mirror of
https://ghfast.top/https://github.com/aeroxw/easy_tdx_max.git
synced 2026-09-12 15:44:18 +08:00
feat: multi-factor combo backtest engine (v1.9.0)
- Add backtest/combo.py: CombinationRunner, extract_factor_signals, combine_masks - Signal merge modes: AND / OR / MAJORITY (majority default) - CLI: --combo-strategies and --combo-mode for easy-tdx backtest - run_all_strategies.py: --combo 2 --combo 3 auto-screen best combos - Fix MyTT MFI/CR divide-by-zero RuntimeWarning - 14 new unit tests, 328 total passing
This commit is contained in:
Justin Gu
@@ -0,0 +1,396 @@
|
||||
"""测试多因子组合回测模块。"""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import numpy as np
|
||||
import pandas as pd
|
||||
import pytest
|
||||
from easy_tdx import MyTT
|
||||
from easy_tdx.backtest.strategy import Strategy
|
||||
|
||||
# ── 辅助函数 ─────────────────────────────────────────────────────────────────────
|
||||
|
||||
|
||||
def _make_df(n: int = 100, seed: int = 42) -> pd.DataFrame:
|
||||
"""构造合成 OHLCV DataFrame。"""
|
||||
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(1_000_000, 10_000_000, 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,
|
||||
}
|
||||
)
|
||||
|
||||
|
||||
# ── 测试用策略 ─────────────────────────────────────────────────────────────────
|
||||
|
||||
|
||||
class BuyOnBar3Strategy(Strategy):
|
||||
"""在第 3 根 bar 买入，第 8 根 bar 卖出（确定性策略，方便断言）。"""
|
||||
|
||||
def init(self) -> None:
|
||||
pass
|
||||
|
||||
def next(self) -> None:
|
||||
if self._bar_index == 3 and self.position["size"] == 0:
|
||||
self.buy(size=0)
|
||||
elif self._bar_index == 8 and self.position["size"] > 0:
|
||||
self.sell(size=0)
|
||||
|
||||
|
||||
class BuyOnBar5Strategy(Strategy):
|
||||
"""在第 5 根 bar 买入，第 10 根 bar 卖出。"""
|
||||
|
||||
def init(self) -> None:
|
||||
pass
|
||||
|
||||
def next(self) -> None:
|
||||
if self._bar_index == 5 and self.position["size"] == 0:
|
||||
self.buy(size=0)
|
||||
elif self._bar_index == 10 and self.position["size"] > 0:
|
||||
self.sell(size=0)
|
||||
|
||||
|
||||
class BuyOnBar7Strategy(Strategy):
|
||||
"""在第 7 根 bar 买入，第 12 根 bar 卖出。"""
|
||||
|
||||
def init(self) -> None:
|
||||
pass
|
||||
|
||||
def next(self) -> None:
|
||||
if self._bar_index == 7 and self.position["size"] == 0:
|
||||
self.buy(size=0)
|
||||
elif self._bar_index == 12 and self.position["size"] > 0:
|
||||
self.sell(size=0)
|
||||
|
||||
|
||||
class NeverTradeStrategy(Strategy):
|
||||
"""从不交易的策略。"""
|
||||
|
||||
def init(self) -> None:
|
||||
pass
|
||||
|
||||
def next(self) -> None:
|
||||
pass
|
||||
|
||||
|
||||
# ── extract_factor_signals 测试 ──────────────────────────────────────────────
|
||||
|
||||
|
||||
class TestExtractFactorSignals:
|
||||
"""测试因子信号提取。"""
|
||||
|
||||
def test_basic_buy_sell_signals(self) -> None:
|
||||
"""BuyOnBar3Strategy 应在第 3 根 bar 买入、第 8 根 bar 卖出。"""
|
||||
from easy_tdx.backtest.combo import extract_factor_signals
|
||||
|
||||
df = _make_df()
|
||||
fs = extract_factor_signals(BuyOnBar3Strategy, df)
|
||||
|
||||
assert fs.name == "BuyOnBar3Strategy"
|
||||
assert len(fs.buy_mask) == len(df)
|
||||
assert len(fs.sell_mask) == len(df)
|
||||
|
||||
# 第 3 根 bar 买入
|
||||
assert fs.buy_mask[3] is np.True_
|
||||
# 其他 bar 不买入
|
||||
buy_count = int(np.sum(fs.buy_mask))
|
||||
assert buy_count == 1
|
||||
|
||||
# 第 8 根 bar 卖出
|
||||
assert fs.sell_mask[8] is np.True_
|
||||
sell_count = int(np.sum(fs.sell_mask))
|
||||
assert sell_count == 1
|
||||
|
||||
def test_never_trade_strategy(self) -> None:
|
||||
"""从不交易的策略应产生全空遮罩。"""
|
||||
from easy_tdx.backtest.combo import extract_factor_signals
|
||||
|
||||
df = _make_df()
|
||||
fs = extract_factor_signals(NeverTradeStrategy, df)
|
||||
|
||||
assert int(np.sum(fs.buy_mask)) == 0
|
||||
assert int(np.sum(fs.sell_mask)) == 0
|
||||
|
||||
def test_real_ma_strategy(self) -> None:
|
||||
"""真实 MA 策略的信号提取应有交易信号。"""
|
||||
from easy_tdx.backtest.combo import extract_factor_signals
|
||||
|
||||
# 用较长数据确保有交叉
|
||||
df = _make_df(n=200, seed=123)
|
||||
|
||||
class MAStrategy(Strategy):
|
||||
def init(self) -> None:
|
||||
self.ma5 = self.I(MyTT.MA, self.data.close, 5)
|
||||
self.ma20 = self.I(MyTT.MA, self.data.close, 20)
|
||||
|
||||
def next(self) -> None:
|
||||
if self._bar_index < 1:
|
||||
return
|
||||
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:
|
||||
self.buy(size=0)
|
||||
elif prev_ma5 >= prev_ma20 and curr_ma5 < curr_ma20:
|
||||
if self.position["size"] > 0:
|
||||
self.sell(size=0)
|
||||
|
||||
fs = extract_factor_signals(MAStrategy, df)
|
||||
# 200 bar 数据应该至少有一次买卖
|
||||
assert int(np.sum(fs.buy_mask)) >= 1
|
||||
assert int(np.sum(fs.sell_mask)) >= 0 # 可能没有卖（最后还在持仓）
|
||||
|
||||
|
||||
# ── combine_masks 测试 ──────────────────────────────────────────────────────
|
||||
|
||||
|
||||
class TestCombineMasks:
|
||||
"""测试信号遮罩合并。"""
|
||||
|
||||
def test_and_mode(self) -> None:
|
||||
"""AND 模式：只有所有因子都看多才买入。"""
|
||||
from easy_tdx.backtest.combo import FactorSignals, combine_masks
|
||||
|
||||
n = 20
|
||||
# 因子 A: bar 3 买入, bar 8 卖出
|
||||
fs_a = FactorSignals(
|
||||
name="A",
|
||||
buy_mask=np.array([i == 3 for i in range(n)]),
|
||||
sell_mask=np.array([i == 8 for i in range(n)]),
|
||||
)
|
||||
# 因子 B: bar 3 买入, bar 10 卖出
|
||||
fs_b = FactorSignals(
|
||||
name="B",
|
||||
buy_mask=np.array([i == 3 for i in range(n)]),
|
||||
sell_mask=np.array([i == 10 for i in range(n)]),
|
||||
)
|
||||
|
||||
buy, sell = combine_masks([fs_a, fs_b], mode="AND")
|
||||
|
||||
# bar 3 两因子都看多 → AND 买入
|
||||
assert buy[3] is np.True_
|
||||
# 其他 bar 不买入
|
||||
assert int(np.sum(buy)) == 1
|
||||
|
||||
# 卖出：只有 bar 8 两因子都看空？不是，A 在 8 卖，B 在 10 卖 → AND 无交集
|
||||
assert int(np.sum(sell)) == 0
|
||||
|
||||
def test_or_mode(self) -> None:
|
||||
"""OR 模式：任一因子看多就买入。"""
|
||||
from easy_tdx.backtest.combo import FactorSignals, combine_masks
|
||||
|
||||
n = 20
|
||||
fs_a = FactorSignals(
|
||||
name="A",
|
||||
buy_mask=np.array([i == 3 for i in range(n)]),
|
||||
sell_mask=np.array([i == 8 for i in range(n)]),
|
||||
)
|
||||
fs_b = FactorSignals(
|
||||
name="B",
|
||||
buy_mask=np.array([i == 5 for i in range(n)]),
|
||||
sell_mask=np.array([i == 10 for i in range(n)]),
|
||||
)
|
||||
|
||||
buy, sell = combine_masks([fs_a, fs_b], mode="OR")
|
||||
|
||||
# bar 3 和 bar 5 都应该买入
|
||||
assert buy[3] is np.True_
|
||||
assert buy[5] is np.True_
|
||||
assert int(np.sum(buy)) == 2
|
||||
|
||||
# bar 8 和 bar 10 都应该卖出
|
||||
assert sell[8] is np.True_
|
||||
assert sell[10] is np.True_
|
||||
assert int(np.sum(sell)) == 2
|
||||
|
||||
def test_majority_mode_2_factors(self) -> None:
|
||||
"""MAJORITY 模式（2 因子）：过半 = >1 → 需要两个都同意（等同 AND）。"""
|
||||
from easy_tdx.backtest.combo import FactorSignals, combine_masks
|
||||
|
||||
n = 20
|
||||
fs_a = FactorSignals(
|
||||
name="A",
|
||||
buy_mask=np.array([i == 3 for i in range(n)]),
|
||||
sell_mask=np.array([i == 8 for i in range(n)]),
|
||||
)
|
||||
fs_b = FactorSignals(
|
||||
name="B",
|
||||
buy_mask=np.array([i == 3 for i in range(n)]),
|
||||
sell_mask=np.array([i == 10 for i in range(n)]),
|
||||
)
|
||||
|
||||
buy, sell = combine_masks([fs_a, fs_b], mode="MAJORITY")
|
||||
|
||||
# 2 因子 MAJORITY: threshold = 2/2 = 1, need > 1 → need 2 个同意
|
||||
assert buy[3] is np.True_
|
||||
assert int(np.sum(buy)) == 1
|
||||
|
||||
def test_majority_mode_3_factors(self) -> None:
|
||||
"""MAJORITY 模式（3 因子）：过半 = >1.5 → 需要至少 2 个同意。"""
|
||||
from easy_tdx.backtest.combo import FactorSignals, combine_masks
|
||||
|
||||
n = 20
|
||||
fs_a = FactorSignals(
|
||||
name="A",
|
||||
buy_mask=np.array([i == 5 for i in range(n)]),
|
||||
sell_mask=np.array([i == 10 for i in range(n)]),
|
||||
)
|
||||
fs_b = FactorSignals(
|
||||
name="B",
|
||||
buy_mask=np.array([i == 5 for i in range(n)]),
|
||||
sell_mask=np.array([i == 12 for i in range(n)]),
|
||||
)
|
||||
fs_c = FactorSignals(
|
||||
name="C",
|
||||
buy_mask=np.array([i == 7 for i in range(n)]),
|
||||
sell_mask=np.array([i == 10 for i in range(n)]),
|
||||
)
|
||||
|
||||
buy, sell = combine_masks([fs_a, fs_b, fs_c], mode="MAJORITY")
|
||||
|
||||
# bar 5: A 和 B 同意（2/3），满足 > 1.5 → 买入
|
||||
assert buy[5] is np.True_
|
||||
# bar 7: 只有 C 同意（1/3），不满足 → 不买入
|
||||
assert buy[7] is np.False_
|
||||
assert int(np.sum(buy)) == 1
|
||||
|
||||
# bar 10: A 和 C 同意卖出（2/3），满足 → 卖出
|
||||
assert sell[10] is np.True_
|
||||
# bar 12: 只有 B 同意（1/3），不满足 → 不卖出
|
||||
assert sell[12] is np.False_
|
||||
assert int(np.sum(sell)) == 1
|
||||
|
||||
def test_invalid_mode_raises(self) -> None:
|
||||
"""无效合并模式应抛出 ValueError。"""
|
||||
from easy_tdx.backtest.combo import FactorSignals, combine_masks
|
||||
|
||||
n = 10
|
||||
fs = FactorSignals(
|
||||
name="X",
|
||||
buy_mask=np.zeros(n, dtype=bool),
|
||||
sell_mask=np.zeros(n, dtype=bool),
|
||||
)
|
||||
with pytest.raises(ValueError, match="不支持的合并模式"):
|
||||
combine_masks([fs, fs], mode="INVALID")
|
||||
|
||||
|
||||
# ── CombinationRunner 测试 ──────────────────────────────────────────────────
|
||||
|
||||
|
||||
class TestCombinationRunner:
|
||||
"""测试组合回测运行器。"""
|
||||
|
||||
def test_run_combination_and(self) -> None:
|
||||
"""运行 AND 组合：两个确定性策略在第 3 和第 5 bar 都不重叠 → AND 无交易。"""
|
||||
from easy_tdx.backtest.combo import CombinationRunner
|
||||
|
||||
df = _make_df()
|
||||
runner = CombinationRunner(
|
||||
strategy_classes=[BuyOnBar3Strategy, BuyOnBar5Strategy],
|
||||
df=df,
|
||||
cash=100000.0,
|
||||
)
|
||||
result = runner.run_combination(indices=[0, 1], mode="AND")
|
||||
|
||||
# BuyOnBar3 在 bar 3 买, BuyOnBar5 在 bar 5 买 → AND 无交集
|
||||
assert result.performance["total_trades"] == 0
|
||||
|
||||
def test_run_combination_or(self) -> None:
|
||||
"""运行 OR 组合：两个策略买入信号取并集。"""
|
||||
from easy_tdx.backtest.combo import CombinationRunner
|
||||
|
||||
df = _make_df()
|
||||
runner = CombinationRunner(
|
||||
strategy_classes=[BuyOnBar3Strategy, BuyOnBar5Strategy],
|
||||
df=df,
|
||||
cash=100000.0,
|
||||
)
|
||||
result = runner.run_combination(indices=[0, 1], mode="OR")
|
||||
|
||||
# OR 模式：bar 3 和 bar 5 都会触发买入信号
|
||||
# 但由于仓位管理，实际成交可能有限
|
||||
assert result.performance["total_trades"] >= 1
|
||||
|
||||
def test_screen_2_factors(self) -> None:
|
||||
"""screen 遍历所有 2 因子组合。"""
|
||||
from easy_tdx.backtest.combo import CombinationRunner
|
||||
|
||||
df = _make_df()
|
||||
runner = CombinationRunner(
|
||||
strategy_classes=[BuyOnBar3Strategy, BuyOnBar5Strategy, BuyOnBar7Strategy],
|
||||
df=df,
|
||||
cash=100000.0,
|
||||
)
|
||||
results = runner.screen(combo_sizes=(2,), mode="OR")
|
||||
|
||||
# C(3, 2) = 3 组合
|
||||
assert len(results) == 3
|
||||
|
||||
# 每个结果应有名称和绩效
|
||||
for r in results:
|
||||
assert " + " in r.name
|
||||
assert r.result.performance is not None
|
||||
|
||||
def test_screen_2_and_3_factors(self) -> None:
|
||||
"""screen 同时遍历 2 和 3 因子组合。"""
|
||||
from easy_tdx.backtest.combo import CombinationRunner
|
||||
|
||||
df = _make_df()
|
||||
runner = CombinationRunner(
|
||||
strategy_classes=[BuyOnBar3Strategy, BuyOnBar5Strategy, BuyOnBar7Strategy],
|
||||
df=df,
|
||||
cash=100000.0,
|
||||
)
|
||||
results = runner.screen(combo_sizes=(2, 3), mode="OR")
|
||||
|
||||
# C(3, 2) + C(3, 3) = 3 + 1 = 4
|
||||
assert len(results) == 4
|
||||
|
||||
# 应按总收益率降序排列
|
||||
returns = [r.result.performance["total_return"] for r in results]
|
||||
assert returns == sorted(returns, reverse=True)
|
||||
|
||||
def test_screen_with_never_trade(self) -> None:
|
||||
"""包含 NeverTradeStrategy 的组合不应崩溃。"""
|
||||
from easy_tdx.backtest.combo import CombinationRunner
|
||||
|
||||
df = _make_df()
|
||||
runner = CombinationRunner(
|
||||
strategy_classes=[BuyOnBar3Strategy, NeverTradeStrategy],
|
||||
df=df,
|
||||
cash=100000.0,
|
||||
)
|
||||
results = runner.screen(combo_sizes=(2,), mode="AND", filter_zero_trades=False)
|
||||
|
||||
# C(2, 2) = 1，AND 模式下无交易
|
||||
assert len(results) == 1
|
||||
assert results[0].result.performance["total_trades"] == 0
|
||||
|
||||
def test_screen_with_zero_trade_filtered(self) -> None:
|
||||
"""screen 默认过滤零交易组合（filter_zero_trades=True）。"""
|
||||
from easy_tdx.backtest.combo import CombinationRunner
|
||||
|
||||
df = _make_df()
|
||||
runner = CombinationRunner(
|
||||
strategy_classes=[BuyOnBar3Strategy, BuyOnBar5Strategy, NeverTradeStrategy],
|
||||
df=df,
|
||||
cash=100000.0,
|
||||
)
|
||||
# AND 模式：大部分组合无交集 → 零交易
|
||||
results = runner.screen(combo_sizes=(2,), mode="AND", filter_zero_trades=True)
|
||||
|
||||
# 所有 AND 组合应该都是零交易（信号不重叠），过滤后为空
|
||||
assert len(results) == 0
|
||||
Reference in New Issue
Block a user