# pragma pylint: disable=missing-docstring,W0212 import logging import math import pandas as pd from unittest.mock import MagicMock from freqtrade import exchange, optimize from freqtrade.exchange import Bittrex from freqtrade.optimize import preprocess from freqtrade.optimize.backtesting import backtest, generate_text_table, get_timeframe import freqtrade.optimize.backtesting as backtesting from freqtrade.strategy.strategy import Strategy def setup_strategy(): strategy = Strategy() strategy.init({'strategy': 'default_strategy'}) return strategy def trim_dictlist(dl, num): new = {} for pair, pair_data in dl.items(): new[pair] = pair_data[num:] return new def test_generate_text_table(): results = pd.DataFrame( { 'currency': ['BTC_ETH', 'BTC_ETH'], 'profit_percent': [0.1, 0.2], 'profit_BTC': [0.2, 0.4], 'duration': [10, 30], 'profit': [2, 0], 'loss': [0, 0] } ) print(generate_text_table({'BTC_ETH': {}}, results, 'BTC', 5)) assert generate_text_table({'BTC_ETH': {}}, results, 'BTC', 5) == ( 'pair buy count avg profit % total profit BTC avg duration profit loss\n' # noqa '------- ----------- -------------- ------------------ -------------- -------- ------\n' # noqa 'BTC_ETH 2 15.00 0.60000000 100.0 2 0\n' # noqa 'TOTAL 2 15.00 0.60000000 100.0 2 0') # noqa def test_get_timeframe(): data = preprocess(optimize.load_data( None, ticker_interval=1, pairs=['BTC_UNITEST'])) min_date, max_date = get_timeframe(data) assert min_date.isoformat() == '2017-11-04T23:02:00+00:00' assert max_date.isoformat() == '2017-11-14T22:59:00+00:00' def test_backtest(default_conf, mocker): setup_strategy() mocker.patch.dict('freqtrade.main._CONF', default_conf) exchange._API = Bittrex({'key': '', 'secret': ''}) data = optimize.load_data(None, ticker_interval=5, pairs=['BTC_ETH']) data = trim_dictlist(data, -200) results = backtest({'stake_amount': default_conf['stake_amount'], 'processed': optimize.preprocess(data), 'max_open_trades': 10, 'realistic': True}) assert not results.empty def test_backtest_1min_ticker_interval(default_conf, mocker): mocker.patch.dict('freqtrade.main._CONF', default_conf) exchange._API = Bittrex({'key': '', 'secret': ''}) setup_strategy() # Run a backtesting for an exiting 5min ticker_interval data = optimize.load_data(None, ticker_interval=1, pairs=['BTC_UNITEST']) data = trim_dictlist(data, -200) results = backtest({'stake_amount': default_conf['stake_amount'], 'processed': optimize.preprocess(data), 'max_open_trades': 1, 'realistic': True}) assert not results.empty def load_data_test(what): timerange = ((None, 'line'), None, -100) data = optimize.load_data(None, ticker_interval=1, pairs=['BTC_UNITEST'], timerange=timerange) pair = data['BTC_UNITEST'] datalen = len(pair) # Depending on the what parameter we now adjust the # loaded data looks: # pair :: [{'O': 0.123, 'H': 0.123, 'L': 0.123, # 'C': 0.123, 'V': 123.123, # 'T': '2017-11-04T23:02:00', 'BV': 0.123}] base = 0.001 if what == 'raise': return {'BTC_UNITEST': [{'T': pair[x]['T'], # Keep old dates 'V': pair[x]['V'], # Keep old volume 'BV': pair[x]['BV'], # keep too 'O': x * base, # But replace O,H,L,C 'H': x * base + 0.0001, 'L': x * base - 0.0001, 'C': x * base} for x in range(0, datalen)]} if what == 'lower': return {'BTC_UNITEST': [{'T': pair[x]['T'], # Keep old dates 'V': pair[x]['V'], # Keep old volume 'BV': pair[x]['BV'], # keep too 'O': 1 - x * base, # But replace O,H,L,C 'H': 1 - x * base + 0.0001, 'L': 1 - x * base - 0.0001, 'C': 1 - x * base} for x in range(0, datalen)]} if what == 'sine': hz = 0.1 # frequency return {'BTC_UNITEST': [{'T': pair[x]['T'], # Keep old dates 'V': pair[x]['V'], # Keep old volume 'BV': pair[x]['BV'], # keep too # But replace O,H,L,C 'O': math.sin(x * hz) / 1000 + base, 'H': math.sin(x * hz) / 1000 + base + 0.0001, 'L': math.sin(x * hz) / 1000 + base - 0.0001, 'C': math.sin(x * hz) / 1000 + base} for x in range(0, datalen)]} return data def simple_backtest(config, contour, num_results): setup_strategy() data = load_data_test(contour) processed = optimize.preprocess(data) assert isinstance(processed, dict) results = backtest({'stake_amount': config['stake_amount'], 'processed': processed, 'max_open_trades': 1, 'realistic': True}) # results :: assert len(results) == num_results # Test backtest on offline data # loaded by freqdata/optimize/__init__.py::load_data() def test_backtest2(default_conf, mocker): setup_strategy() mocker.patch.dict('freqtrade.main._CONF', default_conf) data = optimize.load_data(None, ticker_interval=5, pairs=['BTC_ETH']) data = trim_dictlist(data, -200) results = backtest({'stake_amount': default_conf['stake_amount'], 'processed': optimize.preprocess(data), 'max_open_trades': 10, 'realistic': True}) assert not results.empty def test_processed(default_conf, mocker): setup_strategy() mocker.patch.dict('freqtrade.main._CONF', default_conf) dict_of_tickerrows = load_data_test('raise') dataframes = optimize.preprocess(dict_of_tickerrows) dataframe = dataframes['BTC_UNITEST'] cols = dataframe.columns # assert the dataframe got some of the indicator columns for col in ['close', 'high', 'low', 'open', 'date', 'ema50', 'ao', 'macd', 'plus_dm']: assert col in cols def test_backtest_pricecontours(default_conf, mocker): mocker.patch.dict('freqtrade.main._CONF', default_conf) tests = [['raise', 17], ['lower', 0], ['sine', 17]] for [contour, numres] in tests: simple_backtest(default_conf, contour, numres) def mocked_load_data(datadir, pairs=[], ticker_interval=0, refresh_pairs=False, timerange=None): tickerdata = optimize.load_tickerdata_file(datadir, 'BTC_UNITEST', 1, timerange=timerange) pairdata = {'BTC_UNITEST': tickerdata} return pairdata def test_backtest_start(default_conf, mocker, caplog): default_conf['exchange']['pair_whitelist'] = ['BTC_UNITEST'] mocker.patch.dict('freqtrade.main._CONF', default_conf) mocker.patch('freqtrade.misc.load_config', new=lambda s: default_conf) mocker.patch.multiple('freqtrade.optimize', load_data=mocked_load_data) args = MagicMock() args.ticker_interval = 1 args.level = 10 args.live = False args.datadir = None args.export = None args.timerange = '-100' # needed due to MagicMock malleability backtesting.start(args) # check the logs, that will contain the backtest result exists = ['Using max_open_trades: 1 ...', 'Using stake_amount: 0.001 ...', 'Measuring data from 2017-11-14T21:17:00+00:00 up to 2017-11-14T22:59:00+00:00 ...'] for line in exists: assert ('freqtrade.optimize.backtesting', logging.INFO, line) in caplog.record_tuples