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

Move Backtesting to a class and add unit tests

Browse Source
This commit is contained in:
Gerald Lonlas
2018-02-08 23:35:38 -08:00
parent db67b10605
commit 1d251d6151
9 changed files with 942 additions and 427 deletions
Download Patch File Download Diff File Expand all files Collapse all files

35
freqtrade/optimize/__init__.py
View File

@@ -1,18 +1,16 @@
# pragma pylint: disable=missing-docstring
import logging
import json
import os
from typing import Optional, List, Dict
from pandas import DataFrame
import gzip
from freqtrade.exchange import get_ticker_history
from freqtrade.analyze import populate_indicators, parse_ticker_dataframe
from freqtrade import misc
from freqtrade.logger import Logger
from user_data.hyperopt_conf import hyperopt_optimize_conf
import gzip
logger = logging.getLogger(__name__)
logger = Logger(name=__name__).get_logger()
def trim_tickerlist(tickerlist, timerange):
@@ -84,21 +82,13 @@ def load_data(datadir: str, ticker_interval: int, pairs: Optional[List[str]] = N
return result
def tickerdata_to_dataframe(data):
preprocessed = preprocess(data)
return preprocessed
def preprocess(tickerdata: Dict[str, List]) -> Dict[str, DataFrame]:
"""Creates a dataframe and populates indicators for given ticker data"""
return {pair: populate_indicators(parse_ticker_dataframe(pair_data))
for pair, pair_data in tickerdata.items()}
def make_testdata_path(datadir: str) -> str:
"""Return the path where testdata files are stored"""
return datadir or os.path.abspath(os.path.join(os.path.dirname(__file__),
'..', 'tests', 'testdata'))
return datadir or os.path.abspath(
os.path.join(
os.path.dirname(__file__), '..', 'tests', 'testdata'
)
)
def download_pairs(datadir, pairs: List[str], ticker_interval: int) -> bool:
@@ -115,11 +105,6 @@ def download_pairs(datadir, pairs: List[str], ticker_interval: int) -> bool:
return True
def file_dump_json(filename, data):
with open(filename, "wt") as fp:
json.dump(data, fp)
# FIX: 20180110, suggest rename interval to tick_interval
def download_backtesting_testdata(datadir: str, pair: str, interval: int = 5) -> bool:
"""
@@ -142,8 +127,8 @@ def download_backtesting_testdata(datadir: str, pair: str, interval: int = 5) ->
))
if os.path.isfile(filename):
with open(filename, "rt") as fp:
data = json.load(fp)
with open(filename, "rt") as file:
data = json.load(file)
logger.debug("Current Start: {}".format(data[1]['T']))
logger.debug("Current End: {}".format(data[-1:][0]['T']))
else:

504
freqtrade/optimize/backtesting.py
View File

@@ -1,235 +1,321 @@
# pragma pylint: disable=missing-docstring,W0212
# pragma pylint: disable=missing-docstring, W0212, too-many-arguments
"""
This module contains the backtesting logic
"""
from typing import Dict, Tuple, Any
import logging
from typing import Dict, Tuple
import arrow
from pandas import DataFrame, Series
from tabulate import tabulate
import freqtrade.misc as misc
import freqtrade.optimize as optimize
from freqtrade import exchange
from freqtrade.analyze import populate_buy_trend, populate_sell_trend
from freqtrade.arguments import Arguments
from freqtrade.exchange import Bittrex
from freqtrade.main import should_sell
from freqtrade.configuration import Configuration
from freqtrade import exchange
from freqtrade.analyze import Analyze
from freqtrade.logger import Logger
from freqtrade.misc import file_dump_json
from freqtrade.persistence import Trade
from freqtrade.strategy.strategy import Strategy
logger = logging.getLogger(__name__)
def get_timeframe(data: Dict[str, DataFrame]) -> Tuple[arrow.Arrow, arrow.Arrow]:
class Backtesting(object):
"""
Get the maximum timeframe for the given backtest data
:param data: dictionary with preprocessed backtesting data
:return: tuple containing min_date, max_date
"""
all_dates = Series([])
for pair_data in data.values():
all_dates = all_dates.append(pair_data['date'])
all_dates.sort_values(inplace=True)
return arrow.get(all_dates.iloc[0]), arrow.get(all_dates.iloc[-1])
Backtesting class, this class contains all the logic to run a backtest
To run a backtest:
backtesting = Backtesting(config)
backtesting.start()
"""
def __init__(self, config: Dict[str, Any]) -> None:
self.logging = Logger(name=__name__)
self.logger = self.logging.get_logger()
def generate_text_table(
data: Dict[str, Dict], results: DataFrame, stake_currency, ticker_interval) -> str:
"""
Generates and returns a text table for the given backtest data and the results dataframe
:return: pretty printed table with tabulate as str
"""
floatfmt = ('s', 'd', '.2f', '.8f', '.1f')
tabular_data = []
headers = ['pair', 'buy count', 'avg profit %',
'total profit ' + stake_currency, 'avg duration', 'profit', 'loss']
for pair in data:
result = results[results.currency == pair]
self.config = config
self.analyze = None
self.ticker_interval = None
self.tickerdata_to_dataframe = None
self.populate_buy_trend = None
self.populate_sell_trend = None
self._init()
def _init(self) -> None:
"""
Init objects required for backtesting
:return: None
"""
self.analyze = Analyze(self.config)
self.ticker_interval = self.analyze.strategy.ticker_interval
self.tickerdata_to_dataframe = self.analyze.tickerdata_to_dataframe
self.populate_buy_trend = self.analyze.populate_buy_trend
self.populate_sell_trend = self.analyze.populate_sell_trend
exchange._API = Bittrex({'key': '', 'secret': ''})
@staticmethod
def get_timeframe(data: Dict[str, DataFrame]) -> Tuple[arrow.Arrow, arrow.Arrow]:
"""
Get the maximum timeframe for the given backtest data
:param data: dictionary with preprocessed backtesting data
:return: tuple containing min_date, max_date
"""
all_dates = Series([])
for pair_data in data.values():
all_dates = all_dates.append(pair_data['date'])
all_dates.sort_values(inplace=True)
return arrow.get(all_dates.iloc[0]), arrow.get(all_dates.iloc[-1])
def _generate_text_table(self, data: Dict[str, Dict], results: DataFrame) -> str:
"""
Generates and returns a text table for the given backtest data and the results dataframe
:return: pretty printed table with tabulate as str
"""
stake_currency = self.config.get('stake_currency')
ticker_interval = self.ticker_interval
floatfmt = ('s', 'd', '.2f', '.8f', '.1f')
tabular_data = []
headers = ['pair', 'buy count', 'avg profit %',
'total profit ' + stake_currency, 'avg duration', 'profit', 'loss']
for pair in data:
result = results[results.currency == pair]
tabular_data.append([
pair,
len(result.index),
result.profit_percent.mean() * 100.0,
result.profit_BTC.sum(),
result.duration.mean() * ticker_interval,
len(result[result.profit_BTC > 0]),
len(result[result.profit_BTC < 0])
])
# Append Total
tabular_data.append([
pair,
len(result.index),
result.profit_percent.mean() * 100.0,
result.profit_BTC.sum(),
result.duration.mean() * ticker_interval,
len(result[result.profit_BTC > 0]),
len(result[result.profit_BTC < 0])
'TOTAL',
len(results.index),
results.profit_percent.mean() * 100.0,
results.profit_BTC.sum(),
results.duration.mean() * ticker_interval,
len(results[results.profit_BTC > 0]),
len(results[results.profit_BTC < 0])
])
return tabulate(tabular_data, headers=headers, floatfmt=floatfmt)
# Append Total
tabular_data.append([
'TOTAL',
len(results.index),
results.profit_percent.mean() * 100.0,
results.profit_BTC.sum(),
results.duration.mean() * ticker_interval,
len(results[results.profit_BTC > 0]),
len(results[results.profit_BTC < 0])
])
return tabulate(tabular_data, headers=headers, floatfmt=floatfmt)
def _get_sell_trade_entry(self, pair, row, buy_subset, ticker, trade_count_lock, args):
stake_amount = args['stake_amount']
max_open_trades = args.get('max_open_trades', 0)
trade = Trade(
open_rate=row.close,
open_date=row.date,
stake_amount=stake_amount,
amount=stake_amount / row.open,
fee=exchange.get_fee()
)
def get_sell_trade_entry(pair, row, buy_subset, ticker, trade_count_lock, args):
stake_amount = args['stake_amount']
max_open_trades = args.get('max_open_trades', 0)
trade = Trade(open_rate=row.close,
open_date=row.date,
stake_amount=stake_amount,
amount=stake_amount / row.open,
fee=exchange.get_fee()
)
# calculate win/lose forwards from buy point
sell_subset = ticker[ticker.date > row.date][['close', 'date', 'sell']]
for row2 in sell_subset.itertuples(index=True):
if max_open_trades > 0:
# Increase trade_count_lock for every iteration
trade_count_lock[row2.date] = trade_count_lock.get(row2.date, 0) + 1
# Buy is on is in the buy_subset there is a row that matches the date
# of the sell event
buy_signal = not buy_subset[buy_subset.date == row2.date].empty
if(should_sell(trade, row2.close, row2.date, buy_signal, row2.sell)):
return row2, (pair,
trade.calc_profit_percent(rate=row2.close),
trade.calc_profit(rate=row2.close),
row2.Index - row.Index
), row2.date
return None
def backtest(args) -> DataFrame:
"""
Implements backtesting functionality
:param args: a dict containing:
stake_amount: btc amount to use for each trade
processed: a processed dictionary with format {pair, data}
max_open_trades: maximum number of concurrent trades (default: 0, disabled)
realistic: do we try to simulate realistic trades? (default: True)
sell_profit_only: sell if profit only
use_sell_signal: act on sell-signal
stoploss: use stoploss
:return: DataFrame
"""
processed = args['processed']
max_open_trades = args.get('max_open_trades', 0)
realistic = args.get('realistic', True)
record = args.get('record', None)
records = []
trades = []
trade_count_lock: dict = {}
exchange._API = Bittrex({'key': '', 'secret': ''})
for pair, pair_data in processed.items():
pair_data['buy'], pair_data['sell'] = 0, 0
ticker = populate_sell_trend(populate_buy_trend(pair_data))
# for each buy point
lock_pair_until = None
headers = ['buy', 'open', 'close', 'date', 'sell']
buy_subset = ticker[(ticker.buy == 1) & (ticker.sell == 0)][headers]
for row in buy_subset.itertuples(index=True):
if realistic:
if lock_pair_until is not None and row.date <= lock_pair_until:
continue
# calculate win/lose forwards from buy point
sell_subset = ticker[ticker.date > row.date][['close', 'date', 'sell']]
for row2 in sell_subset.itertuples(index=True):
if max_open_trades > 0:
# Check if max_open_trades has already been reached for the given date
if not trade_count_lock.get(row.date, 0) < max_open_trades:
continue
# Increase trade_count_lock for every iteration
trade_count_lock[row2.date] = trade_count_lock.get(row2.date, 0) + 1
if max_open_trades > 0:
# Increase lock
trade_count_lock[row.date] = trade_count_lock.get(row.date, 0) + 1
# Buy is on is in the buy_subset there is a row that matches the date
# of the sell event
buy_signal = not buy_subset[buy_subset.date == row2.date].empty
if(
self.analyze.should_sell(
trade=trade,
rate=row2.close,
date=row2.date,
buy=buy_signal,
sell=row2.sell
)
):
return \
row2, \
(
pair,
trade.calc_profit_percent(rate=row2.close),
trade.calc_profit(rate=row2.close),
row2.Index - row.Index
),\
row2.date
return None
ret = get_sell_trade_entry(pair, row, buy_subset, ticker,
trade_count_lock, args)
if ret:
row2, trade_entry, next_date = ret
lock_pair_until = next_date
trades.append(trade_entry)
if record:
# Note, need to be json.dump friendly
# record a tuple of pair, current_profit_percent,
# entry-date, duration
records.append((pair, trade_entry[1],
row.date.strftime('%s'),
row2.date.strftime('%s'),
row.Index, trade_entry[3]))
# For now export inside backtest(), maybe change so that backtest()
# returns a tuple like: (dataframe, records, logs, etc)
if record and record.find('trades') >= 0:
logger.info('Dumping backtest results')
misc.file_dump_json('backtest-result.json', records)
labels = ['currency', 'profit_percent', 'profit_BTC', 'duration']
return DataFrame.from_records(trades, columns=labels)
def backtest(self, args) -> DataFrame:
"""
Implements backtesting functionality
NOTE: This method is used by Hyperopt at each iteration. Please keep it optimized.
Of course try to not have ugly code. By some accessor are sometime slower than functions.
Avoid, logging on this method
:param args: a dict containing:
stake_amount: btc amount to use for each trade
processed: a processed dictionary with format {pair, data}
max_open_trades: maximum number of concurrent trades (default: 0, disabled)
realistic: do we try to simulate realistic trades? (default: True)
sell_profit_only: sell if profit only
use_sell_signal: act on sell-signal
stoploss: use stoploss
:return: DataFrame
"""
processed = args['processed']
max_open_trades = args.get('max_open_trades', 0)
realistic = args.get('realistic', True)
record = args.get('record', None)
records = []
trades = []
trade_count_lock = {}
for pair, pair_data in processed.items():
pair_data['buy'], pair_data['sell'] = 0, 0
ticker = self.populate_sell_trend(
self.populate_buy_trend(pair_data)
)
# for each buy point
lock_pair_until = None
headers = ['buy', 'open', 'close', 'date', 'sell']
buy_subset = ticker[(ticker.buy == 1) & (ticker.sell == 0)][headers]
for row in buy_subset.itertuples(index=True):
if realistic:
if lock_pair_until is not None and row.date <= lock_pair_until:
continue
if max_open_trades > 0:
# Check if max_open_trades has already been reached for the given date
if not trade_count_lock.get(row.date, 0) < max_open_trades:
continue
if max_open_trades > 0:
# Increase lock
trade_count_lock[row.date] = trade_count_lock.get(row.date, 0) + 1
ret = self._get_sell_trade_entry(
pair=pair,
row=row,
buy_subset=buy_subset,
ticker=ticker,
trade_count_lock=trade_count_lock,
args=args
)
if ret:
row2, trade_entry, next_date = ret
lock_pair_until = next_date
trades.append(trade_entry)
if record:
# Note, need to be json.dump friendly
# record a tuple of pair, current_profit_percent,
# entry-date, duration
records.append((pair, trade_entry[1],
row.date.strftime('%s'),
row2.date.strftime('%s'),
row.Index, trade_entry[3]))
# For now export inside backtest(), maybe change so that backtest()
# returns a tuple like: (dataframe, records, logs, etc)
if record and record.find('trades') >= 0:
self.logger.info('Dumping backtest results')
file_dump_json('backtest-result.json', records)
labels = ['currency', 'profit_percent', 'profit_BTC', 'duration']
return DataFrame.from_records(trades, columns=labels)
def start(self) -> None:
"""
Run a backtesting end-to-end
:return: None
"""
data = {}
pairs = self.config['exchange']['pair_whitelist']
if self.config.get('live'):
self.logger.info('Downloading data for all pairs in whitelist ...')
for pair in pairs:
data[pair] = exchange.get_ticker_history(pair, self.ticker_interval)
else:
self.logger.info('Using local backtesting data (using whitelist in given config) ...')
self.logger.info('Using stake_currency: %s ...', self.config['stake_currency'])
self.logger.info('Using stake_amount: %s ...', self.config['stake_amount'])
timerange = Arguments.parse_timerange(self.config.get('timerange'))
data = optimize.load_data(
self.config['datadir'],
pairs=pairs,
ticker_interval=self.ticker_interval,
refresh_pairs=self.config.get('refresh_pairs', False),
timerange=timerange
)
max_open_trades = self.config.get('max_open_trades', 0)
preprocessed = self.tickerdata_to_dataframe(data)
# Print timeframe
min_date, max_date = self.get_timeframe(preprocessed)
self.logger.info(
'Measuring data from %s up to %s (%s days)..',
min_date.isoformat(),
max_date.isoformat(),
(max_date - min_date).days
)
# Execute backtest and print results
sell_profit_only = self.config.get('experimental', {}).get('sell_profit_only', False)
use_sell_signal = self.config.get('experimental', {}).get('use_sell_signal', False)
results = self.backtest(
{
'stake_amount': self.config.get('stake_amount'),
'processed': preprocessed,
'max_open_trades': max_open_trades,
'realistic': self.config.get('realistic_simulation', False),
'sell_profit_only': sell_profit_only,
'use_sell_signal': use_sell_signal,
'stoploss': self.analyze.strategy.stoploss,
'record': self.config.get('export')
}
)
self.logging.set_format('%(message)s')
self.logger.info(
'\n==================================== '
'BACKTESTING REPORT'
' ====================================\n'
'%s',
self._generate_text_table(
data,
results
)
)
def start(args):
def setup_configuration(args) -> Dict[str, Any]:
"""
Prepare the configuration for the backtesting
:param args: Cli args from Arguments()
:return: Configuration
"""
configuration = Configuration(args)
config = configuration.get_config()
# Ensure we do not use Exchange credentials
config['exchange']['key'] = ''
config['exchange']['secret'] = ''
return config
def start(args) -> None:
"""
Start Backtesting script
:param args: Cli args from Arguments()
:return: None
"""
# Initialize logger
logging.basicConfig(
level=args.loglevel,
format='%(asctime)s - %(name)s - %(levelname)s - %(message)s',
)
logger = Logger(name=__name__).get_logger()
logger.info('Starting freqtrade in Backtesting mode')
exchange._API = Bittrex({'key': '', 'secret': ''})
# Initialize configuration
config = setup_configuration(args)
logger.info('Using config: %s ...', args.config)
config = misc.load_config(args.config)
# If -i/--ticker-interval is use we override the configuration parameter
# (that will override the strategy configuration)
if args.ticker_interval:
config.update({'ticker_interval': args.ticker_interval})
# init the strategy to use
config.update({'strategy': args.strategy})
strategy = Strategy()
strategy.init(config)
logger.info('Using ticker_interval: %d ...', strategy.ticker_interval)
data = {}
pairs = config['exchange']['pair_whitelist']
if args.live:
logger.info('Downloading data for all pairs in whitelist ...')
for pair in pairs:
data[pair] = exchange.get_ticker_history(pair, strategy.ticker_interval)
else:
logger.info('Using local backtesting data (using whitelist in given config) ...')
logger.info('Using stake_currency: %s ...', config['stake_currency'])
logger.info('Using stake_amount: %s ...', config['stake_amount'])
timerange = misc.parse_timerange(args.timerange)
data = optimize.load_data(args.datadir,
pairs=pairs,
ticker_interval=strategy.ticker_interval,
refresh_pairs=args.refresh_pairs,
timerange=timerange)
max_open_trades = 0
if args.realistic_simulation:
logger.info('Using max_open_trades: %s ...', config['max_open_trades'])
max_open_trades = config['max_open_trades']
# Monkey patch config
from freqtrade import main
main._CONF = config
preprocessed = optimize.tickerdata_to_dataframe(data)
# Print timeframe
min_date, max_date = get_timeframe(preprocessed)
logger.info('Measuring data from %s up to %s (%s days)..',
min_date.isoformat(),
max_date.isoformat(),
(max_date-min_date).days)
# Execute backtest and print results
sell_profit_only = config.get('experimental', {}).get('sell_profit_only', False)
use_sell_signal = config.get('experimental', {}).get('use_sell_signal', False)
results = backtest({'stake_amount': config['stake_amount'],
'processed': preprocessed,
'max_open_trades': max_open_trades,
'realistic': args.realistic_simulation,
'sell_profit_only': sell_profit_only,
'use_sell_signal': use_sell_signal,
'stoploss': strategy.stoploss,
'record': args.export
})
logger.info(
'\n==================================== BACKTESTING REPORT ====================================\n%s', # noqa
generate_text_table(data, results, config['stake_currency'], strategy.ticker_interval)
)
# Initialize backtesting object
backtesting = Backtesting(config)
backtesting.start()