stable/docs/data-analysis.md

213 lines
7.1 KiB
Markdown
Raw Normal View History

2019-08-11 16:52:37 +00:00
# Analyzing bot data with Jupyter notebooks
2019-06-22 14:18:22 +00:00
2019-08-11 16:52:37 +00:00
You can analyze the results of backtests and trading history easily using Jupyter notebooks. Sample notebooks are located at `user_data/notebooks/`.
2019-06-22 14:18:22 +00:00
2019-08-11 16:52:37 +00:00
## Pro tips
2019-06-22 14:18:22 +00:00
2019-08-11 16:52:37 +00:00
* See [jupyter.org](https://jupyter.org/documentation) for usage instructions.
2019-08-20 04:47:10 +00:00
* Don't forget to start a Jupyter notebook server from within your conda or venv environment or use [nb_conda_kernels](https://github.com/Anaconda-Platform/nb_conda_kernels)*
* Copy the example notebook before use so your changes don't get clobbered with the next freqtrade update.
2019-08-11 16:52:37 +00:00
## Fine print
2019-08-20 04:47:10 +00:00
Some tasks don't work especially well in notebooks. For example, anything using asynchronous execution is a problem for Jupyter. Also, freqtrade's primary entry point is the shell cli, so using pure python in a notebook bypasses arguments that provide required objects and parameters to helper functions. You may need to set those values or create expected objects manually.
2019-08-11 16:52:37 +00:00
## Recommended workflow
| Task | Tool |
--- | ---
Bot operations | CLI
2019-08-20 04:47:10 +00:00
Repetitive tasks | Shell scripts
2019-08-11 16:52:37 +00:00
Data analysis & visualization | Notebook
1. Use the CLI to
* download historical data
* run a backtest
* run with real-time data
* export results
1. Collect these actions in shell scripts
* save complicated commands with arguments
2019-08-20 04:47:10 +00:00
* execute multi-step operations
* automate testing strategies and preparing data for analysis
2019-08-11 16:52:37 +00:00
1. Use a notebook to
2019-08-20 04:47:10 +00:00
* visualize data
2019-08-11 16:52:37 +00:00
* munge and plot to generate insights
2019-08-20 04:47:10 +00:00
## Example utility snippets
2019-08-11 16:52:37 +00:00
### Change directory to root
2019-08-20 04:47:10 +00:00
Jupyter notebooks execute from the notebook directory. The following snippet searches for the project root, so relative paths remain consistent.
2019-08-11 16:52:37 +00:00
```python
import os
from pathlib import Path
2019-08-20 04:47:10 +00:00
# Change directory
# Modify this cell to insure that the output shows the correct path.
# Define all paths relative to the project root shown in the cell output
2019-08-11 16:52:37 +00:00
project_root = "somedir/freqtrade"
i=0
try:
os.chdirdir(project_root)
assert Path('LICENSE').is_file()
except:
while i<4 and (not Path('LICENSE').is_file()):
os.chdir(Path(Path.cwd(), '../'))
i+=1
project_root = Path.cwd()
print(Path.cwd())
```
## Load existing objects into a Jupyter notebook
2019-08-20 04:47:10 +00:00
These examples assume that you have already generated data using the cli. They will allow you to drill deeper into your results, and perform analysis which otherwise would make the output very difficult to digest due to information overload.
2019-06-22 14:18:22 +00:00
2019-08-09 02:09:15 +00:00
### Load backtest results into a pandas dataframe
```python
2019-08-09 21:06:19 +00:00
from freqtrade.data.btanalysis import load_backtest_data
2019-08-20 04:47:10 +00:00
2019-08-09 02:09:15 +00:00
# Load backtest results
df = load_backtest_data("user_data/backtest_results/backtest-result.json")
2019-08-09 02:09:15 +00:00
# Show value-counts per pair
df.groupby("pair")["sell_reason"].value_counts()
```
2019-08-02 17:50:12 +00:00
2019-08-09 02:09:15 +00:00
### Load live trading results into a pandas dataframe
2019-08-09 02:09:15 +00:00
``` python
2019-08-09 21:06:19 +00:00
from freqtrade.data.btanalysis import load_trades_from_db
2019-08-09 02:09:15 +00:00
# Fetch trades from database
df = load_trades_from_db("sqlite:///tradesv3.sqlite")
2019-08-09 02:09:15 +00:00
# Display results
df.groupby("pair")["sell_reason"].value_counts()
```
2019-08-10 17:55:33 +00:00
### Load multiple configuration files
2019-09-13 05:08:42 +00:00
This option can be useful to inspect the results of passing in multiple configs.
This will also run through the whole Configuration initialization, so the configuration is completely initialized to be passed to other methods.
2019-08-10 17:55:33 +00:00
``` python
2019-08-20 04:47:10 +00:00
import json
2019-08-10 17:55:33 +00:00
from freqtrade.configuration import Configuration
2019-08-20 04:47:10 +00:00
# Load config from multiple files
2019-08-10 17:55:33 +00:00
config = Configuration.from_files(["config1.json", "config2.json"])
2019-08-11 16:52:37 +00:00
# Show the config in memory
print(json.dumps(config['original_config'], indent=2))
2019-09-13 05:08:42 +00:00
```
For Interactive environments, have an additional configuration specifying `user_data_dir` and pass this in last, so you don't have to change directories while running the bot.
2019-09-14 07:54:40 +00:00
Best avoid relative paths, since this starts at the storage location of the jupyter notebook, unless the directory is changed.
2019-09-13 05:08:42 +00:00
``` json
{
"user_data_dir": "~/.freqtrade/"
}
2019-08-11 16:52:37 +00:00
```
2019-08-11 16:52:37 +00:00
### Load exchange data to a pandas dataframe
2019-08-11 16:52:37 +00:00
This loads candle data to a dataframe
```python
from pathlib import Path
from freqtrade.data.history import load_pair_history
2019-08-20 04:47:10 +00:00
# Load data using values passed to function
2019-08-11 16:52:37 +00:00
ticker_interval = "5m"
data_location = Path('user_data', 'data', 'bitrex')
pair = "BTC_USDT"
candles = load_pair_history(datadir=data_location,
ticker_interval=ticker_interval,
pair=pair)
# Confirm success
2019-08-20 04:47:10 +00:00
print(f"Loaded len(candles) rows of data for {pair} from {data_location}")
candles.head()
2019-08-11 16:52:37 +00:00
```
## Strategy debugging example
Debugging a strategy can be time-consuming. FreqTrade offers helper functions to visualize raw data.
### Define variables used in analyses
You can override strategy settings as demonstrated below.
```python
# Customize these according to your needs.
# Define some constants
ticker_interval = "5m"
# Name of the strategy class
2019-08-27 04:41:07 +00:00
strategy_name = 'SampleStrategy'
2019-08-07 02:35:14 +00:00
# Path to user data
user_data_dir = 'user_data'
# Location of the strategy
strategy_location = Path(user_data_dir, 'strategies')
# Location of the data
data_location = Path(user_data_dir, 'data', 'binance')
2019-08-11 16:52:37 +00:00
# Pair to analyze - Only use one pair here
2019-08-07 02:35:14 +00:00
pair = "BTC_USDT"
```
2019-08-07 02:35:14 +00:00
### Load exchange data
2019-08-07 02:35:14 +00:00
```python
2019-08-11 16:52:37 +00:00
from pathlib import Path
from freqtrade.data.history import load_pair_history
2019-08-07 02:35:14 +00:00
# Load data using values set above
2019-08-11 16:52:37 +00:00
candles = load_pair_history(datadir=data_location,
2019-08-07 02:35:14 +00:00
ticker_interval=ticker_interval,
pair=pair)
2019-08-07 02:35:14 +00:00
# Confirm success
2019-08-20 04:47:10 +00:00
print(f"Loaded {len(candles)} rows of data for {pair} from {data_location}")
candles.head()
```
2019-08-07 02:35:14 +00:00
### Load and run strategy
2019-08-07 02:35:14 +00:00
* Rerun each time the strategy file is changed
2019-08-07 02:35:14 +00:00
```python
2019-08-11 16:52:37 +00:00
from freqtrade.resolvers import StrategyResolver
2019-08-20 04:47:10 +00:00
2019-08-07 02:35:14 +00:00
# Load strategy using values set above
strategy = StrategyResolver({'strategy': strategy_name,
'user_data_dir': user_data_dir,
'strategy_path': strategy_location}).strategy
2019-08-09 02:09:15 +00:00
# Generate buy/sell signals using strategy
2019-08-11 16:52:37 +00:00
df = strategy.analyze_ticker(candles, {'pair': pair})
```
2019-08-09 02:09:15 +00:00
### Display the trade details
2019-08-20 04:47:10 +00:00
* Note that using `data.tail()` is preferable to `data.head()` as most indicators have some "startup" data at the top of the dataframe.
2019-08-11 16:52:37 +00:00
* Some possible problems
* Columns with NaN values at the end of the dataframe
* Columns used in `crossed*()` functions with completely different units
* Comparison with full backtest
* having 200 buy signals as output for one pair from `analyze_ticker()` does not necessarily mean that 200 trades will be made during backtesting.
* Assuming you use only one condition such as, `df['rsi'] < 30` as buy condition, this will generate multiple "buy" signals for each pair in sequence (until rsi returns > 29). The bot will only buy on the first of these signals (and also only if a trade-slot ("max_open_trades") is still available), or on one of the middle signals, as soon as a "slot" becomes available.
2019-06-22 14:18:22 +00:00
2019-08-09 02:09:15 +00:00
```python
# Report results
print(f"Generated {df['buy'].sum()} buy signals")
data = df.set_index('date', drop=True)
data.tail()
2019-06-22 14:18:22 +00:00
```
2019-06-24 15:20:41 +00:00
Feel free to submit an issue or Pull Request enhancing this document if you would like to share ideas on how to best analyze the data.