.Long) or
+                    (action == Actions.Long_enter.value and self._position == Positions.Short) or
+                    (action == Actions.Long_exit.value and self._position == Positions.Short) or
+                    (action == Actions.Long_exit.value and self._position == Positions.Neutral))
+
+    def _is_valid(self, action: int) -> bool:
+        # trade signal
+        """
+        Determine if the signal is valid.
+        e.g.: agent wants a Actions.Long_exit while it is in a Positions.short
+        """
+        # Agent should only try to exit if it is in position
+        if action in (Actions.Short_exit.value, Actions.Long_exit.value):
+            if self._position not in (Positions.Short, Positions.Long):
+                return False
+
+        # Agent should only try to enter if it is not in position
+        if action in (Actions.Short_enter.value, Actions.Long_enter.value):
+            if self._position != Positions.Neutral:
+                return False
+
+        return True

freqtrade/freqai/RL/BaseEnvironment.py

@@ -0,0 +1,307 @@
+import logging
+import random
+from abc import abstractmethod
+from enum import Enum
+from typing import Optional
+
+import gym
+import numpy as np
+import pandas as pd
+from gym import spaces
+from gym.utils import seeding
+from pandas import DataFrame
+
+from freqtrade.data.dataprovider import DataProvider
+
+
+logger = logging.getLogger(__name__)
+
+
+class Positions(Enum):
+    Short = 0
+    Long = 1
+    Neutral = 0.5
+
+    def opposite(self):
+        return Positions.Short if self == Positions.Long else Positions.Long
+
+
+class BaseEnvironment(gym.Env):
+    """
+    Base class for environments. This class is agnostic to action count.
+    Inherited classes customize this to include varying action counts/types,
+    See RL/Base5ActionRLEnv.py and RL/Base4ActionRLEnv.py
+    """
+
+    def __init__(self, df: DataFrame = DataFrame(), prices: DataFrame = DataFrame(),
+                 reward_kwargs: dict = {}, window_size=10, starting_point=True,
+                 id: str = 'baseenv-1', seed: int = 1, config: dict = {},
+                 dp: Optional[DataProvider] = None):
+        """
+        Initializes the training/eval environment.
+        :param df: dataframe of features
+        :param prices: dataframe of prices to be used in the training environment
+        :param window_size: size of window (temporal) to pass to the agent
+        :param reward_kwargs: extra config settings assigned by user in `rl_config`
+        :param starting_point: start at edge of window or not
+        :param id: string id of the environment (used in backend for multiprocessed env)
+        :param seed: Sets the seed of the environment higher in the gym.Env object
+        :param config: Typical user configuration file
+        :param dp: dataprovider from freqtrade
+        """
+        self.config = config
+        self.rl_config = config['freqai']['rl_config']
+        self.add_state_info = self.rl_config.get('add_state_info', False)
+        self.id = id
+        self.seed(seed)
+        self.reset_env(df, prices, window_size, reward_kwargs, starting_point)
+        self.max_drawdown = 1 - self.rl_config.get('max_training_drawdown_pct', 0.8)
+        self.compound_trades = config['stake_amount'] == 'unlimited'
+        if self.config.get('fee', None) is not None:
+            self.fee = self.config['fee']
+        elif dp is not None:
+            self.fee = dp._exchange.get_fee(symbol=dp.current_whitelist()[0])  # type: ignore
+        else:
+            self.fee = 0.0015
+
+    def reset_env(self, df: DataFrame, prices: DataFrame, window_size: int,
+                  reward_kwargs: dict, starting_point=True):
+        """
+        Resets the environment when the agent fails (in our case, if the drawdown
+        exceeds the user set max_training_drawdown_pct)
+        :param df: dataframe of features
+        :param prices: dataframe of prices to be used in the training environment
+        :param window_size: size of window (temporal) to pass to the agent
+        :param reward_kwargs: extra config settings assigned by user in `rl_config`
+        :param starting_point: start at edge of window or not
+        """
+        self.df = df
+        self.signal_features = self.df
+        self.prices = prices
+        self.window_size = window_size
+        self.starting_point = starting_point
+        self.rr = reward_kwargs["rr"]
+        self.profit_aim = reward_kwargs["profit_aim"]
+
+        # # spaces
+        if self.add_state_info:
+            self.total_features = self.signal_features.shape[1] + 3
+        else:
+            self.total_features = self.signal_features.shape[1]
+        self.shape = (window_size, self.total_features)
+        self.set_action_space()
+        self.observation_space = spaces.Box(
+            low=-1, high=1, shape=self.shape, dtype=np.float32)
+
+        # episode
+        self._start_tick: int = self.window_size
+        self._end_tick: int = len(self.prices) - 1
+        self._done: bool = False
+        self._current_tick: int = self._start_tick
+        self._last_trade_tick: Optional[int] = None
+        self._position = Positions.Neutral
+        self._position_history: list = [None]
+        self.total_reward: float = 0
+        self._total_profit: float = 1
+        self._total_unrealized_profit: float = 1
+        self.history: dict = {}
+        self.trade_history: list = []
+
+    @abstractmethod
+    def set_action_space(self):
+        """
+        Unique to the environment action count. Must be inherited.
+        """
+
+    def seed(self, seed: int = 1):
+        self.np_random, seed = seeding.np_random(seed)
+        return [seed]
+
+    def reset(self):
+
+        self._done = False
+
+        if self.starting_point is True:
+            if self.rl_config.get('randomize_starting_position', False):
+                length_of_data = int(self._end_tick / 4)
+                start_tick = random.randint(self.window_size + 1, length_of_data)
+                self._start_tick = start_tick
+            self._position_history = (self._start_tick * [None]) + [self._position]
+        else:
+            self._position_history = (self.window_size * [None]) + [self._position]
+
+        self._current_tick = self._start_tick
+        self._last_trade_tick = None
+        self._position = Positions.Neutral
+
+        self.total_reward = 0.
+        self._total_profit = 1.  # unit
+        self.history = {}
+        self.trade_history = []
+        self.portfolio_log_returns = np.zeros(len(self.prices))
+
+        self._profits = [(self._start_tick, 1)]
+        self.close_trade_profit = []
+        self._total_unrealized_profit = 1
+
+        return self._get_observation()
+
+    @abstractmethod
+    def step(self, action: int):
+        """
+        Step depeneds on action types, this must be inherited.
+        """
+        return
+
+    def _get_observation(self):
+        """
+        This may or may not be independent of action types, user can inherit
+        this in their custom "MyRLEnv"
+        """
+        features_window = self.signal_features[(
+            self._current_tick - self.window_size):self._current_tick]
+        if self.add_state_info:
+            features_and_state = DataFrame(np.zeros((len(features_window), 3)),
+                                           columns=['current_profit_pct',
+                                                    'position',
+                                                    'trade_duration'],
+                                           index=features_window.index)
+
+            features_and_state['current_profit_pct'] = self.get_unrealized_profit()
+            features_and_state['position'] = self._position.value
+            features_and_state['trade_duration'] = self.get_trade_duration()
+            features_and_state = pd.concat([features_window, features_and_state], axis=1)
+            return features_and_state
+        else:
+            return features_window
+
+    def get_trade_duration(self):
+        """
+        Get the trade duration if the agent is in a trade
+        """
+        if self._last_trade_tick is None:
+            return 0
+        else:
+            return self._current_tick - self._last_trade_tick
+
+    def get_unrealized_profit(self):
+        """
+        Get the unrealized profit if the agent is in a trade
+        """
+        if self._last_trade_tick is None:
+            return 0.
+
+        if self._position == Positions.Neutral:
+            return 0.
+        elif self._position == Positions.Short:
+            current_price = self.add_exit_fee(self.prices.iloc[self._current_tick].open)
+            last_trade_price = self.add_entry_fee(self.prices.iloc[self._last_trade_tick].open)
+            return (last_trade_price - current_price) / last_trade_price
+        elif self._position == Positions.Long:
+            current_price = self.add_entry_fee(self.prices.iloc[self._current_tick].open)
+            last_trade_price = self.add_exit_fee(self.prices.iloc[self._last_trade_tick].open)
+            return (current_price - last_trade_price) / last_trade_price
+        else:
+            return 0.
+
+    @abstractmethod
+    def is_tradesignal(self, action: int) -> bool:
+        """
+        Determine if the signal is a trade signal. This is
+        unique to the actions in the environment, and therefore must be
+        inherited.
+        """
+        return True
+
+    def _is_valid(self, action: int) -> bool:
+        """
+        Determine if the signal is valid.This is
+        unique to the actions in the environment, and therefore must be
+        inherited.
+        """
+        return True
+
+    def add_entry_fee(self, price):
+        return price * (1 + self.fee)
+
+    def add_exit_fee(self, price):
+        return price / (1 + self.fee)
+
+    def _update_history(self, info):
+        if not self.history:
+            self.history = {key: [] for key in info.keys()}
+
+        for key, value in info.items():
+            self.history[key].append(value)
+
+    @abstractmethod
+    def calculate_reward(self, action: int) -> float:
+        """
+        An example reward function. This is the one function that users will likely
+        wish to inject their own creativity into.
+        :param action: int = The action made by the agent for the current candle.
+        :return:
+        float = the reward to give to the agent for current step (used for optimization
+            of weights in NN)
+        """
+
+    def _update_unrealized_total_profit(self):
+        """
+        Update the unrealized total profit incase of episode end.
+        """
+        if self._position in (Positions.Long, Positions.Short):
+            pnl = self.get_unrealized_profit()
+            if self.compound_trades:
+                # assumes unit stake and compounding
+                unrl_profit = self._total_profit * (1 + pnl)
+            else:
+                # assumes unit stake and no compounding
+                unrl_profit = self._total_profit + pnl
+            self._total_unrealized_profit = unrl_profit
+
+    def _update_total_profit(self):
+        pnl = self.get_unrealized_profit()
+        if self.compound_trades:
+            # assumes unit stake and compounding
+            self._total_profit = self._total_profit * (1 + pnl)
+        else:
+            # assumes unit stake and no compounding
+            self._total_profit += pnl
+
+    def current_price(self) -> float:
+        return self.prices.iloc[self._current_tick].open
+
+    # Keeping around incase we want to start building more complex environment
+    # templates in the future.
+    # def most_recent_return(self):
+    #     """
+    #     Calculate the tick to tick return if in a trade.
+    #     Return is generated from rising prices in Long
+    #     and falling prices in Short positions.
+    #     The actions Sell/Buy or Hold during a Long position trigger the sell/buy-fee.
+    #     """
+    #     # Long positions
+    #     if self._position == Positions.Long:
+    #         current_price = self.prices.iloc[self._current_tick].open
+    #         previous_price = self.prices.iloc[self._current_tick - 1].open
+
+    #         if (self._position_history[self._current_tick - 1] == Positions.Short
+    #                 or self._position_history[self._current_tick - 1] == Positions.Neutral):
+    #             previous_price = self.add_entry_fee(previous_price)
+
+    #         return np.log(current_price) - np.log(previous_price)
+
+    #     # Short positions
+    #     if self._position == Positions.Short:
+    #         current_price = self.prices.iloc[self._current_tick].open
+    #         previous_price = self.prices.iloc[self._current_tick - 1].open
+    #         if (self._position_history[self._current_tick - 1] == Positions.Long
+    #                 or self._position_history[self._current_tick - 1] == Positions.Neutral):
+    #             previous_price = self.add_exit_fee(previous_price)
+
+    #         return np.log(previous_price) - np.log(current_price)
+
+    #     return 0
+
+    # def update_portfolio_log_returns(self, action):
+    #     self.portfolio_log_returns[self._current_tick] = self.most_recent_return(action)

freqtrade/freqai/RL/BaseReinforcementLearningModel.py

@@ -0,0 +1,396 @@
+import importlib
+import logging
+from abc import abstractmethod
+from datetime import datetime, timezone
+from pathlib import Path
+from typing import Any, Callable, Dict, Optional, Tuple, Type, Union
+
+import gym
+import numpy as np
+import numpy.typing as npt
+import pandas as pd
+import torch as th
+import torch.multiprocessing
+from pandas import DataFrame
+from stable_baselines3.common.callbacks import EvalCallback
+from stable_baselines3.common.monitor import Monitor
+from stable_baselines3.common.utils import set_random_seed
+from stable_baselines3.common.vec_env import SubprocVecEnv
+
+from freqtrade.exceptions import OperationalException
+from freqtrade.freqai.data_kitchen import FreqaiDataKitchen
+from freqtrade.freqai.freqai_interface import IFreqaiModel
+from freqtrade.freqai.RL.Base5ActionRLEnv import Actions, Base5ActionRLEnv
+from freqtrade.freqai.RL.BaseEnvironment import Positions
+from freqtrade.persistence import Trade
+
+
+logger = logging.getLogger(__name__)
+
+torch.multiprocessing.set_sharing_strategy('file_system')
+
+SB3_MODELS = ['PPO', 'A2C', 'DQN']
+SB3_CONTRIB_MODELS = ['TRPO', 'ARS', 'RecurrentPPO', 'MaskablePPO']
+
+
+class BaseReinforcementLearningModel(IFreqaiModel):
+    """
+    User created Reinforcement Learning Model prediction class
+    """
+
+    def __init__(self, **kwargs) -> None:
+        super().__init__(config=kwargs['config'])
+        self.max_threads = min(self.freqai_info['rl_config'].get(
+            'cpu_count', 1), max(int(self.max_system_threads / 2), 1))
+        th.set_num_threads(self.max_threads)
+        self.reward_params = self.freqai_info['rl_config']['model_reward_parameters']
+        self.train_env: Union[SubprocVecEnv, gym.Env] = None
+        self.eval_env: Union[SubprocVecEnv, gym.Env] = None
+        self.eval_callback: Optional[EvalCallback] = None
+        self.model_type = self.freqai_info['rl_config']['model_type']
+        self.rl_config = self.freqai_info['rl_config']
+        self.continual_learning = self.freqai_info.get('continual_learning', False)
+        if self.model_type in SB3_MODELS:
+            import_str = 'stable_baselines3'
+        elif self.model_type in SB3_CONTRIB_MODELS:
+            import_str = 'sb3_contrib'
+        else:
+            raise OperationalException(f'{self.model_type} not available in stable_baselines3 or '
+                                       f'sb3_contrib. please choose one of {SB3_MODELS} or '
+                                       f'{SB3_CONTRIB_MODELS}')
+
+        mod = importlib.import_module(import_str, self.model_type)
+        self.MODELCLASS = getattr(mod, self.model_type)
+        self.policy_type = self.freqai_info['rl_config']['policy_type']
+        self.unset_outlier_removal()
+        self.net_arch = self.rl_config.get('net_arch', [128, 128])
+        self.dd.model_type = "stable_baselines"
+
+    def unset_outlier_removal(self):
+        """
+        If user has activated any function that may remove training points, this
+        function will set them to false and warn them
+        """
+        if self.ft_params.get('use_SVM_to_remove_outliers', False):
+            self.ft_params.update({'use_SVM_to_remove_outliers': False})
+            logger.warning('User tried to use SVM with RL. Deactivating SVM.')
+        if self.ft_params.get('use_DBSCAN_to_remove_outliers', False):
+            self.ft_params.update({'use_DBSCAN_to_remove_outliers': False})
+            logger.warning('User tried to use DBSCAN with RL. Deactivating DBSCAN.')
+        if self.freqai_info['data_split_parameters'].get('shuffle', False):
+            self.freqai_info['data_split_parameters'].update({'shuffle': False})
+            logger.warning('User tried to shuffle training data. Setting shuffle to False')
+
+    def train(
+        self, unfiltered_df: DataFrame, pair: str, dk: FreqaiDataKitchen, **kwargs
+    ) -> Any:
+        """
+        Filter the training data and train a model to it. Train makes heavy use of the datakitchen
+        for storing, saving, loading, and analyzing the data.
+        :param unfiltered_df: Full dataframe for the current training period
+        :param metadata: pair metadata from strategy.
+        :returns:
+        :model: Trained model which can be used to inference (self.predict)
+        """
+
+        logger.info("--------------------Starting training " f"{pair} --------------------")
+
+        features_filtered, labels_filtered = dk.filter_features(
+            unfiltered_df,
+            dk.training_features_list,
+            dk.label_list,
+            training_filter=True,
+        )
+
+        data_dictionary: Dict[str, Any] = dk.make_train_test_datasets(
+            features_filtered, labels_filtered)
+        dk.fit_labels()  # FIXME useless for now, but just satiating append methods
+
+        # normalize all data based on train_dataset only
+        prices_train, prices_test = self.build_ohlc_price_dataframes(dk.data_dictionary, pair, dk)
+        data_dictionary = dk.normalize_data(data_dictionary)
+
+        # data cleaning/analysis
+        self.data_cleaning_train(dk)
+
+        logger.info(
+            f'Training model on {len(dk.data_dictionary["train_features"].columns)}'
+            f' features and {len(data_dictionary["train_features"])} data points'
+        )
+
+        self.set_train_and_eval_environments(data_dictionary, prices_train, prices_test, dk)
+
+        model = self.fit(data_dictionary, dk)
+
+        logger.info(f"--------------------done training {pair}--------------------")
+
+        return model
+
+    def set_train_and_eval_environments(self, data_dictionary: Dict[str, DataFrame],
+                                        prices_train: DataFrame, prices_test: DataFrame,
+                                        dk: FreqaiDataKitchen):
+        """
+        User can override this if they are using a custom MyRLEnv
+        :param data_dictionary: dict = common data dictionary containing train and test
+            features/labels/weights.
+        :param prices_train/test: DataFrame = dataframe comprised of the prices to be used in the
+            environment during training or testing
+        :param dk: FreqaiDataKitchen = the datakitchen for the current pair
+        """
+        train_df = data_dictionary["train_features"]
+        test_df = data_dictionary["test_features"]
+
+        self.train_env = self.MyRLEnv(df=train_df,
+                                      prices=prices_train,
+                                      window_size=self.CONV_WIDTH,
+                                      reward_kwargs=self.reward_params,
+                                      config=self.config,
+                                      dp=self.data_provider)
+        self.eval_env = Monitor(self.MyRLEnv(df=test_df,
+                                             prices=prices_test,
+                                             window_size=self.CONV_WIDTH,
+                                             reward_kwargs=self.reward_params,
+                                             config=self.config,
+                                             dp=self.data_provider))
+        self.eval_callback = EvalCallback(self.eval_env, deterministic=True,
+                                          render=False, eval_freq=len(train_df),
+                                          best_model_save_path=str(dk.data_path))
+
+    @abstractmethod
+    def fit(self, data_dictionary: Dict[str, Any], dk: FreqaiDataKitchen, **kwargs):
+        """
+        Agent customizations and abstract Reinforcement Learning customizations
+        go in here. Abstract method, so this function must be overridden by
+        user class.
+        """
+        return
+
+    def get_state_info(self, pair: str) -> Tuple[float, float, int]:
+        """
+        State info during dry/live (not backtesting) which is fed back
+        into the model.
+        :param pair: str = COIN/STAKE to get the environment information for
+        :return:
+        :market_side: float = representing short, long, or neutral for
+            pair
+        :current_profit: float = unrealized profit of the current trade
+        :trade_duration: int = the number of candles that the trade has
+            been open for
+        """
+        open_trades = Trade.get_trades_proxy(is_open=True)
+        market_side = 0.5
+        current_profit: float = 0
+        trade_duration = 0
+        for trade in open_trades:
+            if trade.pair == pair:
+                if self.data_provider._exchange is None:  # type: ignore
+                    logger.error('No exchange available.')
+                    return 0, 0, 0
+                else:
+                    current_rate = self.data_provider._exchange.get_rate(  # type: ignore
+                                pair, refresh=False, side="exit", is_short=trade.is_short)
+
+                now = datetime.now(timezone.utc).timestamp()
+                trade_duration = int((now - trade.open_date_utc.timestamp()) / self.base_tf_seconds)
+                current_profit = trade.calc_profit_ratio(current_rate)
+
+        return market_side, current_profit, int(trade_duration)
+
+    def predict(
+        self, unfiltered_df: DataFrame, dk: FreqaiDataKitchen, **kwargs
+    ) -> Tuple[DataFrame, npt.NDArray[np.int_]]:
+        """
+        Filter the prediction features data and predict with it.
+        :param unfiltered_dataframe: Full dataframe for the current backtest period.
+        :return:
+        :pred_df: dataframe containing the predictions
+        :do_predict: np.array of 1s and 0s to indicate places where freqai needed to remove
+        data (NaNs) or felt uncertain about data (PCA and DI index)
+        """
+
+        dk.find_features(unfiltered_df)
+        filtered_dataframe, _ = dk.filter_features(
+            unfiltered_df, dk.training_features_list, training_filter=False
+        )
+        filtered_dataframe = dk.normalize_data_from_metadata(filtered_dataframe)
+        dk.data_dictionary["prediction_features"] = filtered_dataframe
+
+        # optional additional data cleaning/analysis
+        self.data_cleaning_predict(dk)
+
+        pred_df = self.rl_model_predict(
+            dk.data_dictionary["prediction_features"], dk, self.model)
+        pred_df.fillna(0, inplace=True)
+
+        return (pred_df, dk.do_predict)
+
+    def rl_model_predict(self, dataframe: DataFrame,
+                         dk: FreqaiDataKitchen, model: Any) -> DataFrame:
+        """
+        A helper function to make predictions in the Reinforcement learning module.
+        :param dataframe: DataFrame = the dataframe of features to make the predictions on
+        :param dk: FreqaiDatakitchen = data kitchen for the current pair
+        :param model: Any = the trained model used to inference the features.
+        """
+        output = pd.DataFrame(np.zeros(len(dataframe)), columns=dk.label_list)
+
+        def _predict(window):
+            observations = dataframe.iloc[window.index]
+            if self.live and self.rl_config.get('add_state_info', False):
+                market_side, current_profit, trade_duration = self.get_state_info(dk.pair)
+                observations['current_profit_pct'] = current_profit
+                observations['position'] = market_side
+                observations['trade_duration'] = trade_duration
+            res, _ = model.predict(observations, deterministic=True)
+            return res
+
+        output = output.rolling(window=self.CONV_WIDTH).apply(_predict)
+
+        return output
+
+    def build_ohlc_price_dataframes(self, data_dictionary: dict,
+                                    pair: str, dk: FreqaiDataKitchen) -> Tuple[DataFrame,
+                                                                               DataFrame]:
+        """
+        Builds the train prices and test prices for the environment.
+        """
+
+        pair = pair.replace(':', '')
+        train_df = data_dictionary["train_features"]
+        test_df = data_dictionary["test_features"]
+
+        # price data for model training and evaluation
+        tf = self.config['timeframe']
+        ohlc_list = [f'%-{pair}raw_open_{tf}', f'%-{pair}raw_low_{tf}',
+                     f'%-{pair}raw_high_{tf}', f'%-{pair}raw_close_{tf}']
+        rename_dict = {f'%-{pair}raw_open_{tf}': 'open', f'%-{pair}raw_low_{tf}': 'low',
+                       f'%-{pair}raw_high_{tf}': ' high', f'%-{pair}raw_close_{tf}': 'close'}
+
+        prices_train = train_df.filter(ohlc_list, axis=1)
+        if prices_train.empty:
+            raise OperationalException('Reinforcement learning module didnt find the raw prices '
+                                       'assigned in populate_any_indicators. Please assign them '
+                                       'with:\n'
+                                       'informative[f"%-{pair}raw_close"] = informative["close"]\n'
+                                       'informative[f"%-{pair}raw_open"] = informative["open"]\n'
+                                       'informative[f"%-{pair}raw_high"] = informative["high"]\n'
+                                       'informative[f"%-{pair}raw_low"] = informative["low"]\n')
+        prices_train.rename(columns=rename_dict, inplace=True)
+        prices_train.reset_index(drop=True)
+
+        prices_test = test_df.filter(ohlc_list, axis=1)
+        prices_test.rename(columns=rename_dict, inplace=True)
+        prices_test.reset_index(drop=True)
+
+        return prices_train, prices_test
+
+    def load_model_from_disk(self, dk: FreqaiDataKitchen) -> Any:
+        """
+        Can be used by user if they are trying to limit_ram_usage *and*
+        perform continual learning.
+        For now, this is unused.
+        """
+        exists = Path(dk.data_path / f"{dk.model_filename}_model").is_file()
+        if exists:
+            model = self.MODELCLASS.load(dk.data_path / f"{dk.model_filename}_model")
+        else:
+            logger.info('No model file on disk to continue learning from.')
+
+        return model
+
+    def _on_stop(self):
+        """
+        Hook called on bot shutdown. Close SubprocVecEnv subprocesses for clean shutdown.
+        """
+
+        if self.train_env:
+            self.train_env.close()
+
+        if self.eval_env:
+            self.eval_env.close()
+
+    # Nested class which can be overridden by user to customize further
+    class MyRLEnv(Base5ActionRLEnv):
+        """
+        User can override any function in BaseRLEnv and gym.Env. Here the user
+        sets a custom reward based on profit and trade duration.
+        """
+
+        def calculate_reward(self, action: int) -> float:
+            """
+            An example reward function. This is the one function that users will likely
+            wish to inject their own creativity into.
+            :param action: int = The action made by the agent for the current candle.
+            :return:
+            float = the reward to give to the agent for current step (used for optimization
+                of weights in NN)
+            """
+            # first, penalize if the action is not valid
+            if not self._is_valid(action):
+                return -2
+
+            pnl = self.get_unrealized_profit()
+            factor = 100.
+
+            # reward agent for entering trades
+            if (action in (Actions.Long_enter.value, Actions.Short_enter.value)
+                    and self._position == Positions.Neutral):
+                return 25
+            # discourage agent from not entering trades
+            if action == Actions.Neutral.value and self._position == Positions.Neutral:
+                return -1
+
+            max_trade_duration = self.rl_config.get('max_trade_duration_candles', 300)
+            if self._last_trade_tick:
+                trade_duration = self._current_tick - self._last_trade_tick
+            else:
+                trade_duration = 0
+
+            if trade_duration <= max_trade_duration:
+                factor *= 1.5
+            elif trade_duration > max_trade_duration:
+                factor *= 0.5
+
+            # discourage sitting in position
+            if (self._position in (Positions.Short, Positions.Long) and
+               action == Actions.Neutral.value):
+                return -1 * trade_duration / max_trade_duration
+
+            # close long
+            if action == Actions.Long_exit.value and self._position == Positions.Long:
+                if pnl > self.profit_aim * self.rr:
+                    factor *= self.rl_config['model_reward_parameters'].get('win_reward_factor', 2)
+                return float(pnl * factor)
+
+            # close short
+            if action == Actions.Short_exit.value and self._position == Positions.Short:
+                if pnl > self.profit_aim * self.rr:
+                    factor *= self.rl_config['model_reward_parameters'].get('win_reward_factor', 2)
+                return float(pnl * factor)
+
+            return 0.
+
+
+def make_env(MyRLEnv: Type[gym.Env], env_id: str, rank: int,
+             seed: int, train_df: DataFrame, price: DataFrame,
+             reward_params: Dict[str, int], window_size: int, monitor: bool = False,
+             config: Dict[str, Any] = {}) -> Callable:
+    """
+    Utility function for multiprocessed env.
+
+    :param env_id: (str) the environment ID
+    :param num_env: (int) the number of environment you wish to have in subprocesses
+    :param seed: (int) the inital seed for RNG
+    :param rank: (int) index of the subprocess
+    :return: (Callable)
+    """
+
+    def _init() -> gym.Env:
+
+        env = MyRLEnv(df=train_df, prices=price, window_size=window_size,
+                      reward_kwargs=reward_params, id=env_id, seed=seed + rank, config=config)
+        if monitor:
+            env = Monitor(env)
+        return env
+    set_random_seed(seed)
+    return _init

freqtrade/freqai/data_drawer.py

@@ -1,4 +1,5 @@
 import collections
+import importlib
 import logging
 import re
 import shutil
@@ -99,6 +100,7 @@ class FreqaiDataDrawer:
         self.empty_pair_dict: pair_info = {
                 "model_filename": "", "trained_timestamp": 0,
                 "data_path": "", "extras": {}}
+        self.model_type = self.freqai_info.get('model_save_type', 'joblib')
 
     def update_metric_tracker(self, metric: str, value: float, pair: str) -> None:
         """
@@ -497,10 +499,12 @@ class FreqaiDataDrawer:
         save_path = Path(dk.data_path)
 
         # Save the trained model
-        if not dk.keras:
+        if self.model_type == 'joblib':
             dump(model, save_path / f"{dk.model_filename}_model.joblib")
-        else:
+        elif self.model_type == 'keras':
             model.save(save_path / f"{dk.model_filename}_model.h5")
+        elif 'stable_baselines' in self.model_type:
+            model.save(save_path / f"{dk.model_filename}_model.zip")
 
         if dk.svm_model is not None:
             dump(dk.svm_model, save_path / f"{dk.model_filename}_svm_model.joblib")
@@ -527,11 +531,10 @@ class FreqaiDataDrawer:
                 dk.pca, open(dk.data_path / f"{dk.model_filename}_pca_object.pkl", "wb")
             )
 
-        # if self.live:
-        # store as much in ram as possible to increase performance
         self.model_dictionary[coin] = model
         self.pair_dict[coin]["model_filename"] = dk.model_filename
         self.pair_dict[coin]["data_path"] = str(dk.data_path)
+
         if coin not in self.meta_data_dictionary:
             self.meta_data_dictionary[coin] = {}
         self.meta_data_dictionary[coin]["train_df"] = dk.data_dictionary["train_features"]
@@ -563,14 +566,6 @@ class FreqaiDataDrawer:
         if dk.live:
             dk.model_filename = self.pair_dict[coin]["model_filename"]
             dk.data_path = Path(self.pair_dict[coin]["data_path"])
-            if self.freqai_info.get("follow_mode", False):
-                # follower can be on a different system which is rsynced from the leader:
-                dk.data_path = Path(
-                    self.config["user_data_dir"]
-                    / "models"
-                    / dk.data_path.parts[-2]
-                    / dk.data_path.parts[-1]
-                )
 
         if coin in self.meta_data_dictionary:
             dk.data = self.meta_data_dictionary[coin]["meta_data"]
@@ -589,12 +584,16 @@ class FreqaiDataDrawer:
         # try to access model in memory instead of loading object from disk to save time
         if dk.live and coin in self.model_dictionary:
             model = self.model_dictionary[coin]
-        elif not dk.keras:
+        elif self.model_type == 'joblib':
             model = load(dk.data_path / f"{dk.model_filename}_model.joblib")
-        else:
+        elif self.model_type == 'keras':
             from tensorflow import keras
-
             model = keras.models.load_model(dk.data_path / f"{dk.model_filename}_model.h5")
+        elif self.model_type == 'stable_baselines':
+            mod = importlib.import_module(
+                'stable_baselines3', self.freqai_info['rl_config']['model_type'])
+            MODELCLASS = getattr(mod, self.freqai_info['rl_config']['model_type'])
+            model = MODELCLASS.load(dk.data_path / f"{dk.model_filename}_model")
 
         if Path(dk.data_path / f"{dk.model_filename}_svm_model.joblib").is_file():
             dk.svm_model = load(dk.data_path / f"{dk.model_filename}_svm_model.joblib")
@@ -604,6 +603,10 @@ class FreqaiDataDrawer:
                 f"Unable to load model, ensure model exists at " f"{dk.data_path} "
             )
 
+        # load it into ram if it was loaded from disk
+        if coin not in self.model_dictionary:
+            self.model_dictionary[coin] = model
+
         if self.config["freqai"]["feature_parameters"]["principal_component_analysis"]:
             dk.pca = cloudpickle.load(
                 open(dk.data_path / f"{dk.model_filename}_pca_object.pkl", "rb")

freqtrade/freqai/data_kitchen.py

@@ -9,6 +9,7 @@ from typing import Any, Dict, List, Tuple
 import numpy as np
 import numpy.typing as npt
 import pandas as pd
+import psutil
 from pandas import DataFrame, HDFStore
 from scipy import stats
 from sklearn import linear_model
@@ -98,7 +99,10 @@ class FreqaiDataKitchen:
                 )
 
         self.data['extra_returns_per_train'] = self.freqai_config.get('extra_returns_per_train', {})
-        self.thread_count = self.freqai_config.get("data_kitchen_thread_count", -1)
+        if not self.freqai_config.get("data_kitchen_thread_count", 0):
+            self.thread_count = max(int(psutil.cpu_count() * 2 - 2), 1)
+        else:
+            self.thread_count = self.freqai_config["data_kitchen_thread_count"]
         self.train_dates: DataFrame = pd.DataFrame()
         self.unique_classes: Dict[str, list] = {}
         self.unique_class_list: list = []

freqtrade/freqai/freqai_interface.py

@@ -5,15 +5,17 @@ from abc import ABC, abstractmethod
 from collections import deque
 from datetime import datetime, timezone
 from pathlib import Path
-from typing import Any, Dict, List, Literal, Tuple
+from typing import Any, Dict, List, Literal, Optional, Tuple
 
 import numpy as np
 import pandas as pd
+import psutil
 from numpy.typing import NDArray
 from pandas import DataFrame
 
 from freqtrade.configuration import TimeRange
 from freqtrade.constants import Config
+from freqtrade.data.dataprovider import DataProvider
 from freqtrade.enums import RunMode
 from freqtrade.exceptions import OperationalException
 from freqtrade.exchange import timeframe_to_seconds
@@ -101,6 +103,8 @@ class IFreqaiModel(ABC):
         self._threads: List[threading.Thread] = []
         self._stop_event = threading.Event()
         self.metadata = self.dd.load_global_metadata_from_disk()
+        self.data_provider: Optional[DataProvider] = None
+        self.max_system_threads = max(int(psutil.cpu_count() * 2 - 2), 1)
 
         record_params(config, self.full_path)
 
@@ -129,6 +133,7 @@ class IFreqaiModel(ABC):
 
         self.live = strategy.dp.runmode in (RunMode.DRY_RUN, RunMode.LIVE)
         self.dd.set_pair_dict_info(metadata)
+        self.data_provider = strategy.dp
 
         if self.live:
             self.inference_timer('start')
@@ -175,6 +180,13 @@ class IFreqaiModel(ABC):
         self.model = None
         self.dk = None
 
+    def _on_stop(self):
+        """
+        Callback for Subclasses to override to include logic for shutting down resources
+        when SIGINT is sent.
+        """
+        return
+
     def shutdown(self):
         """
         Cleans up threads on Shutdown, set stop event. Join threads to wait
@@ -183,6 +195,9 @@ class IFreqaiModel(ABC):
         logger.info("Stopping FreqAI")
         self._stop_event.set()
 
+        self.data_provider = None
+        self._on_stop()
+
         logger.info("Waiting on Training iteration")
         for _thread in self._threads:
             _thread.join()
@@ -663,7 +678,7 @@ class IFreqaiModel(ABC):
             hist_preds_df['DI_values'] = 0
 
         for return_str in dk.data['extra_returns_per_train']:
-            hist_preds_df[return_str] = 0
+            hist_preds_df[return_str] = dk.data['extra_returns_per_train'][return_str]
 
         hist_preds_df['close_price'] = strat_df['close']
         hist_preds_df['date_pred'] = strat_df['date']

freqtrade/freqai/prediction_models/ReinforcementLearner.py

@@ -0,0 +1,141 @@
+import logging
+from pathlib import Path
+from typing import Any, Dict
+
+import torch as th
+
+from freqtrade.freqai.data_kitchen import FreqaiDataKitchen
+from freqtrade.freqai.RL.Base5ActionRLEnv import Actions, Base5ActionRLEnv, Positions
+from freqtrade.freqai.RL.BaseReinforcementLearningModel import BaseReinforcementLearningModel
+
+
+logger = logging.getLogger(__name__)
+
+
+class ReinforcementLearner(BaseReinforcementLearningModel):
+    """
+    Reinforcement Learning Model prediction model.
+
+    Users can inherit from this class to make their own RL model with custom
+    environment/training controls. Define the file as follows:
+
+    ```
+    from freqtrade.freqai.prediction_models.ReinforcementLearner import ReinforcementLearner
+
+    class MyCoolRLModel(ReinforcementLearner):
+    ```
+
+    Save the file to `user_data/freqaimodels`, then run it with:
+
+    freqtrade trade --freqaimodel MyCoolRLModel --config config.json --strategy SomeCoolStrat
+
+    Here the users can override any of the functions
+    available in the `IFreqaiModel` inheritance tree. Most importantly for RL, this
+    is where the user overrides `MyRLEnv` (see below), to define custom
+    `calculate_reward()` function, or to override any other parts of the environment.
+
+    This class also allows users to override any other part of the IFreqaiModel tree.
+    For example, the user can override `def fit()` or `def train()` or `def predict()`
+    to take fine-tuned control over these processes.
+
+    Another common override may be `def data_cleaning_predict()` where the user can
+    take fine-tuned control over the data handling pipeline.
+    """
+
+    def fit(self, data_dictionary: Dict[str, Any], dk: FreqaiDataKitchen, **kwargs):
+        """
+        User customizable fit method
+        :param data_dictionary: dict = common data dictionary containing all train/test
+            features/labels/weights.
+        :param dk: FreqaiDatakitchen = data kitchen for current pair.
+        :return:
+        model Any = trained model to be used for inference in dry/live/backtesting
+        """
+        train_df = data_dictionary["train_features"]
+        total_timesteps = self.freqai_info["rl_config"]["train_cycles"] * len(train_df)
+
+        policy_kwargs = dict(activation_fn=th.nn.ReLU,
+                             net_arch=self.net_arch)
+
+        if dk.pair not in self.dd.model_dictionary or not self.continual_learning:
+            model = self.MODELCLASS(self.policy_type, self.train_env, policy_kwargs=policy_kwargs,
+                                    tensorboard_log=Path(
+                                        dk.full_path / "tensorboard" / dk.pair.split('/')[0]),
+                                    **self.freqai_info['model_training_parameters']
+                                    )
+        else:
+            logger.info('Continual training activated - starting training from previously '
+                        'trained agent.')
+            model = self.dd.model_dictionary[dk.pair]
+            model.set_env(self.train_env)
+
+        model.learn(
+            total_timesteps=int(total_timesteps),
+            callback=self.eval_callback
+        )
+
+        if Path(dk.data_path / "best_model.zip").is_file():
+            logger.info('Callback found a best model.')
+            best_model = self.MODELCLASS.load(dk.data_path / "best_model")
+            return best_model
+
+        logger.info('Couldnt find best model, using final model instead.')
+
+        return model
+
+    class MyRLEnv(Base5ActionRLEnv):
+        """
+        User can override any function in BaseRLEnv and gym.Env. Here the user
+        sets a custom reward based on profit and trade duration.
+        """
+
+        def calculate_reward(self, action: int) -> float:
+            """
+            An example reward function. This is the one function that users will likely
+            wish to inject their own creativity into.
+            :param action: int = The action made by the agent for the current candle.
+            :return:
+            float = the reward to give to the agent for current step (used for optimization
+                of weights in NN)
+            """
+            # first, penalize if the action is not valid
+            if not self._is_valid(action):
+                return -2
+
+            pnl = self.get_unrealized_profit()
+            factor = 100.
+
+            # reward agent for entering trades
+            if (action in (Actions.Long_enter.value, Actions.Short_enter.value)
+                    and self._position == Positions.Neutral):
+                return 25
+            # discourage agent from not entering trades
+            if action == Actions.Neutral.value and self._position == Positions.Neutral:
+                return -1
+
+            max_trade_duration = self.rl_config.get('max_trade_duration_candles', 300)
+            trade_duration = self._current_tick - self._last_trade_tick  # type: ignore
+
+            if trade_duration <= max_trade_duration:
+                factor *= 1.5
+            elif trade_duration > max_trade_duration:
+                factor *= 0.5
+
+            # discourage sitting in position
+            if (self._position in (Positions.Short, Positions.Long) and
+                    action == Actions.Neutral.value):
+                return -1 * trade_duration / max_trade_duration
+
+            # close long
+            if action == Actions.Long_exit.value and self._position == Positions.Long:
+                if pnl > self.profit_aim * self.rr:
+                    factor *= self.rl_config['model_reward_parameters'].get('win_reward_factor', 2)
+                return float(pnl * factor)
+
+            # close short
+            if action == Actions.Short_exit.value and self._position == Positions.Short:
+                if pnl > self.profit_aim * self.rr:
+                    factor *= self.rl_config['model_reward_parameters'].get('win_reward_factor', 2)
+                return float(pnl * factor)
+
+            return 0.

freqtrade/freqai/prediction_models/ReinforcementLearner_multiproc.py

@@ -0,0 +1,51 @@
+import logging
+from typing import Any, Dict  # , Tuple
+
+# import numpy.typing as npt
+from pandas import DataFrame
+from stable_baselines3.common.callbacks import EvalCallback
+from stable_baselines3.common.vec_env import SubprocVecEnv
+
+from freqtrade.freqai.data_kitchen import FreqaiDataKitchen
+from freqtrade.freqai.prediction_models.ReinforcementLearner import ReinforcementLearner
+from freqtrade.freqai.RL.BaseReinforcementLearningModel import make_env
+
+
+logger = logging.getLogger(__name__)
+
+
+class ReinforcementLearner_multiproc(ReinforcementLearner):
+    """
+    Demonstration of how to build vectorized environments
+    """
+
+    def set_train_and_eval_environments(self, data_dictionary: Dict[str, Any],
+                                        prices_train: DataFrame, prices_test: DataFrame,
+                                        dk: FreqaiDataKitchen):
+        """
+        User can override this if they are using a custom MyRLEnv
+        :param data_dictionary: dict = common data dictionary containing train and test
+            features/labels/weights.
+        :param prices_train/test: DataFrame = dataframe comprised of the prices to be used in
+            the environment during training
+        or testing
+        :param dk: FreqaiDataKitchen = the datakitchen for the current pair
+        """
+        train_df = data_dictionary["train_features"]
+        test_df = data_dictionary["test_features"]
+
+        env_id = "train_env"
+        self.train_env = SubprocVecEnv([make_env(self.MyRLEnv, env_id, i, 1, train_df, prices_train,
+                                        self.reward_params, self.CONV_WIDTH, monitor=True,
+                                        config=self.config) for i
+                                        in range(self.max_threads)])
+
+        eval_env_id = 'eval_env'
+        self.eval_env = SubprocVecEnv([make_env(self.MyRLEnv, eval_env_id, i, 1,
+                                                test_df, prices_test,
+                                                self.reward_params, self.CONV_WIDTH, monitor=True,
+                                                config=self.config) for i
+                                       in range(self.max_threads)])
+        self.eval_callback = EvalCallback(self.eval_env, deterministic=True,
+                                          render=False, eval_freq=len(train_df),
+                                          best_model_save_path=str(dk.data_path))

freqtrade/freqtradebot.py

@@ -191,10 +191,10 @@ class FreqtradeBot(LoggingMixin):
         # Check whether markets have to be reloaded and reload them when it's needed
         self.exchange.reload_markets()
 
-        self.update_closed_trades_without_assigned_fees()
+        self.update_trades_without_assigned_fees()
 
         # Query trades from persistence layer
-        trades = Trade.get_open_trades()
+        trades: List[Trade] = Trade.get_open_trades()
 
         self.active_pair_whitelist = self._refresh_active_whitelist(trades)
 
@@ -354,7 +354,7 @@ class FreqtradeBot(LoggingMixin):
         if self.trading_mode == TradingMode.FUTURES:
             self._schedule.run_pending()
 
-    def update_closed_trades_without_assigned_fees(self) -> None:
+    def update_trades_without_assigned_fees(self) -> None:
         """
         Update closed trades without close fees assigned.
         Only acts when Orders are in the database, otherwise the last order-id is unknown.
@@ -381,15 +381,16 @@ class FreqtradeBot(LoggingMixin):
 
         trades = Trade.get_open_trades_without_assigned_fees()
         for trade in trades:
-            if trade.is_open and not trade.fee_updated(trade.entry_side):
+            with self._exit_lock:
-                order = trade.select_order(trade.entry_side, False)
+                if trade.is_open and not trade.fee_updated(trade.entry_side):
-                open_order = trade.select_order(trade.entry_side, True)
+                    order = trade.select_order(trade.entry_side, False)
-                if order and open_order is None:
+                    open_order = trade.select_order(trade.entry_side, True)
-                    logger.info(
+                    if order and open_order is None:
-                        f"Updating {trade.entry_side}-fee on trade {trade}"
+                        logger.info(
-                        f"for order {order.order_id}."
+                            f"Updating {trade.entry_side}-fee on trade {trade}"
-                    )
+                            f"for order {order.order_id}."
-                    self.update_trade_state(trade, order.order_id, send_msg=False)
+                        )
+                        self.update_trade_state(trade, order.order_id, send_msg=False)
 
     def handle_insufficient_funds(self, trade: Trade):
         """
@@ -826,6 +827,8 @@ class FreqtradeBot(LoggingMixin):
                 co = self.exchange.cancel_stoploss_order_with_result(
                     trade.stoploss_order_id, trade.pair, trade.amount)
                 trade.update_order(co)
+                # Reset stoploss order id.
+                trade.stoploss_order_id = None
             except InvalidOrderException:
                 logger.exception(f"Could not cancel stoploss order {trade.stoploss_order_id}")
         return trade
@@ -982,7 +985,7 @@ class FreqtradeBot(LoggingMixin):
 # SELL / exit positions / close trades logic and methods
 #
 
-    def exit_positions(self, trades: List[Any]) -> int:
+    def exit_positions(self, trades: List[Trade]) -> int:
         """
         Tries to execute exit orders for open trades (positions)
         """
@@ -1010,7 +1013,7 @@ class FreqtradeBot(LoggingMixin):
 
     def handle_trade(self, trade: Trade) -> bool:
         """
-        Sells/exits_short the current pair if the threshold is reached and updates the trade record.
+        Exits the current pair if the threshold is reached and updates the trade record.
         :return: True if trade has been sold/exited_short, False otherwise
         """
         if not trade.is_open:
@@ -1148,7 +1151,7 @@ class FreqtradeBot(LoggingMixin):
             stoploss = (
                 self.edge.stoploss(pair=trade.pair)
                 if self.edge else
-                self.strategy.stoploss / trade.leverage
+                trade.stop_loss_pct / trade.leverage
             )
             if trade.is_short:
                 stop_price = trade.open_rate * (1 - stoploss)
@@ -1167,7 +1170,6 @@ class FreqtradeBot(LoggingMixin):
             if self.create_stoploss_order(trade=trade, stop_price=trade.stoploss_or_liquidation):
                 return False
             else:
-                trade.stoploss_order_id = None
                 logger.warning('Stoploss order was cancelled, but unable to recreate one.')
 
         # Finally we check if stoploss on exchange should be moved up because of trailing.

freqtrade/optimize/backtesting.py

@@ -692,10 +692,11 @@ class Backtesting:
         trade.orders.append(order)
         return trade
 
-    def _get_exit_trade_entry(self, trade: LocalTrade, row: Tuple) -> Optional[LocalTrade]:
+    def _get_exit_trade_entry(
+            self, trade: LocalTrade, row: Tuple, is_first: bool) -> Optional[LocalTrade]:
         exit_candle_time: datetime = row[DATE_IDX].to_pydatetime()
 
-        if self.trading_mode == TradingMode.FUTURES:
+        if is_first and self.trading_mode == TradingMode.FUTURES:
             trade.funding_fees = self.exchange.calculate_funding_fees(
                 self.futures_data[trade.pair],
                 amount=trade.amount,
@@ -704,32 +705,7 @@ class Backtesting:
                 close_date=exit_candle_time,
             )
 
-        if self.timeframe_detail and trade.pair in self.detail_data:
+        return self._get_exit_trade_entry_for_candle(trade, row)
-            exit_candle_end = exit_candle_time + timedelta(minutes=self.timeframe_min)
-
-            detail_data = self.detail_data[trade.pair]
-            detail_data = detail_data.loc[
-                (detail_data['date'] >= exit_candle_time) &
-                (detail_data['date'] < exit_candle_end)
-            ].copy()
-            if len(detail_data) == 0:
-                # Fall back to "regular" data if no detail data was found for this candle
-                return self._get_exit_trade_entry_for_candle(trade, row)
-            detail_data.loc[:, 'enter_long'] = row[LONG_IDX]
-            detail_data.loc[:, 'exit_long'] = row[ELONG_IDX]
-            detail_data.loc[:, 'enter_short'] = row[SHORT_IDX]
-            detail_data.loc[:, 'exit_short'] = row[ESHORT_IDX]
-            detail_data.loc[:, 'enter_tag'] = row[ENTER_TAG_IDX]
-            detail_data.loc[:, 'exit_tag'] = row[EXIT_TAG_IDX]
-            for det_row in detail_data[HEADERS].values.tolist():
-                res = self._get_exit_trade_entry_for_candle(trade, det_row)
-                if res:
-                    return res
-
-            return None
-
-        else:
-            return self._get_exit_trade_entry_for_candle(trade, row)
 
     def get_valid_price_and_stake(
         self, pair: str, row: Tuple, propose_rate: float, stake_amount: float,
@@ -1074,7 +1050,7 @@ class Backtesting:
 
     def backtest_loop(
             self, row: Tuple, pair: str, current_time: datetime, end_date: datetime,
-            max_open_trades: int, open_trade_count_start: int) -> int:
+            max_open_trades: int, open_trade_count_start: int, is_first: bool = True) -> int:
         """
         NOTE: This method is used by Hyperopt at each iteration. Please keep it optimized.
 
@@ -1092,9 +1068,11 @@ class Backtesting:
         # without positionstacking, we can only have one open trade per pair.
         # max_open_trades must be respected
         # don't open on the last row
+        # We only open trades on the main candle, not on detail candles
         trade_dir = self.check_for_trade_entry(row)
         if (
             (self._position_stacking or len(LocalTrade.bt_trades_open_pp[pair]) == 0)
+            and is_first
             and self.trade_slot_available(max_open_trades, open_trade_count_start)
             and current_time != end_date
             and trade_dir is not None
@@ -1120,7 +1098,7 @@ class Backtesting:
 
                 # 4. Create exit orders (if any)
             if not trade.open_order_id:
-                self._get_exit_trade_entry(trade, row)  # Place exit order if necessary
+                self._get_exit_trade_entry(trade, row, is_first)  # Place exit order if necessary
 
                 # 5. Process exit orders.
             order = trade.select_order(trade.exit_side, is_open=True)
@@ -1171,7 +1149,6 @@ class Backtesting:
 
         self.progress.init_step(BacktestState.BACKTEST, int(
             (end_date - start_date) / timedelta(minutes=self.timeframe_min)))
-
         # Loop timerange and get candle for each pair at that point in time
         while current_time <= end_date:
             open_trade_count_start = LocalTrade.bt_open_open_trade_count
@@ -1185,9 +1162,37 @@ class Backtesting:
                 row_index += 1
                 indexes[pair] = row_index
                 self.dataprovider._set_dataframe_max_index(row_index)
+                current_detail_time: datetime = row[DATE_IDX].to_pydatetime()
+                if self.timeframe_detail and pair in self.detail_data:
+                    exit_candle_end = current_detail_time + timedelta(minutes=self.timeframe_min)
 
-                open_trade_count_start = self.backtest_loop(
+                    detail_data = self.detail_data[pair]
-                    row, pair, current_time, end_date, max_open_trades, open_trade_count_start)
+                    detail_data = detail_data.loc[
+                        (detail_data['date'] >= current_detail_time) &
+                        (detail_data['date'] < exit_candle_end)
+                    ].copy()
+                    if len(detail_data) == 0:
+                        # Fall back to "regular" data if no detail data was found for this candle
+                        open_trade_count_start = self.backtest_loop(
+                            row, pair, current_time, end_date, max_open_trades,
+                            open_trade_count_start)
+                    detail_data.loc[:, 'enter_long'] = row[LONG_IDX]
+                    detail_data.loc[:, 'exit_long'] = row[ELONG_IDX]
+                    detail_data.loc[:, 'enter_short'] = row[SHORT_IDX]
+                    detail_data.loc[:, 'exit_short'] = row[ESHORT_IDX]
+                    detail_data.loc[:, 'enter_tag'] = row[ENTER_TAG_IDX]
+                    detail_data.loc[:, 'exit_tag'] = row[EXIT_TAG_IDX]
+                    is_first = True
+                    current_time_det = current_time
+                    for det_row in detail_data[HEADERS].values.tolist():
+                        open_trade_count_start = self.backtest_loop(
+                            det_row, pair, current_time_det, end_date, max_open_trades,
+                            open_trade_count_start, is_first)
+                        current_time_det += timedelta(minutes=self.timeframe_detail_min)
+                        is_first = False
+                else:
+                    open_trade_count_start = self.backtest_loop(
+                        row, pair, current_time, end_date, max_open_trades, open_trade_count_start)
 
             # Move time one configured time_interval ahead.
             self.progress.increment()

freqtrade/optimize/hyperopt_tools.py

@@ -17,6 +17,7 @@ from freqtrade.enums import HyperoptState
 from freqtrade.exceptions import OperationalException
 from freqtrade.misc import deep_merge_dicts, round_coin_value, round_dict, safe_value_fallback2
 from freqtrade.optimize.hyperopt_epoch_filters import hyperopt_filter_epochs
+from freqtrade.optimize.optimize_reports import generate_wins_draws_losses
 
 
 logger = logging.getLogger(__name__)
@@ -325,8 +326,10 @@ class HyperoptTools():
 
         # New mode, using backtest result for metrics
         trials['results_metrics.winsdrawslosses'] = trials.apply(
-            lambda x: f"{x['results_metrics.wins']} {x['results_metrics.draws']:>4} "
+            lambda x: generate_wins_draws_losses(
-                      f"{x['results_metrics.losses']:>4}", axis=1)
+                            x['results_metrics.wins'], x['results_metrics.draws'],
+                            x['results_metrics.losses']
+                      ), axis=1)
 
         trials = trials[['Best', 'current_epoch', 'results_metrics.total_trades',
                          'results_metrics.winsdrawslosses',
@@ -337,7 +340,7 @@ class HyperoptTools():
                          'loss', 'is_initial_point', 'is_random', 'is_best']]
 
         trials.columns = [
-            'Best', 'Epoch', 'Trades', ' Win Draw Loss', 'Avg profit',
+            'Best', 'Epoch', 'Trades', ' Win  Draw  Loss  Win%', 'Avg profit',
             'Total profit', 'Profit', 'Avg duration', 'max_drawdown', 'max_drawdown_account',
             'max_drawdown_abs', 'Objective', 'is_initial_point', 'is_random', 'is_best'
             ]
@@ -467,9 +470,9 @@ class HyperoptTools():
 
         base_metrics = ['Best', 'current_epoch', 'results_metrics.total_trades',
                         'results_metrics.profit_mean', 'results_metrics.profit_median',
-                        'results_metrics.profit_total',
+                        'results_metrics.profit_total', 'Stake currency',
-                        'Stake currency',
                         'results_metrics.profit_total_abs', 'results_metrics.holding_avg',
+                        'results_metrics.trade_count_long', 'results_metrics.trade_count_short',
                         'loss', 'is_initial_point', 'is_best']
         perc_multi = 100
 
@@ -477,7 +480,9 @@ class HyperoptTools():
         trials = trials[base_metrics + param_metrics]
 
         base_columns = ['Best', 'Epoch', 'Trades', 'Avg profit', 'Median profit', 'Total profit',
-                        'Stake currency', 'Profit', 'Avg duration', 'Objective',
+                        'Stake currency', 'Profit', 'Avg duration',
+                        'Trade count long', 'Trade count short',
+                        'Objective',
                         'is_initial_point', 'is_best']
         param_columns = list(results[0]['params_dict'].keys())
         trials.columns = base_columns + param_columns

freqtrade/optimize/optimize_reports.py

@@ -86,7 +86,7 @@ def _get_line_header(first_column: str, stake_currency: str,
             'Win  Draw  Loss  Win%']
 
 
-def _generate_wins_draws_losses(wins, draws, losses):
+def generate_wins_draws_losses(wins, draws, losses):
     if wins > 0 and losses == 0:
         wl_ratio = '100'
     elif wins == 0:
@@ -600,7 +600,7 @@ def text_table_bt_results(pair_results: List[Dict[str, Any]], stake_currency: st
     output = [[
         t['key'], t['trades'], t['profit_mean_pct'], t['profit_sum_pct'], t['profit_total_abs'],
         t['profit_total_pct'], t['duration_avg'],
-        _generate_wins_draws_losses(t['wins'], t['draws'], t['losses'])
+        generate_wins_draws_losses(t['wins'], t['draws'], t['losses'])
     ] for t in pair_results]
     # Ignore type as floatfmt does allow tuples but mypy does not know that
     return tabulate(output, headers=headers,
@@ -626,7 +626,7 @@ def text_table_exit_reason(exit_reason_stats: List[Dict[str, Any]], stake_curren
 
     output = [[
         t.get('exit_reason', t.get('sell_reason')), t['trades'],
-        _generate_wins_draws_losses(t['wins'], t['draws'], t['losses']),
+        generate_wins_draws_losses(t['wins'], t['draws'], t['losses']),
         t['profit_mean_pct'], t['profit_sum_pct'],
         round_coin_value(t['profit_total_abs'], stake_currency, False),
         t['profit_total_pct'],
@@ -656,7 +656,7 @@ def text_table_tags(tag_type: str, tag_results: List[Dict[str, Any]], stake_curr
             t['profit_total_abs'],
             t['profit_total_pct'],
             t['duration_avg'],
-            _generate_wins_draws_losses(
+            generate_wins_draws_losses(
                 t['wins'],
                 t['draws'],
                 t['losses'])] for t in tag_results]
@@ -715,7 +715,7 @@ def text_table_strategy(strategy_results, stake_currency: str) -> str:
     output = [[
         t['key'], t['trades'], t['profit_mean_pct'], t['profit_sum_pct'], t['profit_total_abs'],
         t['profit_total_pct'], t['duration_avg'],
-        _generate_wins_draws_losses(t['wins'], t['draws'], t['losses']), drawdown]
+        generate_wins_draws_losses(t['wins'], t['draws'], t['losses']), drawdown]
         for t, drawdown in zip(strategy_results, drawdown)]
     # Ignore type as floatfmt does allow tuples but mypy does not know that
     return tabulate(output, headers=headers,

freqtrade/persistence/pairlock_middleware.py

@@ -87,7 +87,7 @@ class PairLocks():
         Get the lock that expires the latest for the pair given.
         """
         locks = PairLocks.get_pair_locks(pair, now, side=side)
-        locks = sorted(locks, key=lambda l: l.lock_end_time, reverse=True)
+        locks = sorted(locks, key=lambda lock: lock.lock_end_time, reverse=True)
         return locks[0] if locks else None
 
     @staticmethod

freqtrade/rpc/api_server/api_auth.py

@@ -81,8 +81,6 @@ async def validate_ws_token(
     except HTTPException:
         pass
 
-    # No checks passed, deny the connection
-    logger.debug("Denying websocket request.")
     # If it doesn't match, close the websocket connection
     await ws.close(code=status.WS_1008_POLICY_VIOLATION)
 

freqtrade/rpc/api_server/api_ws.py

@@ -1,16 +1,16 @@
 import logging
+import time
 from typing import Any, Dict
 
-from fastapi import APIRouter, Depends, WebSocketDisconnect
+from fastapi import APIRouter, Depends
-from fastapi.websockets import WebSocket, WebSocketState
+from fastapi.websockets import WebSocket
 from pydantic import ValidationError
-from websockets.exceptions import WebSocketException
 
 from freqtrade.enums import RPCMessageType, RPCRequestType
 from freqtrade.rpc.api_server.api_auth import validate_ws_token
-from freqtrade.rpc.api_server.deps import get_channel_manager, get_rpc
+from freqtrade.rpc.api_server.deps import get_message_stream, get_rpc
-from freqtrade.rpc.api_server.ws import WebSocketChannel
+from freqtrade.rpc.api_server.ws.channel import WebSocketChannel, create_channel
-from freqtrade.rpc.api_server.ws.channel import ChannelManager
+from freqtrade.rpc.api_server.ws.message_stream import MessageStream
 from freqtrade.rpc.api_server.ws_schemas import (WSAnalyzedDFMessage, WSMessageSchema,
                                                  WSRequestSchema, WSWhitelistMessage)
 from freqtrade.rpc.rpc import RPC
@@ -22,23 +22,35 @@ logger = logging.getLogger(__name__)
 router = APIRouter()
 
 
-async def is_websocket_alive(ws: WebSocket) -> bool:
+async def channel_reader(channel: WebSocketChannel, rpc: RPC):
     """
-    Check if a FastAPI Websocket is still open
+    Iterate over the messages from the channel and process the request
     """
-    if (
+    async for message in channel:
-        ws.application_state == WebSocketState.CONNECTED and
+        await _process_consumer_request(message, channel, rpc)
-        ws.client_state == WebSocketState.CONNECTED
+
-    ):
+
-        return True
+async def channel_broadcaster(channel: WebSocketChannel, message_stream: MessageStream):
-    return False
+    """
+    Iterate over messages in the message stream and send them
+    """
+    async for message, ts in message_stream:
+        if channel.subscribed_to(message.get('type')):
+            # Log a warning if this channel is behind
+            # on the message stream by a lot
+            if (time.time() - ts) > 60:
+                logger.warning(f"Channel {channel} is behind MessageStream by 1 minute,"
+                               " this can cause a memory leak if you see this message"
+                               " often, consider reducing pair list size or amount of"
+                               " consumers.")
+
+            await channel.send(message, timeout=True)
 
 
 async def _process_consumer_request(
     request: Dict[str, Any],
     channel: WebSocketChannel,
-    rpc: RPC,
+    rpc: RPC
-    channel_manager: ChannelManager
 ):
     """
     Validate and handle a request from a websocket consumer
@@ -74,65 +86,29 @@ async def _process_consumer_request(
 
         # Format response
         response = WSWhitelistMessage(data=whitelist)
-        # Send it back
+        await channel.send(response.dict(exclude_none=True))
-        await channel_manager.send_direct(channel, response.dict(exclude_none=True))
 
     elif type == RPCRequestType.ANALYZED_DF:
-        limit = None
+        # Limit the amount of candles per dataframe to 'limit' or 1500
-
+        limit = min(data.get('limit', 1500), 1500) if data else None
-        if data:
-            # Limit the amount of candles per dataframe to 'limit' or 1500
-            limit = max(data.get('limit', 1500), 1500)
 
         # For every pair in the generator, send a separate message
         for message in rpc._ws_request_analyzed_df(limit):
+            # Format response
             response = WSAnalyzedDFMessage(data=message)
-            await channel_manager.send_direct(channel, response.dict(exclude_none=True))
+            await channel.send(response.dict(exclude_none=True))
 
 
 @router.websocket("/message/ws")
 async def message_endpoint(
-    ws: WebSocket,
+    websocket: WebSocket,
+    token: str = Depends(validate_ws_token),
     rpc: RPC = Depends(get_rpc),
-    channel_manager=Depends(get_channel_manager),
+    message_stream: MessageStream = Depends(get_message_stream)
-    token: str = Depends(validate_ws_token)
 ):
-    """
+    if token:
-    Message WebSocket endpoint, facilitates sending RPC messages
+        async with create_channel(websocket) as channel:
-    """
+            await channel.run_channel_tasks(
-    try:
+                channel_reader(channel, rpc),
-        channel = await channel_manager.on_connect(ws)
+                channel_broadcaster(channel, message_stream)
-        if await is_websocket_alive(ws):
+            )
-
-            logger.info(f"Consumer connected - {channel}")
-
-            # Keep connection open until explicitly closed, and process requests
-            try:
-                while not channel.is_closed():
-                    request = await channel.recv()
-
-                    # Process the request here
-                    await _process_consumer_request(request, channel, rpc, channel_manager)
-
-            except (WebSocketDisconnect, WebSocketException):
-                # Handle client disconnects
-                logger.info(f"Consumer disconnected - {channel}")
-            except RuntimeError:
-                # Handle cases like -
-                # RuntimeError('Cannot call "send" once a closed message has been sent')
-                pass
-            except Exception as e:
-                logger.info(f"Consumer connection failed - {channel}: {e}")
-                logger.debug(e, exc_info=e)
-
-    except RuntimeError:
-        # WebSocket was closed
-        # Do nothing
-        pass
-    except Exception as e:
-        logger.error(f"Failed to serve - {ws.client}")
-        # Log tracebacks to keep track of what errors are happening
-        logger.exception(e)
-    finally:
-        if channel:
-            await channel_manager.on_disconnect(ws)

freqtrade/rpc/api_server/deps.py

@@ -41,8 +41,8 @@ def get_exchange(config=Depends(get_config)):
     return ApiServer._exchange
 
 
-def get_channel_manager():
+def get_message_stream():
-    return ApiServer._ws_channel_manager
+    return ApiServer._message_stream
 
 
 def is_webserver_mode(config=Depends(get_config)):

freqtrade/rpc/api_server/webserver.py

@@ -1,22 +1,17 @@
-import asyncio
 import logging
 from ipaddress import IPv4Address
-from threading import Thread
 from typing import Any, Dict, Optional
 
 import orjson
 import uvicorn
 from fastapi import Depends, FastAPI
 from fastapi.middleware.cors import CORSMiddleware
-# Look into alternatives
-from janus import Queue as ThreadedQueue
 from starlette.responses import JSONResponse
 
 from freqtrade.constants import Config
 from freqtrade.exceptions import OperationalException
 from freqtrade.rpc.api_server.uvicorn_threaded import UvicornServer
-from freqtrade.rpc.api_server.ws import ChannelManager
+from freqtrade.rpc.api_server.ws.message_stream import MessageStream
-from freqtrade.rpc.api_server.ws_schemas import WSMessageSchemaType
 from freqtrade.rpc.rpc import RPC, RPCException, RPCHandler
 
 
@@ -50,10 +45,8 @@ class ApiServer(RPCHandler):
     _config: Config = {}
     # Exchange - only available in webserver mode.
     _exchange = None
-    # websocket message queue stuff
+    # websocket message stuff
-    _ws_channel_manager: ChannelManager
+    _message_stream: Optional[MessageStream] = None
-    _ws_thread = None
-    _ws_loop: Optional[asyncio.AbstractEventLoop] = None
 
     def __new__(cls, *args, **kwargs):
         """
@@ -71,15 +64,11 @@ class ApiServer(RPCHandler):
             return
         self._standalone: bool = standalone
         self._server = None
-        self._ws_queue: Optional[ThreadedQueue] = None
-        self._ws_background_task = None
 
         ApiServer.__initialized = True
 
         api_config = self._config['api_server']
 
-        ApiServer._ws_channel_manager = ChannelManager()
-
         self.app = FastAPI(title="Freqtrade API",
                            docs_url='/docs' if api_config.get('enable_openapi', False) else None,
                            redoc_url=None,
@@ -105,21 +94,9 @@ class ApiServer(RPCHandler):
         del ApiServer._rpc
         if self._server and not self._standalone:
             logger.info("Stopping API Server")
+            # self._server.force_exit, self._server.should_exit = True, True
             self._server.cleanup()
 
-        if self._ws_thread and self._ws_loop:
-            logger.info("Stopping API Server background tasks")
-
-            if self._ws_background_task:
-                # Cancel the queue task
-                self._ws_background_task.cancel()
-
-            self._ws_thread.join()
-
-        self._ws_thread = None
-        self._ws_loop = None
-        self._ws_background_task = None
-
     @classmethod
     def shutdown(cls):
         cls.__initialized = False
@@ -129,9 +106,11 @@ class ApiServer(RPCHandler):
         cls._rpc = None
 
     def send_msg(self, msg: Dict[str, Any]) -> None:
-        if self._ws_queue:
+        """
-            sync_q = self._ws_queue.sync_q
+        Publish the message to the message stream
-            sync_q.put(msg)
+        """
+        if ApiServer._message_stream:
+            ApiServer._message_stream.publish(msg)
 
     def handle_rpc_exception(self, request, exc):
         logger.exception(f"API Error calling: {exc}")
@@ -170,54 +149,30 @@ class ApiServer(RPCHandler):
         )
 
         app.add_exception_handler(RPCException, self.handle_rpc_exception)
+        app.add_event_handler(
+            event_type="startup",
+            func=self._api_startup_event
+        )
+        app.add_event_handler(
+            event_type="shutdown",
+            func=self._api_shutdown_event
+        )
 
-    def start_message_queue(self):
+    async def _api_startup_event(self):
-        if self._ws_thread:
+        """
-            return
+        Creates the MessageStream class on startup
+        so it has access to the same event loop
+        as uvicorn
+        """
+        if not ApiServer._message_stream:
+            ApiServer._message_stream = MessageStream()
 
-        # Create a new loop, as it'll be just for the background thread
+    async def _api_shutdown_event(self):
-        self._ws_loop = asyncio.new_event_loop()
+        """
-
+        Removes the MessageStream class on shutdown
-        # Start the thread
+        """
-        self._ws_thread = Thread(target=self._ws_loop.run_forever)
+        if ApiServer._message_stream:
-        self._ws_thread.start()
+            ApiServer._message_stream = None
-
-        # Finally, submit the coro to the thread
-        self._ws_background_task = asyncio.run_coroutine_threadsafe(
-            self._broadcast_queue_data(), loop=self._ws_loop)
-
-    async def _broadcast_queue_data(self) -> None:
-        # Instantiate the queue in this coroutine so it's attached to our loop
-        self._ws_queue = ThreadedQueue()
-        async_queue = self._ws_queue.async_q
-
-        try:
-            while True:
-                logger.debug("Getting queue messages...")
-                if (qsize := async_queue.qsize()) > 20:
-                    # If the queue becomes too big for too long, this may indicate a problem.
-                    logger.warning(f"Queue size now {qsize}")
-                # Get data from queue
-                message: WSMessageSchemaType = await async_queue.get()
-                logger.debug(f"Found message of type: {message.get('type')}")
-                async_queue.task_done()
-                # Broadcast it
-                await self._ws_channel_manager.broadcast(message)
-        except asyncio.CancelledError:
-            pass
-
-        # For testing, shouldn't happen when stable
-        except Exception as e:
-            logger.exception(f"Exception happened in background task: {e}")
-
-        finally:
-            # Disconnect channels and stop the loop on cancel
-            await self._ws_channel_manager.disconnect_all()
-            if self._ws_loop:
-                self._ws_loop.stop()
-            # Avoid adding more items to the queue if they aren't
-            # going to get broadcasted.
-            self._ws_queue = None
 
     def start_api(self):
         """
@@ -257,7 +212,6 @@ class ApiServer(RPCHandler):
             if self._standalone:
                 self._server.run()
             else:
-                self.start_message_queue()
                 self._server.run_in_thread()
         except Exception:
             logger.exception("Api server failed to start.")

freqtrade/rpc/api_server/ws/__init__.py

@@ -3,4 +3,5 @@
 from freqtrade.rpc.api_server.ws.types import WebSocketType
 from freqtrade.rpc.api_server.ws.proxy import WebSocketProxy
 from freqtrade.rpc.api_server.ws.serializer import HybridJSONWebSocketSerializer
-from freqtrade.rpc.api_server.ws.channel import ChannelManager, WebSocketChannel
+from freqtrade.rpc.api_server.ws.channel import WebSocketChannel
+from freqtrade.rpc.api_server.ws.message_stream import MessageStream

freqtrade/rpc/api_server/ws/channel.py

@@ -1,11 +1,13 @@
 import asyncio
 import logging
 import time
-from threading import RLock
+from collections import deque
-from typing import Any, Dict, List, Optional, Type, Union
+from contextlib import asynccontextmanager
+from typing import Any, AsyncIterator, Deque, Dict, List, Optional, Type, Union
 from uuid import uuid4
 
-from fastapi import WebSocket as FastAPIWebSocket
+from fastapi import WebSocketDisconnect
+from websockets.exceptions import ConnectionClosed
 
 from freqtrade.rpc.api_server.ws.proxy import WebSocketProxy
 from freqtrade.rpc.api_server.ws.serializer import (HybridJSONWebSocketSerializer,
@@ -21,31 +23,27 @@ class WebSocketChannel:
     """
     Object to help facilitate managing a websocket connection
     """
-
     def __init__(
         self,
         websocket: WebSocketType,
         channel_id: Optional[str] = None,
-        drain_timeout: int = 3,
-        throttle: float = 0.01,
         serializer_cls: Type[WebSocketSerializer] = HybridJSONWebSocketSerializer
     ):
-
         self.channel_id = channel_id if channel_id else uuid4().hex[:8]
-
-        # The WebSocket object
         self._websocket = WebSocketProxy(websocket)
 
-        self.drain_timeout = drain_timeout
-        self.throttle = throttle
-
-        self._subscriptions: List[str] = []
-        # 32 is the size of the receiving queue in websockets package
-        self.queue: asyncio.Queue[Dict[str, Any]] = asyncio.Queue(maxsize=32)
-        self._relay_task = asyncio.create_task(self.relay())
-
         # Internal event to signify a closed websocket
         self._closed = asyncio.Event()
+        # The async tasks created for the channel
+        self._channel_tasks: List[asyncio.Task] = []
+
+        # Deque for average send times
+        self._send_times: Deque[float] = deque([], maxlen=10)
+        # High limit defaults to 3 to start
+        self._send_high_limit = 3
+
+        # The subscribed message types
+        self._subscriptions: List[str] = []
 
         # Wrap the WebSocket in the Serializing class
         self._wrapped_ws = serializer_cls(self._websocket)
@@ -61,43 +59,58 @@ class WebSocketChannel:
     def remote_addr(self):
         return self._websocket.remote_addr
 
-    async def _send(self, data):
+    @property
-        """
+    def avg_send_time(self):
-        Send data on the wrapped websocket
+        return sum(self._send_times) / len(self._send_times)
-        """
-        await self._wrapped_ws.send(data)
 
-    async def send(self, data) -> bool:
+    def _calc_send_limit(self):
         """
-        Add the data to the queue to be sent.
+        Calculate the send high limit for this channel
-        :returns: True if data added to queue, False otherwise
         """
 
-        # This block only runs if the queue is full, it will wait
+        # Only update if we have enough data
-        # until self.drain_timeout for the relay to drain the outgoing queue
+        if len(self._send_times) == self._send_times.maxlen:
-        # We can't use asyncio.wait_for here because the queue may have been created with a
+            # At least 1s or twice the average of send times, with a
-        # different eventloop
+            # maximum of 3 seconds per message
-        if not self.is_closed():
+            self._send_high_limit = min(max(self.avg_send_time * 2, 1), 3)
-            start = time.time()
-            while self.queue.full():
-                await asyncio.sleep(1)
-                if (time.time() - start) > self.drain_timeout:
-                    return False
 
-            # If for some reason the queue is still full, just return False
+    async def send(
-            try:
+        self,
-                self.queue.put_nowait(data)
+        message: Union[WSMessageSchemaType, Dict[str, Any]],
-            except asyncio.QueueFull:
+        timeout: bool = False
-                return False
+    ):
+        """
+        Send a message on the wrapped websocket. If the sending
+        takes too long, it will raise a TimeoutError and
+        disconnect the connection.
 
-            # If we got here everything is ok
+        :param message: The message to send
-            return True
+        :param timeout: Enforce send high limit, defaults to False
-        else:
+        """
-            return False
+        try:
+            _ = time.time()
+            # If the send times out, it will raise
+            # a TimeoutError and bubble up to the
+            # message_endpoint to close the connection
+            await asyncio.wait_for(
+                self._wrapped_ws.send(message),
+                timeout=self._send_high_limit if timeout else None
+            )
+            total_time = time.time() - _
+            self._send_times.append(total_time)
+
+            self._calc_send_limit()
+        except asyncio.TimeoutError:
+            logger.info(f"Connection for {self} timed out, disconnecting")
+            raise
+
+        # Explicitly give control back to event loop as
+        # websockets.send does not
+        await asyncio.sleep(0.01)
 
     async def recv(self):
         """
-        Receive data on the wrapped websocket
+        Receive a message on the wrapped websocket
         """
         return await self._wrapped_ws.recv()
 
@@ -107,17 +120,27 @@ class WebSocketChannel:
         """
         return await self._websocket.ping()
 
+    async def accept(self):
+        """
+        Accept the underlying websocket connection,
+        if the connection has been closed before we can
+        accept, just close the channel.
+        """
+        try:
+            return await self._websocket.accept()
+        except RuntimeError:
+            await self.close()
+
     async def close(self):
         """
         Close the WebSocketChannel
         """
 
         self._closed.set()
-        self._relay_task.cancel()
 
         try:
-            await self.raw_websocket.close()
+            await self._websocket.close()
-        except Exception:
+        except RuntimeError:
             pass
 
     def is_closed(self) -> bool:
@@ -142,99 +165,76 @@ class WebSocketChannel:
         """
         return message_type in self._subscriptions
 
-    async def relay(self):
+    async def run_channel_tasks(self, *tasks, **kwargs):
         """
-        Relay messages from the channel's queue and send them out. This is started
+        Create and await on the channel tasks unless an exception
-        as a task.
+        was raised, then cancel them all.
+
+        :params *tasks: All coros or tasks to be run concurrently
+        :param **kwargs: Any extra kwargs to pass to gather
         """
-        while not self._closed.is_set():
+
-            message = await self.queue.get()
+        if not self.is_closed():
+            # Wrap the coros into tasks if they aren't already
+            self._channel_tasks = [
+                task if isinstance(task, asyncio.Task) else asyncio.create_task(task)
+                for task in tasks
+            ]
+
             try:
-                await self._send(message)
+                return await asyncio.gather(*self._channel_tasks, **kwargs)
-                self.queue.task_done()
+            except Exception:
+                # If an exception occurred, cancel the rest of the tasks
+                await self.cancel_channel_tasks()
 
-                # Limit messages per sec.
+    async def cancel_channel_tasks(self):
-                # Could cause problems with queue size if too low, and
-                # problems with network traffik if too high.
-                # 0.01 = 100/s
-                await asyncio.sleep(self.throttle)
-            except RuntimeError:
-                # The connection was closed, just exit the task
-                return
-
-
-class ChannelManager:
-    def __init__(self):
-        self.channels = dict()
-        self._lock = RLock()  # Re-entrant Lock
-
-    async def on_connect(self, websocket: WebSocketType):
         """
-        Wrap websocket connection into Channel and add to list
+        Cancel and wait on all channel tasks
-
-        :param websocket: The WebSocket object to attach to the Channel
         """
-        if isinstance(websocket, FastAPIWebSocket):
+        for task in self._channel_tasks:
+            task.cancel()
+
+            # Wait for tasks to finish cancelling
             try:
-                await websocket.accept()
+                await task
-            except RuntimeError:
+            except (
-                # The connection was closed before we could accept it
+                asyncio.CancelledError,
-                return
+                asyncio.TimeoutError,
+                WebSocketDisconnect,
+                ConnectionClosed,
+                RuntimeError
+            ):
+                pass
+            except Exception as e:
+                logger.info(f"Encountered unknown exception: {e}", exc_info=e)
 
-        ws_channel = WebSocketChannel(websocket)
+        self._channel_tasks = []
 
-        with self._lock:
+    async def __aiter__(self):
-            self.channels[websocket] = ws_channel
-
-        return ws_channel
-
-    async def on_disconnect(self, websocket: WebSocketType):
         """
-        Call close on the channel if it's not, and remove from channel list
+        Generator for received messages
+        """
+        # We can not catch any errors here as websocket.recv is
+        # the first to catch any disconnects and bubble it up
+        # so the connection is garbage collected right away
+        while not self.is_closed():
+            yield await self.recv()
 
-        :param websocket: The WebSocket objet attached to the Channel
-        """
-        with self._lock:
-            channel = self.channels.get(websocket)
-            if channel:
-                logger.info(f"Disconnecting channel {channel}")
-                if not channel.is_closed():
-                    await channel.close()
 
-                del self.channels[websocket]
+@asynccontextmanager
+async def create_channel(
+    websocket: WebSocketType,
+    **kwargs
+) -> AsyncIterator[WebSocketChannel]:
+    """
+    Context manager for safely opening and closing a WebSocketChannel
+    """
+    channel = WebSocketChannel(websocket, **kwargs)
+    try:
+        await channel.accept()
+        logger.info(f"Connected to channel - {channel}")
 
-    async def disconnect_all(self):
+        yield channel
-        """
+    finally:
-        Disconnect all Channels
+        await channel.close()
-        """
+        logger.info(f"Disconnected from channel - {channel}")
-        with self._lock:
-            for websocket in self.channels.copy().keys():
-                await self.on_disconnect(websocket)
-
-    async def broadcast(self, message: WSMessageSchemaType):
-        """
-        Broadcast a message on all Channels
-
-        :param message: The message to send
-        """
-        with self._lock:
-            for channel in self.channels.copy().values():
-                if channel.subscribed_to(message.get('type')):
-                    await self.send_direct(channel, message)
-
-    async def send_direct(
-            self, channel: WebSocketChannel, message: Union[WSMessageSchemaType, Dict[str, Any]]):
-        """
-        Send a message directly through direct_channel only
-
-        :param direct_channel: The WebSocketChannel object to send the message through
-        :param message: The message to send
-        """
-        if not await channel.send(message):
-            await self.on_disconnect(channel.raw_websocket)
-
-    def has_channels(self):
-        """
-        Flag for more than 0 channels
-        """
-        return len(self.channels) > 0

freqtrade/rpc/api_server/ws/message_stream.py

@@ -0,0 +1,31 @@
+import asyncio
+import time
+
+
+class MessageStream:
+    """
+    A message stream for consumers to subscribe to,
+    and for producers to publish to.
+    """
+    def __init__(self):
+        self._loop = asyncio.get_running_loop()
+        self._waiter = self._loop.create_future()
+
+    def publish(self, message):
+        """
+        Publish a message to this MessageStream
+
+        :param message: The message to publish
+        """
+        waiter, self._waiter = self._waiter, self._loop.create_future()
+        waiter.set_result((message, time.time(), self._waiter))
+
+    async def __aiter__(self):
+        """
+        Iterate over the messages in the message stream
+        """
+        waiter = self._waiter
+        while True:
+            # Shield the future from being cancelled by a task waiting on it
+            message, ts, waiter = await asyncio.shield(waiter)
+            yield message, ts

freqtrade/rpc/api_server/ws/serializer.py

@@ -1,5 +1,6 @@
 import logging
 from abc import ABC, abstractmethod
+from typing import Any, Dict, Union
 
 import orjson
 import rapidjson
@@ -7,6 +8,7 @@ from pandas import DataFrame
 
 from freqtrade.misc import dataframe_to_json, json_to_dataframe
 from freqtrade.rpc.api_server.ws.proxy import WebSocketProxy
+from freqtrade.rpc.api_server.ws_schemas import WSMessageSchemaType
 
 
 logger = logging.getLogger(__name__)
@@ -24,17 +26,13 @@ class WebSocketSerializer(ABC):
     def _deserialize(self, data):
         raise NotImplementedError()
 
-    async def send(self, data: bytes):
+    async def send(self, data: Union[WSMessageSchemaType, Dict[str, Any]]):
         await self._websocket.send(self._serialize(data))
 
     async def recv(self) -> bytes:
         data = await self._websocket.recv()
-
         return self._deserialize(data)
 
-    async def close(self, code: int = 1000):
-        await self._websocket.close(code)
-
 
 class HybridJSONWebSocketSerializer(WebSocketSerializer):
     def _serialize(self, data) -> str:

freqtrade/rpc/external_message_consumer.py

@@ -31,6 +31,7 @@ class Producer(TypedDict):
     name: str
     host: str
     port: int
+    secure: bool
     ws_token: str
 
 
@@ -180,7 +181,8 @@ class ExternalMessageConsumer:
                 host, port = producer['host'], producer['port']
                 token = producer['ws_token']
                 name = producer['name']
-                ws_url = f"ws://{host}:{port}/api/v1/message/ws?token={token}"
+                scheme = 'wss' if producer.get('secure', False) else 'ws'
+                ws_url = f"{scheme}://{host}:{port}/api/v1/message/ws?token={token}"
 
                 # This will raise InvalidURI if the url is bad
                 async with websockets.connect(

freqtrade/rpc/rpc.py

@@ -789,17 +789,18 @@ class RPC:
         if not order_type:
             order_type = self._freqtrade.strategy.order_types.get(
                 'force_entry', self._freqtrade.strategy.order_types['entry'])
-        if self._freqtrade.execute_entry(pair, stake_amount, price,
+        with self._freqtrade._exit_lock:
-                                         ordertype=order_type, trade=trade,
+            if self._freqtrade.execute_entry(pair, stake_amount, price,
-                                         is_short=is_short,
+                                             ordertype=order_type, trade=trade,
-                                         enter_tag=enter_tag,
+                                             is_short=is_short,
-                                         leverage_=leverage,
+                                             enter_tag=enter_tag,
-                                         ):
+                                             leverage_=leverage,
-            Trade.commit()
+                                             ):
-            trade = Trade.get_trades([Trade.is_open.is_(True), Trade.pair == pair]).first()
+                Trade.commit()
-            return trade
+                trade = Trade.get_trades([Trade.is_open.is_(True), Trade.pair == pair]).first()
-        else:
+                return trade
-            raise RPCException(f'Failed to enter position for {pair}.')
+            else:
+                raise RPCException(f'Failed to enter position for {pair}.')
 
     def _rpc_delete(self, trade_id: int) -> Dict[str, Union[str, int]]:
         """

freqtrade/templates/FreqaiExampleHybridStrategy.py

@@ -19,7 +19,7 @@ class FreqaiExampleHybridStrategy(IStrategy):
 
     Launching this strategy would be:
 
-    freqtrade trade --strategy FreqaiExampleHyridStrategy --strategy-path freqtrade/templates
+    freqtrade trade --strategy FreqaiExampleHybridStrategy --strategy-path freqtrade/templates
     --freqaimodel CatboostClassifier --config config_examples/config_freqai.example.json
 
     or the user simply adds this to their config:
@@ -86,7 +86,7 @@ class FreqaiExampleHybridStrategy(IStrategy):
     process_only_new_candles = True
     stoploss = -0.05
     use_exit_signal = True
-    startup_candle_count: int = 300
+    startup_candle_count: int = 30
     can_short = True
 
     # Hyperoptable parameters

freqtrade/templates/strategy_analysis_example.ipynb

@@ -328,7 +328,7 @@
     "# Show graph inline\n",
     "# graph.show()\n",
     "\n",
-    "# Render graph in a seperate window\n",
+    "# Render graph in a separate window\n",
     "graph.show(renderer=\"browser\")\n"
    ]
   },

mkdocs.yml

@@ -29,6 +29,7 @@ nav:
         - Parameter table: freqai-parameter-table.md
         - Feature engineering: freqai-feature-engineering.md
         - Running FreqAI: freqai-running.md
+        - Reinforcement Learning: freqai-reinforcement-learning.md
         - Developer guide: freqai-developers.md
     - Short / Leverage: leverage.md
     - Utility Sub-commands: utils.md

requirements-dev.txt

@@ -3,10 +3,11 @@
 -r requirements-plot.txt
 -r requirements-hyperopt.txt
 -r requirements-freqai.txt
+-r requirements-freqai-rl.txt
 -r docs/requirements-docs.txt
 
 coveralls==3.3.1
-flake8==5.0.4
+flake8==6.0.0
 flake8-tidy-imports==4.8.0
 mypy==0.991
 pre-commit==2.20.0

requirements-freqai-rl.txt

@@ -0,0 +1,9 @@
+# Include all requirements to run the bot.
+-r requirements-freqai.txt
+
+# Required for freqai-rl
+torch==1.12.1
+stable-baselines3==1.6.2
+sb3-contrib==1.6.2
+# Gym is forced to this version by stable-baselines3.
+gym==0.21

requirements.txt

@@ -2,18 +2,18 @@ numpy==1.23.5
 pandas==1.5.1
 pandas-ta==0.3.14b
 
-ccxt==2.1.96
+ccxt==2.2.36
 # Pin cryptography for now due to rust build errors with piwheels
 cryptography==38.0.1; platform_machine == 'armv7l'
-cryptography==38.0.3; platform_machine != 'armv7l'
+cryptography==38.0.4; platform_machine != 'armv7l'
 aiohttp==3.8.3
 SQLAlchemy==1.4.44
 python-telegram-bot==13.14
 arrow==1.2.3
 cachetools==4.2.2
 requests==2.28.1
-urllib3==1.26.12
+urllib3==1.26.13
-jsonschema==4.17.0
+jsonschema==4.17.1
 TA-Lib==0.4.25
 technical==1.3.0
 tabulate==0.9.0
@@ -22,7 +22,7 @@ jinja2==3.1.2
 tables==3.7.0
 blosc==1.10.6
 joblib==1.2.0
-pyarrow==10.0.0; platform_machine != 'armv7l'
+pyarrow==10.0.1; platform_machine != 'armv7l'
 
 # find first, C search in arrays
 py_find_1st==1.1.5
@@ -47,7 +47,7 @@ psutil==5.9.4
 colorama==0.4.6
 # Building config files interactively
 questionary==1.10.0
-prompt-toolkit==3.0.32
+prompt-toolkit==3.0.33
 # Extensions to datetime library
 python-dateutil==2.8.2
 

scripts/ws_client.py

@@ -199,6 +199,7 @@ async def create_client(
         host,
         port,
         token,
+        scheme='ws',
         name='default',
         protocol=ClientProtocol(),
         sleep_time=10,
@@ -211,13 +212,14 @@ async def create_client(
     :param host: The host
     :param port: The port
     :param token: The websocket auth token
+    :param scheme: `ws` for most connections, `wss` for ssl
     :param name: The name of the producer
     :param **kwargs: Any extra kwargs passed to websockets.connect
     """
 
     while 1:
         try:
-            websocket_url = f"ws://{host}:{port}/api/v1/message/ws?token={token}"
+            websocket_url = f"{scheme}://{host}:{port}/api/v1/message/ws?token={token}"
             logger.info(f"Attempting to connect to {name} @ {host}:{port}")
 
             async with websockets.connect(websocket_url, **kwargs) as ws:
@@ -304,6 +306,7 @@ async def _main(args):
         producer['host'],
         producer['port'],
         producer['ws_token'],
+        'wss' if producer.get('secure', False) else 'ws',
         producer['name'],
         sleep_time=sleep_time,
         ping_timeout=ping_timeout,

setup.py

@@ -15,6 +15,14 @@ freqai = [
     'scikit-learn',
     'catboost; platform_machine != "aarch64"',
     'lightgbm',
+    'xgboost'
+]
+
+freqai_rl = [
+    'torch',
+    'stable-baselines3',
+    'gym==0.21',
+    'sb3-contrib'
 ]
 
 develop = [
@@ -36,7 +44,7 @@ jupyter = [
     'nbconvert',
 ]
 
-all_extra = plot + develop + jupyter + hyperopt + freqai
+all_extra = plot + develop + jupyter + hyperopt + freqai + freqai_rl
 
 setup(
     tests_require=[
@@ -90,6 +98,7 @@ setup(
         'jupyter': jupyter,
         'hyperopt': hyperopt,
         'freqai': freqai,
+        'freqai_rl': freqai_rl,
         'all': all_extra,
     },
 )

setup.sh

@@ -78,14 +78,21 @@ function updateenv() {
     fi
 
     REQUIREMENTS_FREQAI=""
+    REQUIREMENTS_FREQAI_RL=""
     read -p "Do you want to install dependencies for freqai [y/N]? "
     dev=$REPLY
     if [[ $REPLY =~ ^[Yy]$ ]]
     then
-        REQUIREMENTS_FREQAI="-r requirements-freqai.txt"
+        REQUIREMENTS_FREQAI="-r requirements-freqai.txt --use-pep517"
+        read -p "Do you also want dependencies for freqai-rl (~700mb additional space required) [y/N]? "
+        dev=$REPLY
+        if [[ $REPLY =~ ^[Yy]$ ]]
+        then
+            REQUIREMENTS_FREQAI="-r requirements-freqai-rl.txt"
+        fi
     fi
 
-    ${PYTHON} -m pip install --upgrade -r ${REQUIREMENTS} ${REQUIREMENTS_HYPEROPT} ${REQUIREMENTS_PLOT} ${REQUIREMENTS_FREQAI}
+    ${PYTHON} -m pip install --upgrade -r ${REQUIREMENTS} ${REQUIREMENTS_HYPEROPT} ${REQUIREMENTS_PLOT} ${REQUIREMENTS_FREQAI} ${REQUIREMENTS_FREQAI_RL}
     if [ $? -ne 0 ]; then
         echo "Failed installing dependencies"
         exit 1

tests/commands/test_commands.py

@@ -1271,7 +1271,7 @@ def test_hyperopt_list(mocker, capsys, caplog, saved_hyperopt_results, tmpdir):
     assert csv_file.is_file()
     line = csv_file.read_text()
     assert ('Best,1,2,-1.25%,-1.2222,-0.00125625,,-2.51,"3,930.0 m",0.43662' in line
-            or "Best,1,2,-1.25%,-1.2222,-0.00125625,,-2.51,2 days 17:30:00,0.43662" in line)
+            or "Best,1,2,-1.25%,-1.2222,-0.00125625,,-2.51,2 days 17:30:00,2,0,0.43662" in line)
     csv_file.unlink()
 
 

tests/conftest.py

@@ -2679,7 +2679,7 @@ def saved_hyperopt_results():
             'params_dict': {
                 'mfi-value': 15, 'fastd-value': 20, 'adx-value': 25, 'rsi-value': 28, 'mfi-enabled': False, 'fastd-enabled': True, 'adx-enabled': True, 'rsi-enabled': True, 'trigger': 'macd_cross_signal', 'sell-mfi-value': 88, 'sell-fastd-value': 97, 'sell-adx-value': 51, 'sell-rsi-value': 67, 'sell-mfi-enabled': False, 'sell-fastd-enabled': False, 'sell-adx-enabled': True, 'sell-rsi-enabled': True, 'sell-trigger': 'sell-bb_upper', 'roi_t1': 1190, 'roi_t2': 541, 'roi_t3': 408, 'roi_p1': 0.026035863879169705, 'roi_p2': 0.12508730043628782, 'roi_p3': 0.27766427921605896, 'stoploss': -0.2562930402099556},  # noqa: E501
             'params_details': {'buy': {'mfi-value': 15, 'fastd-value': 20, 'adx-value': 25, 'rsi-value': 28, 'mfi-enabled': False, 'fastd-enabled': True, 'adx-enabled': True, 'rsi-enabled': True, 'trigger': 'macd_cross_signal'}, 'sell': {'sell-mfi-value': 88, 'sell-fastd-value': 97, 'sell-adx-value': 51, 'sell-rsi-value': 67, 'sell-mfi-enabled': False, 'sell-fastd-enabled': False, 'sell-adx-enabled': True, 'sell-rsi-enabled': True, 'sell-trigger': 'sell-bb_upper'}, 'roi': {0: 0.4287874435315165, 408: 0.15112316431545753, 949: 0.026035863879169705, 2139: 0}, 'stoploss': {'stoploss': -0.2562930402099556}},  # noqa: E501
-            'results_metrics': {'total_trades': 2, 'wins': 0, 'draws': 0, 'losses': 2, 'profit_mean': -0.01254995, 'profit_median': -0.012222, 'profit_total': -0.00125625,  'profit_total_abs': -2.50999, 'max_drawdown': 0.23, 'max_drawdown_abs': -0.00125625,  'holding_avg': timedelta(minutes=3930.0), 'stake_currency': 'BTC', 'strategy_name': 'SampleStrategy'},  # noqa: E501
+            'results_metrics': {'total_trades': 2, 'trade_count_long': 2, 'trade_count_short': 0, 'wins': 0, 'draws': 0, 'losses': 2, 'profit_mean': -0.01254995, 'profit_median': -0.012222, 'profit_total': -0.00125625,  'profit_total_abs': -2.50999, 'max_drawdown': 0.23, 'max_drawdown_abs': -0.00125625,  'holding_avg': timedelta(minutes=3930.0), 'stake_currency': 'BTC', 'strategy_name': 'SampleStrategy'},  # noqa: E501
             'results_explanation': '     2 trades. Avg profit  -1.25%. Total profit -0.00125625 BTC (  -2.51Σ%). Avg duration 3930.0 min.',  # noqa: E501
             'total_profit': -0.00125625,
             'current_epoch': 1,
@@ -2696,7 +2696,7 @@ def saved_hyperopt_results():
                 'sell': {'sell-mfi-value': 96, 'sell-fastd-value': 68, 'sell-adx-value': 63, 'sell-rsi-value': 81, 'sell-mfi-enabled': False, 'sell-fastd-enabled': True, 'sell-adx-enabled': True, 'sell-rsi-enabled': True, 'sell-trigger': 'sell-sar_reversal'},  # noqa: E501
                 'roi': {0: 0.4449309386008759, 140: 0.11955965746663, 823: 0.06403981740598495, 1157: 0},  # noqa: E501
                 'stoploss': {'stoploss': -0.338070047333259}},
-            'results_metrics': {'total_trades': 1, 'wins': 0, 'draws': 0, 'losses': 1, 'profit_mean': 0.012357, 'profit_median': -0.012222, 'profit_total': 6.185e-05, 'profit_total_abs': 0.12357, 'max_drawdown': 0.23, 'max_drawdown_abs': -0.00125625, 'holding_avg': timedelta(minutes=1200.0)},  # noqa: E501
+            'results_metrics': {'total_trades': 1, 'trade_count_long': 1, 'trade_count_short': 0, 'wins': 0, 'draws': 0, 'losses': 1, 'profit_mean': 0.012357, 'profit_median': -0.012222, 'profit_total': 6.185e-05, 'profit_total_abs': 0.12357, 'max_drawdown': 0.23, 'max_drawdown_abs': -0.00125625, 'holding_avg': timedelta(minutes=1200.0)},  # noqa: E501
             'results_explanation': '     1 trades. Avg profit   0.12%. Total profit  0.00006185 BTC (   0.12Σ%). Avg duration 1200.0 min.',  # noqa: E501
             'total_profit': 6.185e-05,
             'current_epoch': 2,
@@ -2707,7 +2707,7 @@ def saved_hyperopt_results():
             'loss': 14.241196856510731,
             'params_dict': {'mfi-value': 25, 'fastd-value': 16, 'adx-value': 29, 'rsi-value': 20, 'mfi-enabled': False, 'fastd-enabled': False, 'adx-enabled': False, 'rsi-enabled': False, 'trigger': 'macd_cross_signal', 'sell-mfi-value': 98, 'sell-fastd-value': 72, 'sell-adx-value': 51, 'sell-rsi-value': 82, 'sell-mfi-enabled': True, 'sell-fastd-enabled': True, 'sell-adx-enabled': True, 'sell-rsi-enabled': True, 'sell-trigger': 'sell-macd_cross_signal', 'roi_t1': 889, 'roi_t2': 533, 'roi_t3': 263, 'roi_p1': 0.04759065393663096, 'roi_p2': 0.1488819964638463, 'roi_p3': 0.4102801822104605, 'stoploss': -0.05394588767607611},  # noqa: E501
             'params_details': {'buy': {'mfi-value': 25, 'fastd-value': 16, 'adx-value': 29, 'rsi-value': 20, 'mfi-enabled': False, 'fastd-enabled': False, 'adx-enabled': False, 'rsi-enabled': False, 'trigger': 'macd_cross_signal'}, 'sell': {'sell-mfi-value': 98, 'sell-fastd-value': 72, 'sell-adx-value': 51, 'sell-rsi-value': 82, 'sell-mfi-enabled': True, 'sell-fastd-enabled': True, 'sell-adx-enabled': True, 'sell-rsi-enabled': True, 'sell-trigger': 'sell-macd_cross_signal'}, 'roi': {0: 0.6067528326109377, 263: 0.19647265040047726, 796: 0.04759065393663096, 1685: 0}, 'stoploss': {'stoploss': -0.05394588767607611}},  # noqa: E501
-            'results_metrics': {'total_trades': 621, 'wins': 320, 'draws': 0, 'losses': 301, 'profit_mean': -0.043883302093397747, 'profit_median': -0.012222, 'profit_total': -0.13639474, 'profit_total_abs': -272.515306, 'max_drawdown': 0.25, 'max_drawdown_abs': -272.515306, 'holding_avg': timedelta(minutes=1691.207729468599)},  # noqa: E501
+            'results_metrics': {'total_trades': 621, 'trade_count_long': 621, 'trade_count_short': 0, 'wins': 320, 'draws': 0, 'losses': 301, 'profit_mean': -0.043883302093397747, 'profit_median': -0.012222, 'profit_total': -0.13639474, 'profit_total_abs': -272.515306, 'max_drawdown': 0.25, 'max_drawdown_abs': -272.515306, 'holding_avg': timedelta(minutes=1691.207729468599)},  # noqa: E501
             'results_explanation': '   621 trades. Avg profit  -0.44%. Total profit -0.13639474 BTC (-272.52Σ%). Avg duration 1691.2 min.',  # noqa: E501
             'total_profit': -0.13639474,
             'current_epoch': 3,
@@ -2718,14 +2718,14 @@ def saved_hyperopt_results():
             'loss': 100000,
             'params_dict': {'mfi-value': 13, 'fastd-value': 35, 'adx-value': 39, 'rsi-value': 29, 'mfi-enabled': True, 'fastd-enabled': False, 'adx-enabled': False, 'rsi-enabled': True, 'trigger': 'macd_cross_signal', 'sell-mfi-value': 87, 'sell-fastd-value': 54, 'sell-adx-value': 63, 'sell-rsi-value': 93, 'sell-mfi-enabled': False, 'sell-fastd-enabled': True, 'sell-adx-enabled': True, 'sell-rsi-enabled': True, 'sell-trigger': 'sell-bb_upper', 'roi_t1': 1402, 'roi_t2': 676, 'roi_t3': 215, 'roi_p1': 0.06264755784937427, 'roi_p2': 0.14258587851894644, 'roi_p3': 0.20671291201040828, 'stoploss': -0.11818343570194478},  # noqa: E501
             'params_details': {'buy': {'mfi-value': 13, 'fastd-value': 35, 'adx-value': 39, 'rsi-value': 29, 'mfi-enabled': True, 'fastd-enabled': False, 'adx-enabled': False, 'rsi-enabled': True, 'trigger': 'macd_cross_signal'}, 'sell': {'sell-mfi-value': 87, 'sell-fastd-value': 54, 'sell-adx-value': 63, 'sell-rsi-value': 93, 'sell-mfi-enabled': False, 'sell-fastd-enabled': True, 'sell-adx-enabled': True, 'sell-rsi-enabled': True, 'sell-trigger': 'sell-bb_upper'}, 'roi': {0: 0.411946348378729, 215: 0.2052334363683207, 891: 0.06264755784937427, 2293: 0}, 'stoploss': {'stoploss': -0.11818343570194478}},  # noqa: E501
-            'results_metrics': {'total_trades': 0, 'wins': 0, 'draws': 0, 'losses': 0, 'profit_mean': None, 'profit_median': None, 'profit_total': 0, 'profit': 0.0, 'holding_avg': timedelta()},  # noqa: E501
+            'results_metrics': {'total_trades': 0, 'trade_count_long': 0, 'trade_count_short': 0, 'wins': 0, 'draws': 0, 'losses': 0, 'profit_mean': None, 'profit_median': None, 'profit_total': 0, 'profit': 0.0, 'holding_avg': timedelta()},  # noqa: E501
             'results_explanation': '     0 trades. Avg profit    nan%. Total profit  0.00000000 BTC (   0.00Σ%). Avg duration   nan min.',  # noqa: E501
             'total_profit': 0, 'current_epoch': 4, 'is_initial_point': True, 'is_random': False, 'is_best': False  # noqa: E501
         }, {
             'loss': 0.22195522184191518,
             'params_dict': {'mfi-value': 17, 'fastd-value': 21, 'adx-value': 38, 'rsi-value': 33, 'mfi-enabled': True, 'fastd-enabled': False, 'adx-enabled': True, 'rsi-enabled': False, 'trigger': 'macd_cross_signal', 'sell-mfi-value': 87, 'sell-fastd-value': 82, 'sell-adx-value': 78, 'sell-rsi-value': 69, 'sell-mfi-enabled': True, 'sell-fastd-enabled': False, 'sell-adx-enabled': True, 'sell-rsi-enabled': False, 'sell-trigger': 'sell-macd_cross_signal', 'roi_t1': 1269, 'roi_t2': 601, 'roi_t3': 444, 'roi_p1': 0.07280999507931168, 'roi_p2': 0.08946698095898986, 'roi_p3': 0.1454876733325284, 'stoploss': -0.18181041180901014},   # noqa: E501
             'params_details': {'buy': {'mfi-value': 17, 'fastd-value': 21, 'adx-value': 38, 'rsi-value': 33, 'mfi-enabled': True, 'fastd-enabled': False, 'adx-enabled': True, 'rsi-enabled': False, 'trigger': 'macd_cross_signal'}, 'sell': {'sell-mfi-value': 87, 'sell-fastd-value': 82, 'sell-adx-value': 78, 'sell-rsi-value': 69, 'sell-mfi-enabled': True, 'sell-fastd-enabled': False, 'sell-adx-enabled': True, 'sell-rsi-enabled': False, 'sell-trigger': 'sell-macd_cross_signal'}, 'roi': {0: 0.3077646493708299, 444: 0.16227697603830155, 1045: 0.07280999507931168, 2314: 0}, 'stoploss': {'stoploss': -0.18181041180901014}},  # noqa: E501
-            'results_metrics': {'total_trades': 14, 'wins': 6, 'draws': 0, 'losses': 8, 'profit_mean': -0.003539515, 'profit_median': -0.012222, 'profit_total': -0.002480140000000001, 'profit_total_abs': -4.955321, 'max_drawdown': 0.34, 'max_drawdown_abs': -4.955321, 'holding_avg': timedelta(minutes=3402.8571428571427)},  # noqa: E501
+            'results_metrics': {'total_trades': 14, 'trade_count_long': 14, 'trade_count_short': 0, 'wins': 6, 'draws': 0, 'losses': 8, 'profit_mean': -0.003539515, 'profit_median': -0.012222, 'profit_total': -0.002480140000000001, 'profit_total_abs': -4.955321, 'max_drawdown': 0.34, 'max_drawdown_abs': -4.955321, 'holding_avg': timedelta(minutes=3402.8571428571427)},  # noqa: E501
             'results_explanation': '    14 trades. Avg profit  -0.35%. Total profit -0.00248014 BTC (  -4.96Σ%). Avg duration 3402.9 min.',  # noqa: E501
             'total_profit': -0.002480140000000001,
             'current_epoch': 5,
@@ -2736,7 +2736,7 @@ def saved_hyperopt_results():
             'loss': 0.545315889154162,
             'params_dict': {'mfi-value': 22, 'fastd-value': 43, 'adx-value': 46, 'rsi-value': 20, 'mfi-enabled': False, 'fastd-enabled': False, 'adx-enabled': True, 'rsi-enabled': True, 'trigger': 'bb_lower', 'sell-mfi-value': 87, 'sell-fastd-value': 65, 'sell-adx-value': 94, 'sell-rsi-value': 63, 'sell-mfi-enabled': False, 'sell-fastd-enabled': True, 'sell-adx-enabled': True, 'sell-rsi-enabled': True, 'sell-trigger': 'sell-macd_cross_signal', 'roi_t1': 319, 'roi_t2': 556, 'roi_t3': 216, 'roi_p1': 0.06251955472249589, 'roi_p2': 0.11659519602202795, 'roi_p3': 0.0953744132197762, 'stoploss': -0.024551752215582423},  # noqa: E501
             'params_details': {'buy': {'mfi-value': 22, 'fastd-value': 43, 'adx-value': 46, 'rsi-value': 20, 'mfi-enabled': False, 'fastd-enabled': False, 'adx-enabled': True, 'rsi-enabled': True, 'trigger': 'bb_lower'}, 'sell': {'sell-mfi-value': 87, 'sell-fastd-value': 65, 'sell-adx-value': 94, 'sell-rsi-value': 63, 'sell-mfi-enabled': False, 'sell-fastd-enabled': True, 'sell-adx-enabled': True, 'sell-rsi-enabled': True, 'sell-trigger': 'sell-macd_cross_signal'}, 'roi': {0: 0.2744891639643, 216: 0.17911475074452382, 772: 0.06251955472249589, 1091: 0}, 'stoploss': {'stoploss': -0.024551752215582423}},  # noqa: E501
-            'results_metrics': {'total_trades': 39, 'wins': 20, 'draws': 0, 'losses': 19, 'profit_mean': -0.0021400679487179478, 'profit_median': -0.012222, 'profit_total': -0.0041773, 'profit_total_abs': -8.346264999999997, 'max_drawdown': 0.45, 'max_drawdown_abs': -4.955321, 'holding_avg': timedelta(minutes=636.9230769230769)},  # noqa: E501
+            'results_metrics': {'total_trades': 39, 'trade_count_long': 39, 'trade_count_short': 0, 'wins': 20, 'draws': 0, 'losses': 19, 'profit_mean': -0.0021400679487179478, 'profit_median': -0.012222, 'profit_total': -0.0041773, 'profit_total_abs': -8.346264999999997, 'max_drawdown': 0.45, 'max_drawdown_abs': -4.955321, 'holding_avg': timedelta(minutes=636.9230769230769)},  # noqa: E501
             'results_explanation': '    39 trades. Avg profit  -0.21%. Total profit -0.00417730 BTC (  -8.35Σ%). Avg duration 636.9 min.',  # noqa: E501
             'total_profit': -0.0041773,
             'current_epoch': 6,
@@ -2749,7 +2749,7 @@ def saved_hyperopt_results():
             'params_details': {
                 'buy': {'mfi-value': 13, 'fastd-value': 41, 'adx-value': 21, 'rsi-value': 29, 'mfi-enabled': False, 'fastd-enabled': True, 'adx-enabled': False, 'rsi-enabled': False, 'trigger': 'bb_lower'}, 'sell': {'sell-mfi-value': 99, 'sell-fastd-value': 60, 'sell-adx-value': 81, 'sell-rsi-value': 69, 'sell-mfi-enabled': True, 'sell-fastd-enabled': True, 'sell-adx-enabled': True, 'sell-rsi-enabled': False, 'sell-trigger': 'sell-macd_cross_signal'}, 'roi': {0: 0.4837436938134452, 145: 0.10853310701097472, 765: 0.0586919200378493, 1536: 0},  # noqa: E501
                 'stoploss': {'stoploss': -0.14613268022709905}},  # noqa: E501
-            'results_metrics': {'total_trades': 318, 'wins': 100, 'draws': 0, 'losses': 218, 'profit_mean': -0.0039833954716981146, 'profit_median': -0.012222, 'profit_total': -0.06339929, 'profit_total_abs': -126.67197600000004, 'max_drawdown': 0.50, 'max_drawdown_abs': -200.955321, 'holding_avg': timedelta(minutes=3140.377358490566)},  # noqa: E501
+            'results_metrics': {'total_trades': 318, 'trade_count_long': 318, 'trade_count_short': 0, 'wins': 100, 'draws': 0, 'losses': 218, 'profit_mean': -0.0039833954716981146, 'profit_median': -0.012222, 'profit_total': -0.06339929, 'profit_total_abs': -126.67197600000004, 'max_drawdown': 0.50, 'max_drawdown_abs': -200.955321, 'holding_avg': timedelta(minutes=3140.377358490566)},  # noqa: E501
             'results_explanation': '   318 trades. Avg profit  -0.40%. Total profit -0.06339929 BTC (-126.67Σ%). Avg duration 3140.4 min.',  # noqa: E501
             'total_profit': -0.06339929,
             'current_epoch': 7,
@@ -2760,7 +2760,7 @@ def saved_hyperopt_results():
             'loss': 20.0,  # noqa: E501
             'params_dict': {'mfi-value': 24, 'fastd-value': 43, 'adx-value': 33, 'rsi-value': 20, 'mfi-enabled': False, 'fastd-enabled': True, 'adx-enabled': True, 'rsi-enabled': True, 'trigger': 'sar_reversal', 'sell-mfi-value': 89, 'sell-fastd-value': 74, 'sell-adx-value': 70, 'sell-rsi-value': 70, 'sell-mfi-enabled': False, 'sell-fastd-enabled': False, 'sell-adx-enabled': False, 'sell-rsi-enabled': True, 'sell-trigger': 'sell-sar_reversal', 'roi_t1': 1149, 'roi_t2': 375, 'roi_t3': 289, 'roi_p1': 0.05571820757172588, 'roi_p2': 0.0606240398618907, 'roi_p3': 0.1729012220156157, 'stoploss': -0.1588514289110401},  # noqa: E501
             'params_details': {'buy': {'mfi-value': 24, 'fastd-value': 43, 'adx-value': 33, 'rsi-value': 20, 'mfi-enabled': False, 'fastd-enabled': True, 'adx-enabled': True, 'rsi-enabled': True, 'trigger': 'sar_reversal'}, 'sell': {'sell-mfi-value': 89, 'sell-fastd-value': 74, 'sell-adx-value': 70, 'sell-rsi-value': 70, 'sell-mfi-enabled': False, 'sell-fastd-enabled': False, 'sell-adx-enabled': False, 'sell-rsi-enabled': True, 'sell-trigger': 'sell-sar_reversal'}, 'roi': {0: 0.2892434694492323, 289: 0.11634224743361658, 664: 0.05571820757172588, 1813: 0}, 'stoploss': {'stoploss': -0.1588514289110401}},  # noqa: E501
-            'results_metrics': {'total_trades': 1, 'wins': 0, 'draws': 1, 'losses': 0, 'profit_mean': 0.0, 'profit_median': 0.0, 'profit_total': 0.0, 'profit_total_abs': 0.0, 'max_drawdown': 0.0, 'max_drawdown_abs': 0.52, 'holding_avg': timedelta(minutes=5340.0)},  # noqa: E501
+            'results_metrics': {'total_trades': 1, 'trade_count_long': 1, 'trade_count_short': 0, 'wins': 0, 'draws': 1, 'losses': 0, 'profit_mean': 0.0, 'profit_median': 0.0, 'profit_total': 0.0, 'profit_total_abs': 0.0, 'max_drawdown': 0.0, 'max_drawdown_abs': 0.52, 'holding_avg': timedelta(minutes=5340.0)},  # noqa: E501
             'results_explanation': '     1 trades. Avg profit   0.00%. Total profit  0.00000000 BTC (   0.00Σ%). Avg duration 5340.0 min.',  # noqa: E501
             'total_profit': 0.0,
             'current_epoch': 8,
@@ -2771,7 +2771,7 @@ def saved_hyperopt_results():
             'loss': 2.4731817780991223,
             'params_dict': {'mfi-value': 22, 'fastd-value': 20, 'adx-value': 29, 'rsi-value': 40, 'mfi-enabled': False, 'fastd-enabled': False, 'adx-enabled': False, 'rsi-enabled': False, 'trigger': 'sar_reversal', 'sell-mfi-value': 97, 'sell-fastd-value': 65, 'sell-adx-value': 81, 'sell-rsi-value': 64, 'sell-mfi-enabled': True, 'sell-fastd-enabled': True, 'sell-adx-enabled': True, 'sell-rsi-enabled': True, 'sell-trigger': 'sell-bb_upper', 'roi_t1': 1012, 'roi_t2': 584, 'roi_t3': 422, 'roi_p1': 0.036764323603472565, 'roi_p2': 0.10335480573205287, 'roi_p3': 0.10322347377503042, 'stoploss': -0.2780610808108503},  # noqa: E501
             'params_details': {'buy': {'mfi-value': 22, 'fastd-value': 20, 'adx-value': 29, 'rsi-value': 40, 'mfi-enabled': False, 'fastd-enabled': False, 'adx-enabled': False, 'rsi-enabled': False, 'trigger': 'sar_reversal'}, 'sell': {'sell-mfi-value': 97, 'sell-fastd-value': 65, 'sell-adx-value': 81, 'sell-rsi-value': 64, 'sell-mfi-enabled': True, 'sell-fastd-enabled': True, 'sell-adx-enabled': True, 'sell-rsi-enabled': True, 'sell-trigger': 'sell-bb_upper'}, 'roi': {0: 0.2433426031105559, 422: 0.14011912933552545, 1006: 0.036764323603472565, 2018: 0}, 'stoploss': {'stoploss': -0.2780610808108503}},  # noqa: E501
-            'results_metrics': {'total_trades': 229, 'wins': 150, 'draws': 0, 'losses': 79, 'profit_mean': -0.0038433433624454144, 'profit_median': -0.012222, 'profit_total': -0.044050070000000004, 'profit_total_abs': -88.01256299999999, 'max_drawdown': 0.41, 'max_drawdown_abs': -150.955321, 'holding_avg': timedelta(minutes=6505.676855895196)},  # noqa: E501
+            'results_metrics': {'total_trades': 229, 'trade_count_long': 229, 'trade_count_short': 0, 'wins': 150, 'draws': 0, 'losses': 79, 'profit_mean': -0.0038433433624454144, 'profit_median': -0.012222, 'profit_total': -0.044050070000000004, 'profit_total_abs': -88.01256299999999, 'max_drawdown': 0.41, 'max_drawdown_abs': -150.955321, 'holding_avg': timedelta(minutes=6505.676855895196)},  # noqa: E501
             'results_explanation': '   229 trades. Avg profit  -0.38%. Total profit -0.04405007 BTC ( -88.01Σ%). Avg duration 6505.7 min.',  # noqa: E501
             'total_profit': -0.044050070000000004,  # noqa: E501
             'current_epoch': 9,
@@ -2782,7 +2782,7 @@ def saved_hyperopt_results():
             'loss': -0.2604606005845212,  # noqa: E501
             'params_dict': {'mfi-value': 23, 'fastd-value': 24, 'adx-value': 22, 'rsi-value': 24, 'mfi-enabled': False, 'fastd-enabled': False, 'adx-enabled': False, 'rsi-enabled': True, 'trigger': 'macd_cross_signal', 'sell-mfi-value': 97, 'sell-fastd-value': 70, 'sell-adx-value': 64, 'sell-rsi-value': 80, 'sell-mfi-enabled': False, 'sell-fastd-enabled': True, 'sell-adx-enabled': True, 'sell-rsi-enabled': True, 'sell-trigger': 'sell-sar_reversal', 'roi_t1': 792, 'roi_t2': 464, 'roi_t3': 215, 'roi_p1': 0.04594053535385903, 'roi_p2': 0.09623192684243963, 'roi_p3': 0.04428219070850663, 'stoploss': -0.16992287161634415},  # noqa: E501
             'params_details': {'buy': {'mfi-value': 23, 'fastd-value': 24, 'adx-value': 22, 'rsi-value': 24, 'mfi-enabled': False, 'fastd-enabled': False, 'adx-enabled': False, 'rsi-enabled': True, 'trigger': 'macd_cross_signal'}, 'sell': {'sell-mfi-value': 97, 'sell-fastd-value': 70, 'sell-adx-value': 64, 'sell-rsi-value': 80, 'sell-mfi-enabled': False, 'sell-fastd-enabled': True, 'sell-adx-enabled': True, 'sell-rsi-enabled': True, 'sell-trigger': 'sell-sar_reversal'}, 'roi': {0: 0.18645465290480528, 215: 0.14217246219629864, 679: 0.04594053535385903, 1471: 0}, 'stoploss': {'stoploss': -0.16992287161634415}},  # noqa: E501
-            'results_metrics': {'total_trades': 4, 'wins': 0, 'draws': 0, 'losses': 4, 'profit_mean': 0.001080385, 'profit_median': -0.012222, 'profit_total': 0.00021629, 'profit_total_abs': 0.432154, 'max_drawdown': 0.13, 'max_drawdown_abs': -4.955321, 'holding_avg': timedelta(minutes=2850.0)},  # noqa: E501
+            'results_metrics': {'total_trades': 4, 'trade_count_long': 4, 'trade_count_short': 0, 'wins': 0, 'draws': 0, 'losses': 4, 'profit_mean': 0.001080385, 'profit_median': -0.012222, 'profit_total': 0.00021629, 'profit_total_abs': 0.432154, 'max_drawdown': 0.13, 'max_drawdown_abs': -4.955321, 'holding_avg': timedelta(minutes=2850.0)},  # noqa: E501
             'results_explanation': '     4 trades. Avg profit   0.11%. Total profit  0.00021629 BTC (   0.43Σ%). Avg duration 2850.0 min.',  # noqa: E501
             'total_profit': 0.00021629,
             'current_epoch': 10,
@@ -2794,7 +2794,7 @@ def saved_hyperopt_results():
             'params_dict': {'mfi-value': 20, 'fastd-value': 32, 'adx-value': 49, 'rsi-value': 23, 'mfi-enabled': True, 'fastd-enabled': True, 'adx-enabled': False, 'rsi-enabled': False, 'trigger': 'bb_lower', 'sell-mfi-value': 75, 'sell-fastd-value': 56, 'sell-adx-value': 61, 'sell-rsi-value': 62, 'sell-mfi-enabled': False, 'sell-fastd-enabled': False, 'sell-adx-enabled': True, 'sell-rsi-enabled': True, 'sell-trigger': 'sell-macd_cross_signal', 'roi_t1': 579, 'roi_t2': 614, 'roi_t3': 273, 'roi_p1': 0.05307643172744114, 'roi_p2': 0.1352282078262871, 'roi_p3': 0.1913307406325751, 'stoploss': -0.25728526022513887},  # noqa: E501
             'params_details': {'buy': {'mfi-value': 20, 'fastd-value': 32, 'adx-value': 49, 'rsi-value': 23, 'mfi-enabled': True, 'fastd-enabled': True, 'adx-enabled': False, 'rsi-enabled': False, 'trigger': 'bb_lower'}, 'sell': {'sell-mfi-value': 75, 'sell-fastd-value': 56, 'sell-adx-value': 61, 'sell-rsi-value': 62, 'sell-mfi-enabled': False, 'sell-fastd-enabled': False, 'sell-adx-enabled': True, 'sell-rsi-enabled': True, 'sell-trigger': 'sell-macd_cross_signal'}, 'roi': {0: 0.3796353801863034, 273: 0.18830463955372825, 887: 0.05307643172744114, 1466: 0}, 'stoploss': {'stoploss': -0.25728526022513887}},  # noqa: E501
             # New Hyperopt mode!
-            'results_metrics': {'total_trades': 117, 'wins': 67, 'draws': 0, 'losses': 50, 'profit_mean': -0.012698609145299145, 'profit_median': -0.012222, 'profit_total': -0.07436117, 'profit_total_abs': -148.573727, 'max_drawdown': 0.52, 'max_drawdown_abs': -224.955321, 'holding_avg': timedelta(minutes=4282.5641025641025)},  # noqa: E501
+            'results_metrics': {'total_trades': 117, 'trade_count_long': 117, 'trade_count_short': 0, 'wins': 67, 'draws': 0, 'losses': 50, 'profit_mean': -0.012698609145299145, 'profit_median': -0.012222, 'profit_total': -0.07436117, 'profit_total_abs': -148.573727, 'max_drawdown': 0.52, 'max_drawdown_abs': -224.955321, 'holding_avg': timedelta(minutes=4282.5641025641025)},  # noqa: E501
             'results_explanation': '   117 trades. Avg profit  -1.27%. Total profit -0.07436117 BTC (-148.57Σ%). Avg duration 4282.6 min.',  # noqa: E501
             'total_profit': -0.07436117,
             'current_epoch': 11,
@@ -2805,7 +2805,7 @@ def saved_hyperopt_results():
             'loss': 100000,
             'params_dict': {'mfi-value': 10, 'fastd-value': 36, 'adx-value': 31, 'rsi-value': 22, 'mfi-enabled': True, 'fastd-enabled': True, 'adx-enabled': True, 'rsi-enabled': False, 'trigger': 'sar_reversal', 'sell-mfi-value': 80, 'sell-fastd-value': 71, 'sell-adx-value': 60, 'sell-rsi-value': 85, 'sell-mfi-enabled': False, 'sell-fastd-enabled': False, 'sell-adx-enabled': True, 'sell-rsi-enabled': True, 'sell-trigger': 'sell-bb_upper', 'roi_t1': 1156, 'roi_t2': 581, 'roi_t3': 408, 'roi_p1': 0.06860454019988212, 'roi_p2': 0.12473718444931989, 'roi_p3': 0.2896360635226823, 'stoploss': -0.30889015124682806},  # noqa: E501
             'params_details': {'buy': {'mfi-value': 10, 'fastd-value': 36, 'adx-value': 31, 'rsi-value': 22, 'mfi-enabled': True, 'fastd-enabled': True, 'adx-enabled': True, 'rsi-enabled': False, 'trigger': 'sar_reversal'}, 'sell': {'sell-mfi-value': 80, 'sell-fastd-value': 71, 'sell-adx-value': 60, 'sell-rsi-value': 85, 'sell-mfi-enabled': False, 'sell-fastd-enabled': False, 'sell-adx-enabled': True, 'sell-rsi-enabled': True, 'sell-trigger': 'sell-bb_upper'}, 'roi': {0: 0.4829777881718843, 408: 0.19334172464920202, 989: 0.06860454019988212, 2145: 0}, 'stoploss': {'stoploss': -0.30889015124682806}},  # noqa: E501
-            'results_metrics': {'total_trades': 0, 'wins': 0, 'draws': 0, 'losses': 0, 'profit_mean': None, 'profit_median': None, 'profit_total': 0, 'profit_total_abs': 0.0, 'max_drawdown': 0.0, 'max_drawdown_abs': 0.0, 'holding_avg': timedelta()},  # noqa: E501
+            'results_metrics': {'total_trades': 0, 'trade_count_long': 0, 'trade_count_short': 0, 'wins': 0, 'draws': 0, 'losses': 0, 'profit_mean': None, 'profit_median': None, 'profit_total': 0, 'profit_total_abs': 0.0, 'max_drawdown': 0.0, 'max_drawdown_abs': 0.0, 'holding_avg': timedelta()},  # noqa: E501
             'results_explanation': '     0 trades. Avg profit    nan%. Total profit  0.00000000 BTC (   0.00Σ%). Avg duration   nan min.',  # noqa: E501
             'total_profit': 0,
             'current_epoch': 12,

tests/exchange/test_exchange.py

@@ -1207,12 +1207,17 @@ def test_create_dry_run_order_fees(
     assert order1['fee']['rate'] == fee
 
 
-@pytest.mark.parametrize("side,startprice,endprice", [
+@pytest.mark.parametrize("side,price,filled", [
-    ("buy", 25.563, 25.566),
+    # order_book_l2_usd spread:
-    ("sell", 25.566, 25.563)
+    # best ask: 25.566
+    # best bid: 25.563
+    ("buy", 25.563, False),
+    ("buy", 25.566, True),
+    ("sell", 25.566, False),
+    ("sell", 25.563, True),
 ])
 @pytest.mark.parametrize("exchange_name", EXCHANGES)
-def test_create_dry_run_order_limit_fill(default_conf, mocker, side, startprice, endprice,
+def test_create_dry_run_order_limit_fill(default_conf, mocker, side, price, filled,
                                          exchange_name, order_book_l2_usd):
     default_conf['dry_run'] = True
     exchange = get_patched_exchange(mocker, default_conf, id=exchange_name)
@@ -1226,7 +1231,7 @@ def test_create_dry_run_order_limit_fill(default_conf, mocker, side, startprice,
         ordertype='limit',
         side=side,
         amount=1,
-        rate=startprice,
+        rate=price,
         leverage=1.0
     )
     assert order_book_l2_usd.call_count == 1
@@ -1235,22 +1240,17 @@ def test_create_dry_run_order_limit_fill(default_conf, mocker, side, startprice,
     assert order["side"] == side
     assert order["type"] == "limit"
     assert order["symbol"] == "LTC/USDT"
+    assert order["average"] == price
+    assert order['status'] == 'open' if not filled else 'closed'
     order_book_l2_usd.reset_mock()
 
+    # fetch order again...
     order_closed = exchange.fetch_dry_run_order(order['id'])
-    assert order_book_l2_usd.call_count == 1
+    assert order_book_l2_usd.call_count == (1 if not filled else 0)
-    assert order_closed['status'] == 'open'
+    assert order_closed['status'] == ('open' if not filled else 'closed')
-    assert not order['fee']
+    assert order_closed['filled'] == (0 if not filled else 1)
-    assert order_closed['filled'] == 0
 
     order_book_l2_usd.reset_mock()
-    order_closed['price'] = endprice
-
-    order_closed = exchange.fetch_dry_run_order(order['id'])
-    assert order_closed['status'] == 'closed'
-    assert order['fee']
-    assert order_closed['filled'] == 1
-    assert order_closed['filled'] == order_closed['amount']
 
     # Empty orderbook test
     mocker.patch('freqtrade.exchange.Exchange.fetch_l2_order_book',

tests/freqai/conftest.py

@@ -27,10 +27,9 @@ def freqai_conf(default_conf, tmpdir):
             "timerange": "20180110-20180115",
             "freqai": {
                 "enabled": True,
-                "startup_candles": 10000,
                 "purge_old_models": True,
                 "train_period_days": 2,
-                "backtest_period_days": 2,
+                "backtest_period_days": 10,
                 "live_retrain_hours": 0,
                 "expiration_hours": 1,
                 "identifier": "uniqe-id100",
@@ -58,6 +57,30 @@ def freqai_conf(default_conf, tmpdir):
     return freqaiconf
 
 
+def make_rl_config(conf):
+    conf.update({"strategy": "freqai_rl_test_strat"})
+    conf["freqai"].update({"model_training_parameters": {
+        "learning_rate": 0.00025,
+        "gamma": 0.9,
+        "verbose": 1
+    }})
+    conf["freqai"]["rl_config"] = {
+        "train_cycles": 1,
+        "thread_count": 2,
+        "max_trade_duration_candles": 300,
+        "model_type": "PPO",
+        "policy_type": "MlpPolicy",
+        "max_training_drawdown_pct": 0.5,
+        "net_arch": [32, 32],
+        "model_reward_parameters": {
+            "rr": 1,
+            "profit_aim": 0.02,
+            "win_reward_factor": 2
+        }}
+
+    return conf
+
+
 def get_patched_data_kitchen(mocker, freqaiconf):
     dk = FreqaiDataKitchen(freqaiconf)
     return dk

tests/freqai/test_freqai_interface.py

@@ -13,8 +13,8 @@ from freqtrade.freqai.utils import download_all_data_for_training, get_required_
 from freqtrade.optimize.backtesting import Backtesting
 from freqtrade.persistence import Trade
 from freqtrade.plugins.pairlistmanager import PairListManager
-from tests.conftest import get_patched_exchange, log_has_re
+from tests.conftest import create_mock_trades, get_patched_exchange, log_has_re
-from tests.freqai.conftest import get_patched_freqai_strategy
+from tests.freqai.conftest import get_patched_freqai_strategy, make_rl_config
 
 
 def is_arm() -> bool:
@@ -32,11 +32,17 @@ def is_mac() -> bool:
     ('XGBoostRegressor', False, True, False),
     ('XGBoostRFRegressor', False, False, False),
     ('CatboostRegressor', False, False, False),
+    ('ReinforcementLearner', False, True, False),
+    ('ReinforcementLearner_multiproc', False, False, False),
+    ('ReinforcementLearner_test_4ac', False, False, False)
     ])
 def test_extract_data_and_train_model_Standard(mocker, freqai_conf, model, pca, dbscan, float32):
     if is_arm() and model == 'CatboostRegressor':
         pytest.skip("CatBoost is not supported on ARM")
 
+    if is_mac() and 'Reinforcement' in model:
+        pytest.skip("Reinforcement learning module not available on intel based Mac OS")
+
     model_save_ext = 'joblib'
     freqai_conf.update({"freqaimodel": model})
     freqai_conf.update({"timerange": "20180110-20180130"})
@@ -45,6 +51,26 @@ def test_extract_data_and_train_model_Standard(mocker, freqai_conf, model, pca,
     freqai_conf['freqai']['feature_parameters'].update({"use_DBSCAN_to_remove_outliers": dbscan})
     freqai_conf.update({"reduce_df_footprint": float32})
 
+    if 'ReinforcementLearner' in model:
+        model_save_ext = 'zip'
+        freqai_conf = make_rl_config(freqai_conf)
+        # test the RL guardrails
+        freqai_conf['freqai']['feature_parameters'].update({"use_SVM_to_remove_outliers": True})
+        freqai_conf['freqai']['data_split_parameters'].update({'shuffle': True})
+
+    if 'test_4ac' in model:
+        freqai_conf["freqaimodel_path"] = str(Path(__file__).parents[1] / "freqai" / "test_models")
+
+    if 'ReinforcementLearner' in model:
+        model_save_ext = 'zip'
+        freqai_conf = make_rl_config(freqai_conf)
+        # test the RL guardrails
+        freqai_conf['freqai']['feature_parameters'].update({"use_SVM_to_remove_outliers": True})
+        freqai_conf['freqai']['data_split_parameters'].update({'shuffle': True})
+
+    if 'test_4ac' in model:
+        freqai_conf["freqaimodel_path"] = str(Path(__file__).parents[1] / "freqai" / "test_models")
+
     strategy = get_patched_freqai_strategy(mocker, freqai_conf)
     exchange = get_patched_exchange(mocker, freqai_conf)
     strategy.dp = DataProvider(freqai_conf, exchange)
@@ -52,6 +78,7 @@ def test_extract_data_and_train_model_Standard(mocker, freqai_conf, model, pca,
     freqai = strategy.freqai
     freqai.live = True
     freqai.dk = FreqaiDataKitchen(freqai_conf)
+    freqai.dk.set_paths('ADA/BTC', 10000)
     timerange = TimeRange.parse_timerange("20180110-20180130")
     freqai.dd.load_all_pair_histories(timerange, freqai.dk)
 
@@ -165,25 +192,35 @@ def test_extract_data_and_train_model_Classifiers(mocker, freqai_conf, model):
 @pytest.mark.parametrize(
     "model, num_files, strat",
     [
-        ("LightGBMRegressor", 6, "freqai_test_strat"),
+        ("LightGBMRegressor", 2, "freqai_test_strat"),
-        ("XGBoostRegressor", 6, "freqai_test_strat"),
+        ("XGBoostRegressor", 2, "freqai_test_strat"),
-        ("CatboostRegressor", 6, "freqai_test_strat"),
+        ("CatboostRegressor", 2, "freqai_test_strat"),
-        ("XGBoostClassifier", 6, "freqai_test_classifier"),
+        ("ReinforcementLearner", 3, "freqai_rl_test_strat"),
-        ("LightGBMClassifier", 6, "freqai_test_classifier"),
+        ("XGBoostClassifier", 2, "freqai_test_classifier"),
-        ("CatboostClassifier", 6, "freqai_test_classifier")
+        ("LightGBMClassifier", 2, "freqai_test_classifier"),
+        ("CatboostClassifier", 2, "freqai_test_classifier")
     ],
     )
 def test_start_backtesting(mocker, freqai_conf, model, num_files, strat, caplog):
     freqai_conf.get("freqai", {}).update({"save_backtest_models": True})
     freqai_conf['runmode'] = RunMode.BACKTEST
-    Trade.use_db = False
     if is_arm() and "Catboost" in model:
         pytest.skip("CatBoost is not supported on ARM")
 
+    if is_mac() and 'Reinforcement' in model:
+        pytest.skip("Reinforcement learning module not available on intel based Mac OS")
+    Trade.use_db = False
+
     freqai_conf.update({"freqaimodel": model})
     freqai_conf.update({"timerange": "20180120-20180130"})
     freqai_conf.update({"strategy": strat})
 
+    if 'ReinforcementLearner' in model:
+        freqai_conf = make_rl_config(freqai_conf)
+
+    if 'test_4ac' in model:
+        freqai_conf["freqaimodel_path"] = str(Path(__file__).parents[1] / "freqai" / "test_models")
+
     strategy = get_patched_freqai_strategy(mocker, freqai_conf)
     exchange = get_patched_exchange(mocker, freqai_conf)
     strategy.dp = DataProvider(freqai_conf, exchange)
@@ -207,6 +244,7 @@ def test_start_backtesting(mocker, freqai_conf, model, num_files, strat, caplog)
     model_folders = [x for x in freqai.dd.full_path.iterdir() if x.is_dir()]
 
     assert len(model_folders) == num_files
+    Trade.use_db = True
     assert log_has_re(
         "Removed features ",
         caplog,
@@ -269,7 +307,7 @@ def test_start_backtesting_from_existing_folder(mocker, freqai_conf, caplog):
     freqai.start_backtesting(df, metadata, freqai.dk)
     model_folders = [x for x in freqai.dd.full_path.iterdir() if x.is_dir()]
 
-    assert len(model_folders) == 6
+    assert len(model_folders) == 2
 
     # without deleting the existing folder structure, re-run
 
@@ -507,3 +545,43 @@ def test_download_all_data_for_training(mocker, freqai_conf, caplog, tmpdir):
         "Downloading",
         caplog,
     )
+
+
+@pytest.mark.usefixtures("init_persistence")
+@pytest.mark.parametrize('dp_exists', [(False), (True)])
+def test_get_state_info(mocker, freqai_conf, dp_exists, caplog, tickers):
+
+    if is_mac():
+        pytest.skip("Reinforcement learning module not available on intel based Mac OS")
+
+    freqai_conf.update({"freqaimodel": "ReinforcementLearner"})
+    freqai_conf.update({"timerange": "20180110-20180130"})
+    freqai_conf.update({"strategy": "freqai_rl_test_strat"})
+    freqai_conf = make_rl_config(freqai_conf)
+    freqai_conf['entry_pricing']['price_side'] = 'same'
+    freqai_conf['exit_pricing']['price_side'] = 'same'
+
+    strategy = get_patched_freqai_strategy(mocker, freqai_conf)
+    exchange = get_patched_exchange(mocker, freqai_conf)
+    ticker_mock = MagicMock(return_value=tickers()['ETH/BTC'])
+    mocker.patch("freqtrade.exchange.Exchange.fetch_ticker", ticker_mock)
+    strategy.dp = DataProvider(freqai_conf, exchange)
+
+    if not dp_exists:
+        strategy.dp._exchange = None
+
+    strategy.freqai_info = freqai_conf.get("freqai", {})
+    freqai = strategy.freqai
+    freqai.data_provider = strategy.dp
+    freqai.live = True
+
+    Trade.use_db = True
+    create_mock_trades(MagicMock(return_value=0.0025), False, True)
+    freqai.get_state_info("ADA/BTC")
+    freqai.get_state_info("ETH/BTC")
+
+    if not dp_exists:
+        assert log_has_re(
+            "No exchange available",
+            caplog,
+        )

tests/freqai/test_models/ReinforcementLearner_test_4ac.py

@@ -0,0 +1,66 @@
+import logging
+
+import numpy as np
+
+from freqtrade.freqai.prediction_models.ReinforcementLearner import ReinforcementLearner
+from freqtrade.freqai.RL.Base4ActionRLEnv import Actions, Base4ActionRLEnv, Positions
+
+
+logger = logging.getLogger(__name__)
+
+
+class ReinforcementLearner_test_4ac(ReinforcementLearner):
+    """
+    User created Reinforcement Learning Model prediction model.
+    """
+
+    class MyRLEnv(Base4ActionRLEnv):
+        """
+        User can override any function in BaseRLEnv and gym.Env. Here the user
+        sets a custom reward based on profit and trade duration.
+        """
+
+        def calculate_reward(self, action: int) -> float:
+
+            # first, penalize if the action is not valid
+            if not self._is_valid(action):
+                return -2
+
+            pnl = self.get_unrealized_profit()
+            rew = np.sign(pnl) * (pnl + 1)
+            factor = 100.
+
+            # reward agent for entering trades
+            if (action in (Actions.Long_enter.value, Actions.Short_enter.value)
+                    and self._position == Positions.Neutral):
+                return 25
+            # discourage agent from not entering trades
+            if action == Actions.Neutral.value and self._position == Positions.Neutral:
+                return -1
+
+            max_trade_duration = self.rl_config.get('max_trade_duration_candles', 300)
+            trade_duration = self._current_tick - self._last_trade_tick  # type: ignore
+
+            if trade_duration <= max_trade_duration:
+                factor *= 1.5
+            elif trade_duration > max_trade_duration:
+                factor *= 0.5
+
+            # discourage sitting in position
+            if (self._position in (Positions.Short, Positions.Long) and
+                    action == Actions.Neutral.value):
+                return -1 * trade_duration / max_trade_duration
+
+            # close long
+            if action == Actions.Exit.value and self._position == Positions.Long:
+                if pnl > self.profit_aim * self.rr:
+                    factor *= self.rl_config['model_reward_parameters'].get('win_reward_factor', 2)
+                return float(rew * factor)
+
+            # close short
+            if action == Actions.Exit.value and self._position == Positions.Short:
+                if pnl > self.profit_aim * self.rr:
+                    factor *= self.rl_config['model_reward_parameters'].get('win_reward_factor', 2)
+                return float(rew * factor)
+
+            return 0.

tests/optimize/test_backtesting.py

@@ -663,30 +663,9 @@ def test_backtest__get_sell_trade_entry(default_conf, fee, mocker) -> None:
         '',  # Exit Signal Name
 
     ]
-    row_detail = pd.DataFrame(
-        [
-            [
-                pd.Timestamp(year=2020, month=1, day=1, hour=5, minute=0, tzinfo=timezone.utc),
-                200, 200.1, 197, 199, 1, 0, 0, 0, '', '', '',
-            ], [
-                pd.Timestamp(year=2020, month=1, day=1, hour=5, minute=1, tzinfo=timezone.utc),
-                199, 199.7, 199, 199.5, 0, 0, 0, 0, '', '', '',
-            ], [
-                pd.Timestamp(year=2020, month=1, day=1, hour=5, minute=2, tzinfo=timezone.utc),
-                199.5, 200.8, 199, 200.9, 0, 0, 0, 0, '', '', '',
-            ], [
-                pd.Timestamp(year=2020, month=1, day=1, hour=5, minute=3, tzinfo=timezone.utc),
-                200.5, 210.5, 193, 210.5, 0, 0, 0, 0, '', '', '',  # ROI sell (?)
-            ], [
-                pd.Timestamp(year=2020, month=1, day=1, hour=5, minute=4, tzinfo=timezone.utc),
-                200, 200.1, 193, 199, 0, 0, 0, 0, '', '', '',
-            ],
-        ], columns=['date', 'open', 'high', 'low', 'close', 'enter_long', 'exit_long',
-                    'enter_short', 'exit_short', 'long_tag', 'short_tag', 'exit_tag']
-    )
 
     # No data available.
-    res = backtesting._get_exit_trade_entry(trade, row_sell)
+    res = backtesting._get_exit_trade_entry(trade, row_sell, True)
     assert res is not None
     assert res.exit_reason == ExitType.ROI.value
     assert res.close_date_utc == datetime(2020, 1, 1, 5, 0, tzinfo=timezone.utc)
@@ -699,20 +678,9 @@ def test_backtest__get_sell_trade_entry(default_conf, fee, mocker) -> None:
         [], columns=['date', 'open', 'high', 'low', 'close', 'enter_long', 'exit_long',
                      'enter_short', 'exit_short', 'long_tag', 'short_tag', 'exit_tag'])
 
-    res = backtesting._get_exit_trade_entry(trade, row)
+    res = backtesting._get_exit_trade_entry(trade, row, True)
     assert res is None
 
-    # Assign backtest-detail data
-    backtesting.detail_data[pair] = row_detail
-
-    res = backtesting._get_exit_trade_entry(trade, row_sell)
-    assert res is not None
-    assert res.exit_reason == ExitType.ROI.value
-    # Sell at minute 3 (not available above!)
-    assert res.close_date_utc == datetime(2020, 1, 1, 5, 3, tzinfo=timezone.utc)
-    sell_order = res.select_order('sell', True)
-    assert sell_order is not None
-
 
 def test_backtest_one(default_conf, fee, mocker, testdatadir) -> None:
     default_conf['use_exit_signal'] = False
@@ -788,17 +756,98 @@ def test_backtest_one(default_conf, fee, mocker, testdatadir) -> None:
     for _, t in results.iterrows():
         assert len(t['orders']) == 2
         ln = data_pair.loc[data_pair["date"] == t["open_date"]]
-        # Check open trade rate alignes to open rate
+        # Check open trade rate aligns to open rate
         assert not ln.empty
         assert round(ln.iloc[0]["open"], 6) == round(t["open_rate"], 6)
-        # check close trade rate alignes to close rate or is between high and low
+        # check close trade rate aligns to close rate or is between high and low
         ln1 = data_pair.loc[data_pair["date"] == t["close_date"]]
-        assert not ln1.empty
         assert (round(ln1.iloc[0]["open"], 6) == round(t["close_rate"], 6) or
                 round(ln1.iloc[0]["low"], 6) < round(
                 t["close_rate"], 6) < round(ln1.iloc[0]["high"], 6))
 
 
+@pytest.mark.parametrize('use_detail', [True, False])
+def test_backtest_one_detail(default_conf_usdt, fee, mocker, testdatadir, use_detail) -> None:
+    default_conf_usdt['use_exit_signal'] = False
+    mocker.patch('freqtrade.exchange.Exchange.get_fee', fee)
+    mocker.patch("freqtrade.exchange.Exchange.get_min_pair_stake_amount", return_value=0.00001)
+    mocker.patch("freqtrade.exchange.Exchange.get_max_pair_stake_amount", return_value=float('inf'))
+    if use_detail:
+        default_conf_usdt['timeframe_detail'] = '1m'
+    patch_exchange(mocker)
+
+    def advise_entry(df, *args, **kwargs):
+        # Mock function to force several entries
+        df.loc[(df['rsi'] < 40), 'enter_long'] = 1
+        return df
+
+    def custom_entry_price(proposed_rate, **kwargs):
+        return proposed_rate * 0.997
+
+    backtesting = Backtesting(default_conf_usdt)
+    backtesting._set_strategy(backtesting.strategylist[0])
+    backtesting.strategy.populate_entry_trend = advise_entry
+    backtesting.strategy.custom_entry_price = custom_entry_price
+    pair = 'XRP/ETH'
+    # Pick a timerange adapted to the pair we use to test
+    timerange = TimeRange.parse_timerange('20191010-20191013')
+    data = history.load_data(datadir=testdatadir, timeframe='5m', pairs=['XRP/ETH'],
+                             timerange=timerange)
+    if use_detail:
+        data_1m = history.load_data(datadir=testdatadir, timeframe='1m', pairs=['XRP/ETH'],
+                                    timerange=timerange)
+        backtesting.detail_data = data_1m
+    processed = backtesting.strategy.advise_all_indicators(data)
+    min_date, max_date = get_timerange(processed)
+
+    result = backtesting.backtest(
+        processed=deepcopy(processed),
+        start_date=min_date,
+        end_date=max_date,
+        max_open_trades=10,
+    )
+    results = result['results']
+    assert not results.empty
+    # Timeout settings from default_conf = entry: 10, exit: 30
+    assert len(results) == (2 if use_detail else 3)
+
+    assert 'orders' in results.columns
+    data_pair = processed[pair]
+
+    data_1m_pair = data_1m[pair] if use_detail else pd.DataFrame()
+    late_entry = 0
+    for _, t in results.iterrows():
+        assert len(t['orders']) == 2
+
+        entryo = t['orders'][0]
+        entry_ts = datetime.fromtimestamp(entryo['order_filled_timestamp'] // 1000, tz=timezone.utc)
+        if entry_ts > t['open_date']:
+            late_entry += 1
+
+        # Get "entry fill" candle
+        ln = (data_1m_pair.loc[data_1m_pair["date"] == entry_ts]
+              if use_detail else data_pair.loc[data_pair["date"] == entry_ts])
+        # Check open trade rate aligns to open rate
+        assert not ln.empty
+
+        # assert round(ln.iloc[0]["open"], 6) == round(t["open_rate"], 6)
+        assert round(ln.iloc[0]["low"], 6) <= round(
+                t["open_rate"], 6) <= round(ln.iloc[0]["high"], 6)
+        # check close trade rate aligns to close rate or is between high and low
+        ln1 = data_pair.loc[data_pair["date"] == t["close_date"]]
+        if use_detail:
+            ln1_1m = data_1m_pair.loc[data_1m_pair["date"] == t["close_date"]]
+            assert not ln1.empty or not ln1_1m.empty
+        else:
+            assert not ln1.empty
+        ln2 = ln1_1m if ln1.empty else ln1
+
+        assert (round(ln2.iloc[0]["low"], 6) <= round(
+                t["close_rate"], 6) <= round(ln2.iloc[0]["high"], 6))
+
+    assert late_entry > 0
+
+
 def test_backtest_timedout_entry_orders(default_conf, fee, mocker, testdatadir) -> None:
     # This strategy intentionally places unfillable orders.
     default_conf['strategy'] = 'StrategyTestV3CustomEntryPrice'

tests/rpc/test_rpc_apiserver.py

@@ -57,7 +57,10 @@ def botclient(default_conf, mocker):
     try:
         apiserver = ApiServer(default_conf)
         apiserver.add_rpc_handler(rpc)
-        yield ftbot, TestClient(apiserver.app)
+        # We need to use the TestClient as a context manager to
+        # handle lifespan events correctly
+        with TestClient(apiserver.app) as client:
+            yield ftbot, client
         # Cleanup ... ?
     finally:
         if apiserver:
@@ -438,7 +441,6 @@ def test_api_cleanup(default_conf, mocker, caplog):
     apiserver.cleanup()
     assert apiserver._server.cleanup.call_count == 1
     assert log_has("Stopping API Server", caplog)
-    assert log_has("Stopping API Server background tasks", caplog)
     ApiServer.shutdown()
 
 
@@ -1459,6 +1461,7 @@ def test_api_strategies(botclient, tmpdir):
         'StrategyTestV3',
         'StrategyTestV3CustomEntryPrice',
         'StrategyTestV3Futures',
+        'freqai_rl_test_strat',
         'freqai_test_classifier',
         'freqai_test_multimodel_classifier_strat',
         'freqai_test_multimodel_strat',
@@ -1714,12 +1717,14 @@ def test_api_ws_subscribe(botclient, mocker):
 
     with client.websocket_connect(ws_url) as ws:
         ws.send_json({'type': 'subscribe', 'data': ['whitelist']})
+        time.sleep(1)
 
     # Check call count is now 1 as we sent a valid subscribe request
     assert sub_mock.call_count == 1
 
     with client.websocket_connect(ws_url) as ws:
         ws.send_json({'type': 'subscribe', 'data': 'whitelist'})
+        time.sleep(1)
 
     # Call count hasn't changed as the subscribe request was invalid
     assert sub_mock.call_count == 1
@@ -1773,24 +1778,18 @@ def test_api_ws_send_msg(default_conf, mocker, caplog):
         mocker.patch('freqtrade.rpc.api_server.ApiServer.start_api')
         apiserver = ApiServer(default_conf)
         apiserver.add_rpc_handler(RPC(get_patched_freqtradebot(mocker, default_conf)))
-        apiserver.start_message_queue()
-        # Give the queue thread time to start
-        time.sleep(0.2)
 
-        # Test message_queue coro receives the message
+        # Start test client context manager to run lifespan events
-        test_message = {"type": "status", "data": "test"}
+        with TestClient(apiserver.app):
-        apiserver.send_msg(test_message)
+            # Test message is published on the Message Stream
-        time.sleep(0.1)  # Not sure how else to wait for the coro to receive the data
+            test_message = {"type": "status", "data": "test"}
-        assert log_has("Found message of type: status", caplog)
+            first_waiter = apiserver._message_stream._waiter
+            apiserver.send_msg(test_message)
+            assert first_waiter.result()[0] == test_message
 
-        # Test if exception logged when error occurs in sending
+            second_waiter = apiserver._message_stream._waiter
-        mocker.patch('freqtrade.rpc.api_server.ws.channel.ChannelManager.broadcast',
+            apiserver.send_msg(test_message)
-                     side_effect=Exception)
+            assert first_waiter != second_waiter
-
-        apiserver.send_msg(test_message)
-        time.sleep(0.1)  # Not sure how else to wait for the coro to receive the data
-        assert log_has_re(r"Exception happened in background task.*", caplog)
 
     finally:
-        apiserver.cleanup()
         ApiServer.shutdown()

tests/strategy/strats/freqai_rl_test_strat.py

@@ -0,0 +1,105 @@
+import logging
+from functools import reduce
+
+import pandas as pd
+import talib.abstract as ta
+from pandas import DataFrame
+
+from freqtrade.strategy import IStrategy, merge_informative_pair
+
+
+logger = logging.getLogger(__name__)
+
+
+class freqai_rl_test_strat(IStrategy):
+    """
+    Test strategy - used for testing freqAI functionalities.
+    DO not use in production.
+    """
+
+    minimal_roi = {"0": 0.1, "240": -1}
+
+    process_only_new_candles = True
+    stoploss = -0.05
+    use_exit_signal = True
+    startup_candle_count: int = 30
+    can_short = False
+
+    def populate_any_indicators(
+        self, pair, df, tf, informative=None, set_generalized_indicators=False
+    ):
+
+        if informative is None:
+            informative = self.dp.get_pair_dataframe(pair, tf)
+
+        # first loop is automatically duplicating indicators for time periods
+        for t in self.freqai_info["feature_parameters"]["indicator_periods_candles"]:
+
+            t = int(t)
+            informative[f"%-{pair}rsi-period_{t}"] = ta.RSI(informative, timeperiod=t)
+
+        # The following columns are necessary for RL models.
+        informative[f"%-{pair}raw_close"] = informative["close"]
+        informative[f"%-{pair}raw_open"] = informative["open"]
+        informative[f"%-{pair}raw_high"] = informative["high"]
+        informative[f"%-{pair}raw_low"] = informative["low"]
+
+        indicators = [col for col in informative if col.startswith("%")]
+        # This loop duplicates and shifts all indicators to add a sense of recency to data
+        for n in range(self.freqai_info["feature_parameters"]["include_shifted_candles"] + 1):
+            if n == 0:
+                continue
+            informative_shift = informative[indicators].shift(n)
+            informative_shift = informative_shift.add_suffix("_shift-" + str(n))
+            informative = pd.concat((informative, informative_shift), axis=1)
+
+        df = merge_informative_pair(df, informative, self.config["timeframe"], tf, ffill=True)
+        skip_columns = [
+            (s + "_" + tf) for s in ["date", "open", "high", "low", "close", "volume"]
+        ]
+        df = df.drop(columns=skip_columns)
+
+        # Add generalized indicators here (because in live, it will call this
+        # function to populate indicators during training). Notice how we ensure not to
+        # add them multiple times
+        if set_generalized_indicators:
+            # For RL, there are no direct targets to set. This is filler (neutral)
+            # until the agent sends an action.
+            df["&-action"] = 0
+
+        return df
+
+    def populate_indicators(self, dataframe: DataFrame, metadata: dict) -> DataFrame:
+
+        dataframe = self.freqai.start(dataframe, metadata, self)
+
+        return dataframe
+
+    def populate_entry_trend(self, df: DataFrame, metadata: dict) -> DataFrame:
+
+        enter_long_conditions = [df["do_predict"] == 1, df["&-action"] == 1]
+
+        if enter_long_conditions:
+            df.loc[
+                reduce(lambda x, y: x & y, enter_long_conditions), ["enter_long", "enter_tag"]
+            ] = (1, "long")
+
+        enter_short_conditions = [df["do_predict"] == 1, df["&-action"] == 3]
+
+        if enter_short_conditions:
+            df.loc[
+                reduce(lambda x, y: x & y, enter_short_conditions), ["enter_short", "enter_tag"]
+            ] = (1, "short")
+
+        return df
+
+    def populate_exit_trend(self, df: DataFrame, metadata: dict) -> DataFrame:
+        exit_long_conditions = [df["do_predict"] == 1, df["&-action"] == 2]
+        if exit_long_conditions:
+            df.loc[reduce(lambda x, y: x & y, exit_long_conditions), "exit_long"] = 1
+
+        exit_short_conditions = [df["do_predict"] == 1, df["&-action"] == 4]
+        if exit_short_conditions:
+            df.loc[reduce(lambda x, y: x & y, exit_short_conditions), "exit_short"] = 1
+
+        return df

tests/strategy/test_strategy_loading.py

@@ -34,7 +34,7 @@ def test_search_all_strategies_no_failed():
     directory = Path(__file__).parent / "strats"
     strategies = StrategyResolver._search_all_objects(directory, enum_failed=False)
     assert isinstance(strategies, list)
-    assert len(strategies) == 11
+    assert len(strategies) == 12
     assert isinstance(strategies[0], dict)
 
 
@@ -42,10 +42,10 @@ def test_search_all_strategies_with_failed():
     directory = Path(__file__).parent / "strats"
     strategies = StrategyResolver._search_all_objects(directory, enum_failed=True)
     assert isinstance(strategies, list)
-    assert len(strategies) == 12
+    assert len(strategies) == 13
     # with enum_failed=True search_all_objects() shall find 2 good strategies
     # and 1 which fails to load
-    assert len([x for x in strategies if x['class'] is not None]) == 11
+    assert len([x for x in strategies if x['class'] is not None]) == 12
 
     assert len([x for x in strategies if x['class'] is None]) == 1
 

tests/test_freqtradebot.py

@@ -1498,6 +1498,7 @@ def test_handle_stoploss_on_exchange_trailing(
         })
     )
     assert freqtrade.handle_trade(trade) is True
+    assert trade.stoploss_order_id is None
 
 
 @pytest.mark.parametrize("is_short", [False, True])
@@ -5046,7 +5047,7 @@ def test_startup_backpopulate_precision(mocker, default_conf_usdt, fee, caplog):
 
 @pytest.mark.usefixtures("init_persistence")
 @pytest.mark.parametrize("is_short", [False, True])
-def test_update_closed_trades_without_assigned_fees(mocker, default_conf_usdt, fee, is_short):
+def test_update_trades_without_assigned_fees(mocker, default_conf_usdt, fee, is_short):
     freqtrade = get_patched_freqtradebot(mocker, default_conf_usdt)
 
     def patch_with_fee(order):
@@ -5075,7 +5076,7 @@ def test_update_closed_trades_without_assigned_fees(mocker, default_conf_usdt, f
         assert trade.fee_close_cost is None
         assert trade.fee_close_currency is None
 
-    freqtrade.update_closed_trades_without_assigned_fees()
+    freqtrade.update_trades_without_assigned_fees()
 
     # Does nothing for dry-run
     trades = Trade.get_trades().all()
@@ -5088,7 +5089,7 @@ def test_update_closed_trades_without_assigned_fees(mocker, default_conf_usdt, f
 
     freqtrade.config['dry_run'] = False
 
-    freqtrade.update_closed_trades_without_assigned_fees()
+    freqtrade.update_trades_without_assigned_fees()
 
     trades = Trade.get_trades().all()
     assert len(trades) == MOCK_TRADE_COUNT
@@ -5551,7 +5552,7 @@ def test_position_adjust(mocker, default_conf_usdt, fee) -> None:
     assert trade.stake_amount == 110
 
     # Assume it does nothing since order is closed and trade is open
-    freqtrade.update_closed_trades_without_assigned_fees()
+    freqtrade.update_trades_without_assigned_fees()
 
     trade = Trade.query.first()
     assert trade
@@ -5622,7 +5623,7 @@ def test_position_adjust(mocker, default_conf_usdt, fee) -> None:
     mocker.patch('freqtrade.exchange.Exchange.create_order', fetch_order_mm)
     mocker.patch('freqtrade.exchange.Exchange.fetch_order', fetch_order_mm)
     mocker.patch('freqtrade.exchange.Exchange.fetch_order_or_stoploss_order', fetch_order_mm)
-    freqtrade.update_closed_trades_without_assigned_fees()
+    freqtrade.update_trades_without_assigned_fees()
 
     orders = Order.query.all()
     assert orders
@@ -5839,7 +5840,7 @@ def test_position_adjust2(mocker, default_conf_usdt, fee) -> None:
     assert trade.stake_amount == bid * amount
 
     # Assume it does nothing since order is closed and trade is open
-    freqtrade.update_closed_trades_without_assigned_fees()
+    freqtrade.update_trades_without_assigned_fees()
 
     trade = Trade.query.first()
     assert trade