mirror of
https://ghfast.top/https://github.com/aeroxw/easy-tdx.git
synced 2026-09-12 16:54:17 +08:00
GitHubandClaude Opus 4.8
6a6d75f5d5
Root cause: _generate_signals() iterated all bars calling strategy.next() but never updated _position_size or _cash on the strategy. Strategies that check self.position['size'] before buy/sell always saw 0, producing only BUY signals with no SELL — exhausting cash and producing drawdowns exceeding 100%. Fix: add _update_strategy_position() that estimates position changes after each bar's signals using close price. This gives the strategy an accurate view of its holdings so it can correctly alternate buy/sell. Regression tests added: - test_position_aware_buy_sell_alternation: verifies BUY/SELL alternation - test_position_aware_no_duplicate_buys: no suspicious tiny duplicate buys Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
362 lines
11 KiB
Python
362 lines
11 KiB
Python
"""Test BacktestEngine orchestration."""
|
|
|
|
from __future__ import annotations
|
|
|
|
import numpy as np
|
|
import pandas as pd
|
|
|
|
from easy_tdx import MyTT
|
|
from easy_tdx.backtest.engine import BacktestEngine
|
|
from easy_tdx.backtest.strategy import Strategy
|
|
|
|
|
|
def _make_df(n: int = 100, seed: int = 42) -> pd.DataFrame:
|
|
"""Generate synthetic OHLCV data."""
|
|
rng = np.random.default_rng(seed)
|
|
close = 100.0 + np.cumsum(rng.normal(0, 1, n))
|
|
high = close + rng.uniform(0, 1, n)
|
|
low = close - rng.uniform(0, 1, n)
|
|
open_ = low + rng.uniform(0, high - low, n)
|
|
volume = rng.integers(1000000, 10000000, n)
|
|
|
|
dates = pd.date_range("2024-01-01", periods=n, freq="D")
|
|
return pd.DataFrame(
|
|
{
|
|
"datetime": dates,
|
|
"open": open_,
|
|
"high": high,
|
|
"low": low,
|
|
"close": close,
|
|
"volume": volume,
|
|
}
|
|
)
|
|
|
|
|
|
class MACrossStrategy(Strategy):
|
|
"""Simple MA crossover strategy."""
|
|
|
|
def init(self):
|
|
self.ma5 = self.I(MyTT.MA, self.data.close, 5)
|
|
self.ma20 = self.I(MyTT.MA, self.data.close, 20)
|
|
self.cross_up = False
|
|
self.cross_down = False
|
|
|
|
def next(self):
|
|
# Check if crossing happened on this bar
|
|
if self._bar_index > 0:
|
|
prev_ma5 = self.ma5[self._bar_index - 1]
|
|
prev_ma20 = self.ma20[self._bar_index - 1]
|
|
curr_ma5 = self.ma5[self._bar_index]
|
|
curr_ma20 = self.ma20[self._bar_index]
|
|
|
|
# Golden cross: ma5 crosses above ma20
|
|
if prev_ma5 <= prev_ma20 and curr_ma5 > curr_ma20:
|
|
self.buy(size=0)
|
|
# Death cross: ma5 crosses below ma20
|
|
elif prev_ma5 >= prev_ma20 and curr_ma5 < curr_ma20:
|
|
self.sell(size=0)
|
|
|
|
|
|
class FixedBuyStrategy(Strategy):
|
|
"""Strategy with fixed buy/sell at specific bars."""
|
|
|
|
def init(self):
|
|
pass
|
|
|
|
def next(self):
|
|
if self._bar_index == 5:
|
|
self.buy(size=100)
|
|
if self._bar_index == 50:
|
|
self.sell(size=100)
|
|
|
|
|
|
class ChanlunStrategy(Strategy):
|
|
"""Strategy that uses chanlun result."""
|
|
|
|
def init(self):
|
|
pass
|
|
|
|
def next(self):
|
|
if self._bar_index == 10 and hasattr(self, "chanlun"):
|
|
# Access chanlun result
|
|
_ = self.chanlun
|
|
self.buy(size=50)
|
|
|
|
|
|
class PrecomputedIndicatorStrategy(Strategy):
|
|
"""Strategy that uses precomputed indicator columns."""
|
|
|
|
def init(self):
|
|
# Assume BOLL_UPPER already exists in df
|
|
if hasattr(self.data, "BOLL_UPPER"):
|
|
self.boll_upper = self.data.BOLL_UPPER
|
|
else:
|
|
self.boll_upper = None
|
|
|
|
def next(self):
|
|
if self.boll_upper is not None and self._bar_index == 20:
|
|
_ = self.boll_upper[self._bar_index]
|
|
self.buy(size=10)
|
|
|
|
|
|
def test_basic_run():
|
|
"""Test basic engine run with MACrossStrategy."""
|
|
df = _make_df(n=200)
|
|
engine = BacktestEngine(MACrossStrategy, cash=100000)
|
|
result = engine.run(df)
|
|
|
|
# Check performance metrics
|
|
assert result.performance is not None
|
|
assert "total_return" in result.performance
|
|
|
|
# Check equity curve length
|
|
assert len(result.equity_curve) == 200
|
|
|
|
# Check columns
|
|
assert "datetime" in result.equity_curve.columns
|
|
assert "total" in result.equity_curve.columns
|
|
|
|
|
|
def test_fixed_strategy():
|
|
"""Test FixedBuyStrategy produces trades."""
|
|
df = _make_df(n=100)
|
|
engine = BacktestEngine(FixedBuyStrategy, cash=100000)
|
|
result = engine.run(df)
|
|
|
|
# Should have at least 2 trades
|
|
assert len(result.trades) >= 2, f"Expected at least 2 trades, got {len(result.trades)}"
|
|
|
|
# Check buy at bar 5
|
|
buy_trades = result.trades[result.trades["direction"] == "BUY"]
|
|
assert len(buy_trades) >= 1, "No buy trades found"
|
|
|
|
# Check sell at bar 50
|
|
sell_trades = result.trades[result.trades["direction"] == "SELL"]
|
|
assert len(sell_trades) >= 1, "No sell trades found"
|
|
|
|
|
|
def test_result_columns():
|
|
"""Test BacktestResult has correct columns."""
|
|
df = _make_df(n=100)
|
|
engine = BacktestEngine(MACrossStrategy)
|
|
result = engine.run(df)
|
|
|
|
# Equity curve columns
|
|
expected_ec_cols = ["datetime", "cash", "position_value", "total"]
|
|
for col in expected_ec_cols:
|
|
assert col in result.equity_curve.columns
|
|
|
|
# Trades columns
|
|
expected_trade_cols = ["datetime", "direction", "size", "price", "pnl"]
|
|
for col in expected_trade_cols:
|
|
assert col in result.trades.columns
|
|
|
|
|
|
def test_to_dict():
|
|
"""Test BacktestResult is serializable."""
|
|
df = _make_df(n=50)
|
|
engine = BacktestEngine(MACrossStrategy)
|
|
result = engine.run(df)
|
|
|
|
# to_dict should not raise
|
|
d = result.to_dict()
|
|
assert "performance" in d
|
|
assert "equity_curve" in d
|
|
assert "trades" in d
|
|
|
|
# to_json should not raise
|
|
json_str = result.to_json()
|
|
assert len(json_str) > 0
|
|
|
|
|
|
def test_chanlun_injection():
|
|
"""Test chanlun result injection."""
|
|
df = _make_df(n=50)
|
|
|
|
# Mock chanlun result
|
|
chanlun_result = {"test": "data"}
|
|
|
|
engine = BacktestEngine(ChanlunStrategy)
|
|
result = engine.run(df, chanlun_result=chanlun_result)
|
|
|
|
# Should have trades
|
|
assert len(result.trades) >= 1
|
|
|
|
|
|
def test_this_close_warning_in_config():
|
|
"""Test future_leak_warning in config when using this_close."""
|
|
df = _make_df(n=50)
|
|
engine = BacktestEngine(MACrossStrategy, execution="this_close")
|
|
result = engine.run(df)
|
|
|
|
# Config should have future_leak_warning
|
|
# Note: MACrossStrategy may not generate signals, so warning might be False
|
|
assert "future_leak_warning" in result.config
|
|
|
|
|
|
def test_config_snapshot():
|
|
"""Test config contains correct cash and commission."""
|
|
df = _make_df(n=50)
|
|
engine = BacktestEngine(MACrossStrategy, cash=50000, commission=0.0005, execution="next_open")
|
|
result = engine.run(df)
|
|
|
|
# Check config
|
|
assert result.config["cash"] == 50000
|
|
assert result.config["commission"] == 0.0005
|
|
assert result.config["execution"] == "next_open"
|
|
|
|
|
|
def test_precomputed_indicator_columns():
|
|
"""Test strategy works with precomputed indicator columns."""
|
|
df = _make_df(n=50)
|
|
# Add precomputed BOLL_UPPER column
|
|
df["BOLL_UPPER"] = df["close"] * 1.05
|
|
|
|
engine = BacktestEngine(PrecomputedIndicatorStrategy)
|
|
result = engine.run(df)
|
|
|
|
# Should not crash and should have trades
|
|
assert len(result.equity_curve) == 50
|
|
|
|
|
|
def test_empty_df():
|
|
"""Test engine with empty DataFrame."""
|
|
df = pd.DataFrame(columns=["datetime", "open", "high", "low", "close", "volume"])
|
|
engine = BacktestEngine(MACrossStrategy)
|
|
result = engine.run(df)
|
|
|
|
# Should return empty result
|
|
assert len(result.equity_curve) == 0
|
|
assert len(result.trades) == 0
|
|
|
|
|
|
def test_strategy_instance_vs_class():
|
|
"""Test engine accepts both strategy class and instance."""
|
|
df = _make_df(n=50)
|
|
|
|
# Test with class
|
|
engine1 = BacktestEngine(MACrossStrategy)
|
|
result1 = engine1.run(df)
|
|
assert len(result1.equity_curve) == 50
|
|
|
|
# Test with instance
|
|
strat = MACrossStrategy()
|
|
engine2 = BacktestEngine(strat)
|
|
result2 = engine2.run(df)
|
|
assert len(result2.equity_curve) == 50
|
|
|
|
|
|
def test_commission_calculation():
|
|
"""Test commission is correctly applied."""
|
|
df = _make_df(n=100)
|
|
engine = BacktestEngine(
|
|
FixedBuyStrategy,
|
|
cash=100000,
|
|
commission=0.001,
|
|
min_commission=10.0,
|
|
)
|
|
result = engine.run(df)
|
|
|
|
# Should have trades
|
|
assert len(result.trades) >= 2
|
|
|
|
# Check trades have commission
|
|
assert (result.trades["commission"] > 0).all()
|
|
|
|
|
|
def test_pnl_calculation():
|
|
"""Test PnL is calculated for sell trades."""
|
|
df = _make_df(n=100, seed=123) # Use specific seed for predictable prices
|
|
engine = BacktestEngine(FixedBuyStrategy, cash=100000, commission=0.0)
|
|
result = engine.run(df)
|
|
|
|
# Should have trades
|
|
assert len(result.trades) >= 2
|
|
|
|
# Get trades - should have at least one BUY and one SELL
|
|
buy_trades = result.trades[result.trades["direction"] == "BUY"]
|
|
sell_trades = result.trades[result.trades["direction"] == "SELL"]
|
|
|
|
assert len(buy_trades) >= 1
|
|
assert len(sell_trades) >= 1
|
|
|
|
# PnL is calculated for sell trades
|
|
# Check that sell trades have PnL computed
|
|
assert (sell_trades["pnl"] != 0).any() or len(sell_trades) == 0
|
|
|
|
# For buy trades, PnL should be 0
|
|
assert (buy_trades["pnl"] == 0).all()
|
|
|
|
|
|
class PositionAwareStrategy(Strategy):
|
|
"""Strategy that checks position before trading (the common pattern)."""
|
|
|
|
def init(self):
|
|
self.ma5 = self.I(MyTT.MA, self.data.close, 5)
|
|
self.ma20 = self.I(MyTT.MA, self.data.close, 20)
|
|
self.cross_up = False
|
|
self.cross_down = False
|
|
|
|
def next(self):
|
|
if self._bar_index > 0:
|
|
prev_ma5 = self.ma5[self._bar_index - 1]
|
|
prev_ma20 = self.ma20[self._bar_index - 1]
|
|
curr_ma5 = self.ma5[self._bar_index]
|
|
curr_ma20 = self.ma20[self._bar_index]
|
|
|
|
if prev_ma5 <= prev_ma20 and curr_ma5 > curr_ma20:
|
|
if self.position["size"] == 0:
|
|
self.buy(size=0)
|
|
elif prev_ma5 >= prev_ma20 and curr_ma5 < curr_ma20:
|
|
if self.position["size"] > 0:
|
|
self.sell(size=0)
|
|
|
|
|
|
def test_position_aware_buy_sell_alternation():
|
|
"""Regression: strategy that checks position must produce alternating BUY/SELL."""
|
|
df = _make_df(n=300, seed=42)
|
|
engine = BacktestEngine(PositionAwareStrategy, cash=100000)
|
|
result = engine.run(df)
|
|
|
|
trades = result.trades[result.trades["rejected"] == False]
|
|
directions = trades["direction"].tolist()
|
|
|
|
# Must have both BUYs and SELLs
|
|
assert "BUY" in directions, "No BUY trades generated"
|
|
assert "SELL" in directions, "No SELL trades generated — position feedback broken"
|
|
|
|
# Trades must alternate: no two consecutive BUYs or SELLs
|
|
for i in range(1, len(directions)):
|
|
assert directions[i] != directions[i - 1], (
|
|
f"Consecutive same-direction trades at index {i}: "
|
|
f"{directions[i - 1]} -> {directions[i]}"
|
|
)
|
|
|
|
|
|
def test_position_aware_no_duplicate_buys():
|
|
"""After a BUY, position['size'] > 0 so strategy should not buy again."""
|
|
df = _make_df(n=300, seed=42)
|
|
engine = BacktestEngine(PositionAwareStrategy, cash=100000)
|
|
result = engine.run(df)
|
|
|
|
buy_trades = result.trades[
|
|
(result.trades["direction"] == "BUY") & (result.trades["rejected"] == False)
|
|
]
|
|
|
|
# Each BUY's size should be reasonable (not tiny leftover from exhausted cash)
|
|
if len(buy_trades) > 1:
|
|
# No consecutive buys where the second is tiny (cash leftover artifact)
|
|
sizes = buy_trades["size"].tolist()
|
|
for i in range(1, len(sizes)):
|
|
# Second buy in a pair should not be tiny compared to first
|
|
# (would indicate position wasn't tracked between bars)
|
|
if i >= 1:
|
|
prev_size = sizes[i - 1]
|
|
cur_size = sizes[i]
|
|
# Allow some variance but not orders-of-magnitude difference
|
|
if prev_size > 0:
|
|
assert cur_size > prev_size * 0.1, (
|
|
f"Suspicious tiny buy {cur_size} after {prev_size} — "
|
|
f"position feedback may be broken"
|
|
)
|