tests: more backtesting testing (#496)

* tests: more backtesting testing

* tests: hyperopt

* tests: document kludge

* tests: improve test_dataframe_correct_length

* tests: remove remarks
This commit is contained in:
kryofly 2018-02-08 20:49:43 +01:00 committed by Janne Sinivirta
parent 53b1f7ac4d
commit 12a19e400f
5 changed files with 206 additions and 15 deletions

View File

@ -186,14 +186,15 @@ def start(args):
data = {}
pairs = config['exchange']['pair_whitelist']
logger.info('Using stake_currency: %s ...', config['stake_currency'])
logger.info('Using stake_amount: %s ...', config['stake_amount'])
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,

View File

@ -261,6 +261,7 @@ def ticker_history_without_bv():
]
# FIX: Perhaps change result fixture to use BTC_UNITEST instead?
@pytest.fixture
def result():
with open('freqtrade/tests/testdata/BTC_ETH-1.json') as data_file:

View File

@ -1,9 +1,10 @@
# pragma pylint: disable=missing-docstring, W0212, line-too-long, C0103
import random
import logging
import math
from unittest.mock import MagicMock
import pandas as pd
import numpy as np
from freqtrade import exchange, optimize
from freqtrade.exchange import Bittrex
from freqtrade.optimize import preprocess
@ -18,6 +19,70 @@ def trim_dictlist(dict_list, num):
return new
# use for mock freqtrade.exchange.get_ticker_history'
def _load_pair_as_ticks(pair, tickfreq):
ticks = optimize.load_data(None, ticker_interval=8, pairs=[pair])
ticks = trim_dictlist(ticks, -200)
return ticks[pair]
# FIX: fixturize this?
def _make_backtest_conf(conf=None,
pair='BTC_UNITEST',
record=None):
data = optimize.load_data(None, ticker_interval=8, pairs=[pair])
data = trim_dictlist(data, -200)
return {'stake_amount': conf['stake_amount'],
'processed': optimize.preprocess(data),
'max_open_trades': 10,
'realistic': True,
'record': record}
def _trend(signals, buy_value, sell_value):
n = len(signals['low'])
buy = np.zeros(n)
sell = np.zeros(n)
for i in range(0, len(signals['buy'])):
if random.random() > 0.5: # Both buy and sell signals at same timeframe
buy[i] = buy_value
sell[i] = sell_value
signals['buy'] = buy
signals['sell'] = sell
return signals
def _trend_alternate(dataframe=None):
signals = dataframe
low = signals['low']
n = len(low)
buy = np.zeros(n)
sell = np.zeros(n)
for i in range(0, len(buy)):
if i % 2 == 0:
buy[i] = 1
else:
sell[i] = 1
signals['buy'] = buy
signals['sell'] = sell
return dataframe
def _run_backtest_1(strategy, fun, backtest_conf):
# strategy is a global (hidden as a singleton), so we
# emulate strategy being pure, by override/restore here
# if we dont do this, the override in strategy will carry over
# to other tests
old_buy = strategy.populate_buy_trend
old_sell = strategy.populate_sell_trend
strategy.populate_buy_trend = fun # Override
strategy.populate_sell_trend = fun # Override
results = backtest(backtest_conf)
strategy.populate_buy_trend = old_buy # restore override
strategy.populate_sell_trend = old_sell # restore override
return results
def test_generate_text_table():
results = pd.DataFrame(
{
@ -127,21 +192,90 @@ def simple_backtest(config, contour, num_results):
assert len(results) == num_results
# Test backtest on offline data
# loaded by freqdata/optimize/__init__.py::load_data()
# Test backtest using offline data (testdata directory)
def test_backtest2(default_conf, mocker, default_strategy):
def test_backtest_ticks(default_conf, mocker, default_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})
ticks = [1, 5]
fun = default_strategy.populate_buy_trend
for tick in ticks:
backtest_conf = _make_backtest_conf(conf=default_conf)
results = _run_backtest_1(default_strategy, fun, backtest_conf)
assert not results.empty
def test_backtest_clash_buy_sell(default_conf, mocker, default_strategy):
mocker.patch.dict('freqtrade.main._CONF', default_conf)
# Override the default buy trend function in our default_strategy
def fun(dataframe=None):
buy_value = 1
sell_value = 1
return _trend(dataframe, buy_value, sell_value)
backtest_conf = _make_backtest_conf(conf=default_conf)
results = _run_backtest_1(default_strategy, fun, backtest_conf)
assert results.empty
def test_backtest_only_sell(default_conf, mocker, default_strategy):
mocker.patch.dict('freqtrade.main._CONF', default_conf)
# Override the default buy trend function in our default_strategy
def fun(dataframe=None):
buy_value = 0
sell_value = 1
return _trend(dataframe, buy_value, sell_value)
backtest_conf = _make_backtest_conf(conf=default_conf)
results = _run_backtest_1(default_strategy, fun, backtest_conf)
assert results.empty
def test_backtest_alternate_buy_sell(default_conf, mocker, default_strategy):
mocker.patch.dict('freqtrade.main._CONF', default_conf)
backtest_conf = _make_backtest_conf(conf=default_conf, pair='BTC_UNITEST')
results = _run_backtest_1(default_strategy, _trend_alternate,
backtest_conf)
assert len(results) == 3
def test_backtest_record(default_conf, mocker, default_strategy):
names = []
records = []
mocker.patch.dict('freqtrade.main._CONF', default_conf)
mocker.patch('freqtrade.misc.file_dump_json',
new=lambda n, r: (names.append(n), records.append(r)))
backtest_conf = _make_backtest_conf(
conf=default_conf,
pair='BTC_UNITEST',
record="trades"
)
results = _run_backtest_1(default_strategy, _trend_alternate,
backtest_conf)
assert len(results) == 3
# Assert file_dump_json was only called once
assert names == ['backtest-result.json']
records = records[0]
# Ensure records are of correct type
assert len(records) == 3
# ('BTC_UNITEST', 0.00331158, '1510684320', '1510691700', 0, 117)
# Below follows just a typecheck of the schema/type of trade-records
oix = None
for (pair, profit, date_buy, date_sell, buy_index, dur) in records:
assert pair == 'BTC_UNITEST'
isinstance(profit, float)
# FIX: buy/sell should be converted to ints
isinstance(date_buy, str)
isinstance(date_sell, str)
isinstance(buy_index, pd._libs.tslib.Timestamp)
if oix:
assert buy_index > oix
oix = buy_index
assert dur > 0
def test_processed(default_conf, mocker, default_strategy):
mocker.patch.dict('freqtrade.main._CONF', default_conf)
dict_of_tickerrows = load_data_test('raise')
@ -191,3 +325,29 @@ def test_backtest_start(default_conf, mocker, caplog):
assert ('freqtrade.optimize.backtesting',
logging.INFO,
line) in caplog.record_tuples
def test_backtest_start_live(default_strategy, default_conf, mocker, caplog):
caplog.set_level(logging.INFO)
default_conf['exchange']['pair_whitelist'] = ['BTC_UNITEST']
mocker.patch('freqtrade.exchange.get_ticker_history',
new=lambda n, i: _load_pair_as_ticks(n, i))
mocker.patch.dict('freqtrade.main._CONF', default_conf)
mocker.patch('freqtrade.misc.load_config', new=lambda s: default_conf)
args = MagicMock()
args.ticker_interval = 1
args.level = 10
args.live = True
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 2017-11-14T19:32:00+00:00 '
'up to 2017-11-14T22:59:00+00:00 (0 days)..']
for line in exists:
assert ('freqtrade.optimize.backtesting',
logging.INFO,
line) in caplog.record_tuples

View File

@ -1,9 +1,15 @@
# pragma pylint: disable=missing-docstring,W0212,C0103
import logging
from unittest.mock import MagicMock
import pandas as pd
from freqtrade.optimize.hyperopt import calculate_loss, TARGET_TRADES, EXPECTED_MAX_PROFIT, start, \
log_results, save_trials, read_trials, generate_roi_table
import freqtrade.optimize.hyperopt as hyperopt
def test_loss_calculation_prefer_correct_trade_count():
correct = calculate_loss(1, TARGET_TRADES, 20)
@ -250,3 +256,26 @@ def test_roi_table_generation():
'roi_p3': 3,
}
assert generate_roi_table(params) == {'0': 6, '15': 3, '25': 1, '30': 0}
# test log_trials_result
# test buy_strategy_generator def populate_buy_trend
# test optimizer if 'ro_t1' in params
def test_format_results():
trades = [('BTC_ETH', 2, 2, 123),
('BTC_LTC', 1, 1, 123),
('BTC_XRP', -1, -2, -246)]
labels = ['currency', 'profit_percent', 'profit_BTC', 'duration']
df = pd.DataFrame.from_records(trades, columns=labels)
x = hyperopt.format_results(df)
assert x.find(' 66.67%')
def test_signal_handler(mocker):
m = MagicMock()
mocker.patch('sys.exit', m)
mocker.patch('freqtrade.optimize.hyperopt.save_trials', m)
mocker.patch('freqtrade.optimize.hyperopt.log_trials_result', m)
hyperopt.signal_handler(9, None)
assert m.call_count == 3

View File

@ -19,8 +19,8 @@ def test_dataframe_correct_columns(result):
def test_dataframe_correct_length(result):
# no idea what this check truly does - should we just remove it?
assert len(result.index) == 14397
dataframe = parse_ticker_dataframe(result)
assert len(result.index) == len(dataframe.index)
def test_populates_buy_trend(result):