stable/freqtrade/tests/optimize/test_backtest_detail.py

# pragma pylint: disable=missing-docstring, W0212, line-too-long, C0103, unused-argument
import logging
from unittest.mock import MagicMock
from typing import NamedTuple

from pandas import DataFrame
import pytest
from arrow import get as getdate


from freqtrade.optimize.backtesting import Backtesting
from freqtrade.tests.conftest import patch_exchange, log_has


class BTContainer(NamedTuple):
    """
    NamedTuple Defining BacktestResults inputs.
    """
    data: DataFrame
    stop_loss: float
    roi: float
    trades: int
    profit_perc: float
    sl: bool
    remains: bool


columns = ['date', 'open', 'high', 'low', 'close', 'volume', 'buy', 'sell']
data_profit = DataFrame([
    [getdate('2018-07-08 18:00:00').datetime, 0.0009910,
     0.001011, 0.00098618, 0.001000, 12345, 1, 0],
    [getdate('2018-07-08 19:00:00').datetime, 0.001000,
     0.001010, 0.0009900, 0.0009900, 12345, 0, 0],
    [getdate('2018-07-08 20:00:00').datetime, 0.0009900,
     0.001011, 0.00091618, 0.0009900, 12345, 0, 0],
    [getdate('2018-07-08 21:00:00').datetime, 0.001000,
     0.001011, 0.00098618, 0.001100, 12345, 0, 1],
    [getdate('2018-07-08 22:00:00').datetime, 0.001000,
     0.001011, 0.00098618, 0.0009900, 12345, 0, 0]
], columns=columns)

tc_profit1 = BTContainer(data=data_profit, stop_loss=-0.01, roi=1, trades=1,
                         profit_perc=0.10557, sl=False, remains=False)  # should be stoploss - drops 8%
tc_profit2 = BTContainer(data=data_profit, stop_loss=-0.10, roi=1,
                         trades=1, profit_perc=0.10557, sl=True, remains=False)


tc_loss0 = BTContainer(data=DataFrame([
    [getdate('2018-07-08 18:00:00').datetime, 0.0009910,
     0.001011, 0.00098618, 0.001000, 12345, 1, 0],
    [getdate('2018-07-08 19:00:00').datetime, 0.001000,
     0.001010, 0.0009900, 0.001000, 12345, 0, 0],
    [getdate('2018-07-08 20:00:00').datetime, 0.001000,
     0.001011, 0.0010618, 0.00091618, 12345, 0, 0],
    [getdate('2018-07-08 21:00:00').datetime, 0.001000,
     0.001011, 0.00098618, 0.00091618, 12345, 0, 0],
    [getdate('2018-07-08 22:00:00').datetime, 0.001000,
     0.001011, 0.00098618, 0.00091618, 12345, 0, 0]
], columns=columns),
    stop_loss=-0.05, roi=1, trades=1, profit_perc=-0.08839, sl=True, remains=False)


# Test 1 Minus 8% Close
# Candle Data for test 1 – close at -8% (9200)
# Test with Stop-loss at 1%
# TC1: Stop-Loss Triggered 1% loss
tc1 = BTContainer(data=DataFrame([
    [getdate('2018-06-10 08:00:00').datetime, 10000, 10050, 9950, 9975, 12345, 1, 0],
    [getdate('2018-06-10 09:00:00').datetime, 9975, 10025, 9200, 9200, 12345, 0, 0],
    [getdate('2018-06-10 10:00:00').datetime, 9950, 10000, 9960, 9955, 12345, 0, 0],
    [getdate('2018-06-10 11:00:00').datetime, 9955, 9975, 9955, 9990, 12345, 0, 0],
    [getdate('2018-06-10 12:00:00').datetime, 9990, 9990, 9990, 9900, 12345, 0, 0]
], columns=columns),
    # stop_loss=-0.01, roi=1, trades=1, profit_perc=-0.01, sl=True, remains=False)  # should be
    stop_loss=-0.01, roi=1, trades=1, profit_perc=0.071, sl=False, remains=True)  #


# Test 2 Minus 4% Low, minus 1% close
# Candle Data for test 2
# Test with Stop-Loss at 3%
# TC2: Stop-Loss Triggered 3% Loss
tc2 = BTContainer(data=DataFrame([
    [getdate('2018-06-10 08:00:00').datetime, 10000, 10050, 9950, 9975, 12345, 1, 0],
    [getdate('2018-06-10 09:00:00').datetime, 9975, 10025, 9925, 9950, 12345, 0, 0],
    [getdate('2018-06-10 10:00:00').datetime, 9950, 10000, 9600, 9925, 12345, 0, 0],
    [getdate('2018-06-10 11:00:00').datetime, 9925, 9975, 9875, 9900, 12345, 0, 0],
    [getdate('2018-06-10 12:00:00').datetime, 9900, 9950, 9850, 9900, 12345, 0, 0]
], columns=columns),
    # stop_loss=-0.03, roi=1, trades=1, profit_perc=-0.03, sl=True, remains=False)  #should be
    stop_loss=-0.03, roi=1, trades=1, profit_perc=-0.00999, sl=False, remains=True)  #


# Test 3 Candle drops 4%, Recovers 1%.
#               Entry Criteria Met
# 	            Candle drops 20%
# Candle Data for test 3
# Test with Stop-Loss at 2%
# TC3: Trade-A: Stop-Loss Triggered 2% Loss
#          Trade-B: Stop-Loss Triggered 2% Loss
tc3 = BTContainer(data=DataFrame([
    [getdate('2018-06-10 08:00:00').datetime, 10000, 10050, 9950, 9975, 12345, 1, 0],
    [getdate('2018-06-10 09:00:00').datetime, 9975, 10025, 9600, 9950, 12345, 0, 0],
    [getdate('2018-06-10 10:00:00').datetime, 9950, 10000, 9900, 9925, 12345, 1, 0],
    [getdate('2018-06-10 11:00:00').datetime, 9925, 9975, 8000, 8000, 12345, 0, 0],
    [getdate('2018-06-10 12:00:00').datetime, 9900, 9950, 9950, 9900, 12345, 0, 0]
], columns=columns),
    # stop_loss=-0.02, roi=1, trades=2, profit_perc=-0.4, sl=True, remains=False)  #should be
    stop_loss=-0.02, roi=1, trades=1, profit_perc=-0.19999, sl=True, remains=False)  #


# Test 4 Minus 3% / recovery +15%
# Candle Data for test 4 – Candle drops 3% Closed 15% up
# Test with Stop-loss at 2% ROI 6%
# TC4: Stop-Loss Triggered 2% Loss
tc4 = BTContainer(data=DataFrame([
    [getdate('2018-06-10 08:00:00').datetime, 10000, 10050, 9950, 9975, 12345, 1, 0],
    [getdate('2018-06-10 09:00:00').datetime, 9975, 11500, 9700, 11500, 12345, 0, 0],
    [getdate('2018-06-10 10:00:00').datetime, 9950, 10000, 9900, 9925, 12345, 0, 0],
    [getdate('2018-06-10 11:00:00').datetime, 9925, 9975, 9875, 9900, 12345, 0, 0],
    [getdate('2018-06-10 12:00:00').datetime, 9900, 9950, 9850, 9900, 12345, 0, 0]
], columns=columns),
    # stop_loss=-0.02, roi=0.06, trades=1, profit_perc=-0.02, sl=False, remains=False)  #should be
    stop_loss=-0.02, roi=0.06, trades=1, profit_perc=-0.141, sl=True, remains=False)

# Test 5 / Drops 0.5% Closes +20%
# Candle Data for test 5
# Set stop-loss at 1% ROI 3%
# TC5: ROI triggers 3% Gain
tc5 = BTContainer(data=DataFrame([
    [getdate('2018-06-10 08:00:00').datetime, 10000, 10050, 9960, 9975, 12345, 1, 0],
    [getdate('2018-06-10 09:00:00').datetime, 9975, 10050, 9950, 9975, 12345, 0, 0],
    [getdate('2018-06-10 10:00:00').datetime, 9950, 12000, 9950, 12000, 12345, 0, 0],
    [getdate('2018-06-10 11:00:00').datetime, 9925, 9975, 9945, 9900, 12345, 0, 0],
    [getdate('2018-06-10 12:00:00').datetime, 9900, 9950, 9850, 9900, 12345, 0, 0]
], columns=columns),
    # stop_loss=-0.01, roi=0.03, trades=1, profit_perc=0.03, sl=False, remains=False)  #should be
    stop_loss=-0.01, roi=0.03, trades=1, profit_perc=0.197, sl=False, remains=False)

# Test 6 / Drops 3% / Recovers 6% Positive / Closes 1% positve
# Candle Data for test 6
# Set stop-loss at 2% ROI at 5%
# TC6: Stop-Loss triggers 2% Loss
tc6 = BTContainer(data=DataFrame([
    [getdate('2018-06-10 08:00:00').datetime, 10000, 10050, 9950, 9975, 12345, 1, 0],
    [getdate('2018-06-10 09:00:00').datetime, 9975, 10600, 9700, 10100, 12345, 0, 0],
    [getdate('2018-06-10 10:00:00').datetime, 9950, 10000, 9900, 9925, 12345, 0, 0],
    [getdate('2018-06-10 11:00:00').datetime, 9925, 9975, 9945, 9900, 12345, 0, 0],
    [getdate('2018-06-10 12:00:00').datetime, 9900, 9950, 9850, 9900, 12345, 0, 0]
], columns=columns),
    # stop_loss=-0.02, roi=0.05, trades=1, profit_perc=-0.02, sl=False, remains=False)  #should be
    stop_loss=-0.02, roi=0.05, trades=1, profit_perc=-0.025, sl=False, remains=True)  #

# Test 7 - 6% Positive / 1% Negative / Close 1% Positve
# Candle Data for test 7
# Set stop-loss at 2% ROI at 3%
# TC7: ROI Triggers 3% Gain
tc7 = BTContainer(data=DataFrame([
    [getdate('2018-06-10 08:00:00').datetime, 10000, 10050, 9950, 9975, 12345, 1, 0],
    [getdate('2018-06-10 09:00:00').datetime, 9975, 10600, 9900, 10100, 12345, 0, 0],
    [getdate('2018-06-10 10:00:00').datetime, 9950, 10000, 9900, 9925, 12345, 0, 0],
    [getdate('2018-06-10 11:00:00').datetime, 9925, 9975, 9945, 9900, 12345, 0, 0],
    [getdate('2018-06-10 12:00:00').datetime, 9900, 9950, 9850, 9900, 12345, 0, 0]
], columns=columns),
    # stop_loss=-0.02, roi=0.03, trades=1, profit_perc=-0.03, sl=False, remains=False)  #should be
    stop_loss=-0.02, roi=0.03, trades=1, profit_perc=-0.025, sl=False, remains=True)  #

TESTS = [
    # tc_profit1,
    # tc_profit2,
    # tc_loss0,
    tc1,
    tc2,
    tc3,
    tc4,
    tc5,
    tc6,
    tc7,
]


@pytest.mark.parametrize("data", TESTS)
def test_backtest_results(default_conf, fee, mocker, caplog, data) -> None:
    """
    run functional tests
    """
    default_conf["stoploss"] = data.stop_loss
    default_conf["minimal_roi"] = {"0": data.roi}
    mocker.patch('freqtrade.exchange.Exchange.get_fee', fee)
    mocker.patch.multiple('freqtrade.analyze.Analyze',
                          populate_sell_trend=MagicMock(return_value=data.data),
                          populate_buy_trend=MagicMock(return_value=data.data))
    patch_exchange(mocker)

    backtesting = Backtesting(default_conf)
    caplog.set_level(logging.DEBUG)

    pair = 'UNITTEST/BTC'
    # Dummy data as we mock the analyze functions
    data_processed = {pair: DataFrame()}
    results = backtesting.backtest(
        {
            'stake_amount': default_conf['stake_amount'],
            'processed': data_processed,
            'max_open_trades': 10,
            'realistic': True
        }
    )
    print(results.T)

    assert len(results) == data.trades
    assert round(results["profit_percent"].sum(), 3) == round(data.profit_perc, 3)
    if data.sl:
        assert log_has("Stop loss hit.", caplog.record_tuples)
    else:
        assert not log_has("Stop loss hit.", caplog.record_tuples)
    log_test = (f'Force_selling still open trade UNITTEST/BTC with '
                f'{results.iloc[-1].profit_percent} perc - {results.iloc[-1].profit_abs}')
    if data.remains:
        assert log_has(log_test,
                       caplog.record_tuples)
    else:
        assert not log_has(log_test,
                           caplog.record_tuples)