adelorenzo/stable
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Merge branch 'feat/short' into pr/samgermain/5378

Browse Source
This commit is contained in:
Matthias
2021-08-24 06:28:16 +02:00
parent faf5cfa66d 957551ea97
commit 7a977a8eaf
57 changed files with 1076 additions and 811 deletions
Download Patch File Download Diff File Expand all files Collapse all files

40
freqtrade/optimize/backtesting.py
View File

@@ -133,6 +133,9 @@ class Backtesting:
self.abort = False
def __del__(self):
self.cleanup()
def cleanup(self):
LoggingMixin.show_output = True
PairLocks.use_db = True
Trade.use_db = True
@@ -219,7 +222,7 @@ class Backtesting:
# Every change to this headers list must evaluate further usages of the resulting tuple
# and eventually change the constants for indexes at the top
headers = ['date', 'open', 'high', 'low', 'close', 'enter_long', 'exit_long',
'enter_short', 'exit_short']
'enter_short', 'exit_short', 'long_tag', 'short_tag']
data: Dict = {}
self.progress.init_step(BacktestState.CONVERT, len(processed))
@@ -227,21 +230,15 @@ class Backtesting:
for pair, pair_data in processed.items():
self.check_abort()
self.progress.increment()
has_buy_tag = 'long_tag' in pair_data
has_short_tag = 'short_tag' in pair_data
headers = headers + ['long_tag'] if has_buy_tag else headers
headers = headers + ['short_tag'] if has_short_tag else headers
if not pair_data.empty:
# Cleanup from prior runs
pair_data.loc[:, 'buy'] = 0 # TODO: Should be renamed to enter_long
pair_data.loc[:, 'enter_short'] = 0
pair_data.loc[:, 'sell'] = 0 # TODO: should be renamed to exit_long
pair_data.loc[:, 'exit_short'] = 0
# pair_data.loc[:, 'sell'] = 0
if has_buy_tag:
pair_data.loc[:, 'long_tag'] = None # cleanup if buy_tag is exist
if has_short_tag:
pair_data.loc[:, 'short_tag'] = None # cleanup if short_tag is exist
pair_data.loc[:, 'long_tag'] = None # cleanup if buy_tag is exist
pair_data.loc[:, 'short_tag'] = None # cleanup if short_tag is exist
df_analyzed = self.strategy.advise_sell(
self.strategy.advise_buy(pair_data, {'pair': pair}),
@@ -256,14 +253,15 @@ class Backtesting:
df_analyzed.loc[:, 'enter_short'] = df_analyzed.loc[:, 'enter_short'].shift(1)
df_analyzed.loc[:, 'exit_long'] = df_analyzed.loc[:, 'exit_long'].shift(1)
df_analyzed.loc[:, 'exit_short'] = df_analyzed.loc[:, 'exit_short'].shift(1)
if has_buy_tag:
df_analyzed.loc[:, 'long_tag'] = df_analyzed.loc[:, 'long_tag'].shift(1)
df_analyzed.loc[:, 'long_tag'] = df_analyzed.loc[:, 'long_tag'].shift(1)
df_analyzed.drop(df_analyzed.head(1).index, inplace=True)
# Update dataprovider cache
self.dataprovider._set_cached_df(pair, self.timeframe, df_analyzed)
df_analyzed = df_analyzed.drop(df_analyzed.head(1).index)
# Convert from Pandas to list for performance reasons
# (Looping Pandas is slow.)
data[pair] = df_analyzed[headers].values.tolist()
@@ -337,13 +335,14 @@ class Backtesting:
def _get_sell_trade_entry(self, trade: LocalTrade, sell_row: Tuple) -> Optional[LocalTrade]:
# TODO: short exits
sell_candle_time = sell_row[DATE_IDX].to_pydatetime()
sell = self.strategy.should_sell(trade, sell_row[OPEN_IDX], # type: ignore
sell_row[DATE_IDX].to_pydatetime(), sell_row[BUY_IDX],
sell_candle_time, sell_row[BUY_IDX],
sell_row[SELL_IDX],
low=sell_row[LOW_IDX], high=sell_row[HIGH_IDX])
if sell.sell_flag:
trade.close_date = sell_row[DATE_IDX].to_pydatetime()
trade.close_date = sell_candle_time
trade.sell_reason = sell.sell_reason
trade_dur = int((trade.close_date_utc - trade.open_date_utc).total_seconds() // 60)
closerate = self._get_close_rate(sell_row, trade, sell, trade_dur)
@@ -355,7 +354,7 @@ class Backtesting:
rate=closerate,
time_in_force=time_in_force,
sell_reason=sell.sell_reason,
current_time=sell_row[DATE_IDX].to_pydatetime()):
current_time=sell_candle_time):
return None
trade.close(closerate, show_msg=False)
@@ -494,6 +493,8 @@ class Backtesting:
for i, pair in enumerate(data):
row_index = indexes[pair]
try:
# Row is treated as "current incomplete candle".
# Buy / sell signals are shifted by 1 to compensate for this.
row = data[pair][row_index]
except IndexError:
# missing Data for one pair at the end.
@@ -505,8 +506,8 @@ class Backtesting:
continue
row_index += 1
self.dataprovider._set_dataframe_max_index(row_index)
indexes[pair] = row_index
self.dataprovider._set_dataframe_max_index(row_index)
# without positionstacking, we can only have one open trade per pair.
# max_open_trades must be respected
@@ -530,7 +531,7 @@ class Backtesting:
open_trades[pair].append(trade)
LocalTrade.add_bt_trade(trade)
for trade in open_trades[pair]:
for trade in list(open_trades[pair]):
# also check the buying candle for sell conditions.
trade_entry = self._get_sell_trade_entry(trade, row)
# Sell occurred
@@ -561,7 +562,8 @@ class Backtesting:
'final_balance': self.wallets.get_total(self.strategy.config['stake_currency']),
}
def backtest_one_strategy(self, strat: IStrategy, data: Dict[str, Any], timerange: TimeRange):
def backtest_one_strategy(self, strat: IStrategy, data: Dict[str, DataFrame],
timerange: TimeRange):
self.progress.init_step(BacktestState.ANALYZE, 0)
logger.info("Running backtesting for Strategy %s", strat.get_strategy_name())
@@ -580,7 +582,7 @@ class Backtesting:
max_open_trades = 0
# need to reprocess data every time to populate signals
preprocessed = self.strategy.ohlcvdata_to_dataframe(data)
preprocessed = self.strategy.advise_all_indicators(data)
# Trim startup period from analyzed dataframe
preprocessed_tmp = trim_dataframes(preprocessed, timerange, self.required_startup)

2
freqtrade/optimize/hyperopt.py
View File

@@ -394,7 +394,7 @@ class Hyperopt:
data, timerange = self.backtesting.load_bt_data()
logger.info("Dataload complete. Calculating indicators")
preprocessed = self.backtesting.strategy.ohlcvdata_to_dataframe(data)
preprocessed = self.backtesting.strategy.advise_all_indicators(data)
# Trim startup period from analyzed dataframe to get correct dates for output.
processed = trim_dataframes(preprocessed, timerange, self.backtesting.required_startup)

128
freqtrade/optimize/hyperopt_epoch_filters.py Normal file
View File

@@ -0,0 +1,128 @@
import logging
from typing import List
from freqtrade.exceptions import OperationalException
logger = logging.getLogger(__name__)
def hyperopt_filter_epochs(epochs: List, filteroptions: dict, log: bool = True) -> List:
"""
Filter our items from the list of hyperopt results
"""
if filteroptions['only_best']:
epochs = [x for x in epochs if x['is_best']]
if filteroptions['only_profitable']:
epochs = [x for x in epochs
if x['results_metrics'].get('profit_total', 0) > 0]
epochs = _hyperopt_filter_epochs_trade_count(epochs, filteroptions)
epochs = _hyperopt_filter_epochs_duration(epochs, filteroptions)
epochs = _hyperopt_filter_epochs_profit(epochs, filteroptions)
epochs = _hyperopt_filter_epochs_objective(epochs, filteroptions)
if log:
logger.info(f"{len(epochs)} " +
("best " if filteroptions['only_best'] else "") +
("profitable " if filteroptions['only_profitable'] else "") +
"epochs found.")
return epochs
def _hyperopt_filter_epochs_trade(epochs: List, trade_count: int):
"""
Filter epochs with trade-counts > trades
"""
return [
x for x in epochs if x['results_metrics'].get('total_trades', 0) > trade_count
]
def _hyperopt_filter_epochs_trade_count(epochs: List, filteroptions: dict) -> List:
if filteroptions['filter_min_trades'] > 0:
epochs = _hyperopt_filter_epochs_trade(epochs, filteroptions['filter_min_trades'])
if filteroptions['filter_max_trades'] > 0:
epochs = [
x for x in epochs
if x['results_metrics'].get('total_trades') < filteroptions['filter_max_trades']
]
return epochs
def _hyperopt_filter_epochs_duration(epochs: List, filteroptions: dict) -> List:
def get_duration_value(x):
# Duration in minutes ...
if 'holding_avg_s' in x['results_metrics']:
avg = x['results_metrics']['holding_avg_s']
return avg // 60
raise OperationalException(
"Holding-average not available. Please omit the filter on average time, "
"or rerun hyperopt with this version")
if filteroptions['filter_min_avg_time'] is not None:
epochs = _hyperopt_filter_epochs_trade(epochs, 0)
epochs = [
x for x in epochs
if get_duration_value(x) > filteroptions['filter_min_avg_time']
]
if filteroptions['filter_max_avg_time'] is not None:
epochs = _hyperopt_filter_epochs_trade(epochs, 0)
epochs = [
x for x in epochs
if get_duration_value(x) < filteroptions['filter_max_avg_time']
]
return epochs
def _hyperopt_filter_epochs_profit(epochs: List, filteroptions: dict) -> List:
if filteroptions['filter_min_avg_profit'] is not None:
epochs = _hyperopt_filter_epochs_trade(epochs, 0)
epochs = [
x for x in epochs
if x['results_metrics'].get('profit_mean', 0) * 100
> filteroptions['filter_min_avg_profit']
]
if filteroptions['filter_max_avg_profit'] is not None:
epochs = _hyperopt_filter_epochs_trade(epochs, 0)
epochs = [
x for x in epochs
if x['results_metrics'].get('profit_mean', 0) * 100
< filteroptions['filter_max_avg_profit']
]
if filteroptions['filter_min_total_profit'] is not None:
epochs = _hyperopt_filter_epochs_trade(epochs, 0)
epochs = [
x for x in epochs
if x['results_metrics'].get('profit_total_abs', 0)
> filteroptions['filter_min_total_profit']
]
if filteroptions['filter_max_total_profit'] is not None:
epochs = _hyperopt_filter_epochs_trade(epochs, 0)
epochs = [
x for x in epochs
if x['results_metrics'].get('profit_total_abs', 0)
< filteroptions['filter_max_total_profit']
]
return epochs
def _hyperopt_filter_epochs_objective(epochs: List, filteroptions: dict) -> List:
if filteroptions['filter_min_objective'] is not None:
epochs = _hyperopt_filter_epochs_trade(epochs, 0)
epochs = [x for x in epochs if x['loss'] < filteroptions['filter_min_objective']]
if filteroptions['filter_max_objective'] is not None:
epochs = _hyperopt_filter_epochs_trade(epochs, 0)
epochs = [x for x in epochs if x['loss'] > filteroptions['filter_max_objective']]
return epochs

111
freqtrade/optimize/hyperopt_tools.py
View File

@@ -4,7 +4,7 @@ import logging
from copy import deepcopy
from datetime import datetime, timezone
from pathlib import Path
from typing import Any, Dict, List, Optional
from typing import Any, Dict, Iterator, List, Optional, Tuple
import numpy as np
import rapidjson
@@ -15,6 +15,7 @@ from pandas import isna, json_normalize
from freqtrade.constants import FTHYPT_FILEVERSION, USERPATH_STRATEGIES
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
logger = logging.getLogger(__name__)
@@ -89,46 +90,70 @@ class HyperoptTools():
return any(s in config['spaces'] for s in [space, 'all', 'default'])
@staticmethod
def _read_results_pickle(results_file: Path) -> List:
def _read_results(results_file: Path, batch_size: int = 10) -> Iterator[List[Any]]:
"""
Read hyperopt results from pickle file
LEGACY method - new files are written as json and cannot be read with this method.
"""
from joblib import load
logger.info(f"Reading pickled epochs from '{results_file}'")
data = load(results_file)
return data
@staticmethod
def _read_results(results_file: Path) -> List:
"""
Read hyperopt results from file
Stream hyperopt results from file
"""
import rapidjson
logger.info(f"Reading epochs from '{results_file}'")
with results_file.open('r') as f:
data = [rapidjson.loads(line) for line in f]
return data
data = []
for line in f:
data += [rapidjson.loads(line)]
if len(data) >= batch_size:
yield data
data = []
yield data
@staticmethod
def load_previous_results(results_file: Path) -> List:
"""
Load data for epochs from the file if we have one
"""
epochs: List = []
def _test_hyperopt_results_exist(results_file) -> bool:
if results_file.is_file() and results_file.stat().st_size > 0:
if results_file.suffix == '.pickle':
epochs = HyperoptTools._read_results_pickle(results_file)
else:
epochs = HyperoptTools._read_results(results_file)
# Detection of some old format, without 'is_best' field saved
if epochs[0].get('is_best') is None:
raise OperationalException(
"Legacy hyperopt results are no longer supported."
"Please rerun hyperopt or use an older version to load this file."
)
return True
else:
# No file found.
return False
@staticmethod
def load_filtered_results(results_file: Path, config: Dict[str, Any]) -> Tuple[List, int]:
filteroptions = {
'only_best': config.get('hyperopt_list_best', False),
'only_profitable': config.get('hyperopt_list_profitable', False),
'filter_min_trades': config.get('hyperopt_list_min_trades', 0),
'filter_max_trades': config.get('hyperopt_list_max_trades', 0),
'filter_min_avg_time': config.get('hyperopt_list_min_avg_time', None),
'filter_max_avg_time': config.get('hyperopt_list_max_avg_time', None),
'filter_min_avg_profit': config.get('hyperopt_list_min_avg_profit', None),
'filter_max_avg_profit': config.get('hyperopt_list_max_avg_profit', None),
'filter_min_total_profit': config.get('hyperopt_list_min_total_profit', None),
'filter_max_total_profit': config.get('hyperopt_list_max_total_profit', None),
'filter_min_objective': config.get('hyperopt_list_min_objective', None),
'filter_max_objective': config.get('hyperopt_list_max_objective', None),
}
if not HyperoptTools._test_hyperopt_results_exist(results_file):
# No file found.
return [], 0
epochs = []
total_epochs = 0
for epochs_tmp in HyperoptTools._read_results(results_file):
if total_epochs == 0 and epochs_tmp[0].get('is_best') is None:
raise OperationalException(
"The file with HyperoptTools results is incompatible with this version "
"of Freqtrade and cannot be loaded.")
logger.info(f"Loaded {len(epochs)} previous evaluations from disk.")
return epochs
total_epochs += len(epochs_tmp)
epochs += hyperopt_filter_epochs(epochs_tmp, filteroptions, log=False)
logger.info(f"Loaded {total_epochs} previous evaluations from disk.")
# Final filter run ...
epochs = hyperopt_filter_epochs(epochs, filteroptions, log=True)
return epochs, total_epochs
@staticmethod
def show_epoch_details(results, total_epochs: int, print_json: bool,
@@ -433,21 +458,14 @@ class HyperoptTools():
trials['Best'] = ''
trials['Stake currency'] = config['stake_currency']
if 'results_metrics.total_trades' in trials:
base_metrics = ['Best', 'current_epoch', 'results_metrics.total_trades',
'results_metrics.profit_mean', 'results_metrics.profit_median',
'results_metrics.profit_total',
'Stake currency',
'results_metrics.profit_total_abs', 'results_metrics.holding_avg',
'loss', 'is_initial_point', 'is_best']
perc_multi = 100
else:
perc_multi = 1
base_metrics = ['Best', 'current_epoch', 'results_metrics.trade_count',
'results_metrics.avg_profit', 'results_metrics.median_profit',
'results_metrics.total_profit',
'Stake currency', 'results_metrics.profit', 'results_metrics.duration',
'loss', 'is_initial_point', 'is_best']
base_metrics = ['Best', 'current_epoch', 'results_metrics.total_trades',
'results_metrics.profit_mean', 'results_metrics.profit_median',
'results_metrics.profit_total',
'Stake currency',
'results_metrics.profit_total_abs', 'results_metrics.holding_avg',
'loss', 'is_initial_point', 'is_best']
perc_multi = 100
param_metrics = [("params_dict."+param) for param in results[0]['params_dict'].keys()]
trials = trials[base_metrics + param_metrics]
@@ -475,11 +493,6 @@ class HyperoptTools():
trials['Avg profit'] = trials['Avg profit'].apply(
lambda x: f'{x * perc_multi:,.2f}%' if not isna(x) else ""
)
if perc_multi == 1:
trials['Avg duration'] = trials['Avg duration'].apply(
lambda x: f'{x:,.1f} m' if isinstance(
x, float) else f"{x.total_seconds() // 60:,.1f} m" if not isna(x) else ""
)
trials['Objective'] = trials['Objective'].apply(
lambda x: f'{x:,.5f}' if x != 100000 else ""
)