gahow/quant
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

chore: backtest engine fee model, metrics, and strategy fixes

Browse Source 
- main.py: add IBKR-style tiered fee schedule (fee_base + fee_per_share),
  PIT universe support, and open-to-close execution improvements
- metrics.py: add raw_summary helper for JSON-safe metric export
- Misc strategy fixes: deprecation warnings, NaN handling

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
This commit is contained in:
Gahow Wang
2026-05-21 20:57:56 +08:00
parent 1f50253d13
commit 149a00c458
7 changed files with 154 additions and 74 deletions
Download Patch File Download Diff File Expand all files Collapse all files

119
strategies/ensemble_alpha.py
View File

@@ -84,17 +84,6 @@ class EnsembleAlphaStrategy(Strategy):
row_sums = raw.sum(axis=1).replace(0, np.nan)
signals = raw.div(row_sums, axis=0).fillna(0.0)
# === Tail-risk protection ===
if self.tail_protection:
# Portfolio equity proxy: equal-weight market return
mkt_ret = ret.mean(axis=1)
mkt_eq = (1 + mkt_ret).cumprod()
mkt_dd = mkt_eq / mkt_eq.cummax() - 1
in_tail = mkt_dd < self.tail_threshold
scale = pd.Series(1.0, index=data.index)
scale[in_tail] = self.tail_scale
signals = signals.mul(scale, axis=0)
# === Monthly rebalance ===
warmup = 252
rebal_mask = pd.Series(False, index=data.index)
@@ -105,6 +94,20 @@ class EnsembleAlphaStrategy(Strategy):
signals = signals.ffill().fillna(0.0)
signals.iloc[:warmup] = 0.0
# === Tail-risk protection (rebal-gated; NaN-safe market mean) ===
# Apply AFTER ffill so the scale changes only at rebal points,
# otherwise daily flips force half-out/half-in trades that burn
# the account through fixed per-trade fees.
if self.tail_protection:
mkt_ret = ret.mean(axis=1, skipna=True)
mkt_eq = (1 + mkt_ret.fillna(0.0)).cumprod()
mkt_dd = mkt_eq / mkt_eq.cummax() - 1
in_tail = mkt_dd < self.tail_threshold
scale_raw = pd.Series(1.0, index=data.index)
scale_raw[in_tail] = self.tail_scale
scale = scale_raw.where(rebal_mask, np.nan).ffill().fillna(1.0)
signals = signals.mul(scale, axis=0)
return signals.shift(1).fillna(0.0)
@@ -162,16 +165,6 @@ class EnhancedFactorComboStrategy(Strategy):
row_sums = raw.sum(axis=1).replace(0, np.nan)
signals = raw.div(row_sums, axis=0).fillna(0.0)
# Tail protection
if self.tail_protection:
mkt_ret = ret.mean(axis=1)
mkt_eq = (1 + mkt_ret).cumprod()
mkt_dd = mkt_eq / mkt_eq.cummax() - 1
in_tail = mkt_dd < -0.15
scale = pd.Series(1.0, index=data.index)
scale[in_tail] = 0.5
signals = signals.mul(scale, axis=0)
# Monthly rebalance
warmup = 252
rebal_mask = pd.Series(False, index=data.index)
@@ -182,6 +175,17 @@ class EnhancedFactorComboStrategy(Strategy):
signals = signals.ffill().fillna(0.0)
signals.iloc[:warmup] = 0.0
# Tail protection (rebal-gated to avoid daily-flip turnover)
if self.tail_protection:
mkt_ret = ret.mean(axis=1, skipna=True)
mkt_eq = (1 + mkt_ret.fillna(0.0)).cumprod()
mkt_dd = mkt_eq / mkt_eq.cummax() - 1
in_tail = mkt_dd < -0.15
scale_raw = pd.Series(1.0, index=data.index)
scale_raw[in_tail] = 0.5
scale = scale_raw.where(rebal_mask, np.nan).ffill().fillna(1.0)
signals = signals.mul(scale, axis=0)
return signals.shift(1).fillna(0.0)
@@ -230,31 +234,41 @@ class RiskManagedEnsembleStrategy(Strategy):
# Step 1: Get raw signals from the ensemble (already shifted by 1)
raw = self.ensemble.generate_signals(data)
# Step 2: Compute MARKET returns (equal-weight of all stocks)
daily_rets = data.pct_change().fillna(0.0)
mkt_rets = daily_rets.mean(axis=1)
# Step 2: Compute MARKET returns over valid (non-masked) columns.
# `daily_rets` keeps NaN for PIT-masked tickers so they don't dilute
# the cross-sectional mean to ~0.
daily_rets = data.pct_change()
mkt_rets = daily_rets.mean(axis=1, skipna=True)
# Step 3: Market drawdown dampener
mkt_eq = (1 + mkt_rets).cumprod()
mkt_dd = mkt_eq / mkt_eq.cummax() - 1 # always ≤ 0
# Linear: at DD=0 → 1.0, at DD=-dd_denom → dd_floor
dd_scale = (1.0 + mkt_dd / self.dd_denom).clip(lower=self.dd_floor, upper=1.0)
dd_scale_lagged = dd_scale.shift(1).fillna(1.0) # PIT
mkt_eq = (1 + mkt_rets.fillna(0.0)).cumprod()
mkt_dd = mkt_eq / mkt_eq.cummax() - 1
dd_scale_raw = (1.0 + mkt_dd / self.dd_denom).clip(
lower=self.dd_floor, upper=1.0,
)
# Step 4: Vol spike guard (uses portfolio's own vol for specificity)
# Step 4: Vol spike guard from portfolio returns (NaN-aware sum)
if self.vol_spike_guard:
port_rets = (raw * daily_rets).sum(axis=1)
port_rets = (raw * daily_rets).sum(axis=1, min_count=1).fillna(0.0)
short_vol = port_rets.rolling(self.vol_spike_window, min_periods=5).std() * np.sqrt(252)
vol_90th = short_vol.rolling(self.vol_spike_lookback, min_periods=126).quantile(0.90)
in_spike = short_vol > vol_90th
vol_scale = pd.Series(1.0, index=data.index)
vol_scale[in_spike] = self.vol_spike_floor
vol_scale_lagged = vol_scale.shift(1).fillna(1.0) # PIT
vol_scale_raw = pd.Series(1.0, index=data.index)
vol_scale_raw[in_spike] = self.vol_spike_floor
else:
vol_scale_lagged = 1.0
vol_scale_raw = pd.Series(1.0, index=data.index)
# Step 5: Combined scaling, sampled at the inner ensemble's rebal
# cadence so we don't trade in/out daily (which would incur huge
# fixed-fee costs).
combined = (dd_scale_raw * vol_scale_raw).shift(1).fillna(1.0)
rebal_freq = getattr(self.ensemble, "rebal_freq", 21)
warmup = 252
rebal_mask = pd.Series(False, index=data.index)
rebal_indices = list(range(warmup, len(data), rebal_freq))
rebal_mask.iloc[rebal_indices] = True
final_scale = combined.where(rebal_mask, np.nan).ffill().fillna(1.0)
# Step 5: Combined scaling
final_scale = dd_scale_lagged * vol_scale_lagged
return raw.mul(final_scale, axis=0)
@@ -342,26 +356,31 @@ class SharpeBoostedEnsembleStrategy(Strategy):
signals = signals.shift(1).fillna(0.0) # PIT: 1-day execution lag
# === Asymmetric vol scaling ===
# Only reduce exposure when vol is high AND returns are negative
# High vol + positive returns = riding a trend, don't cut
daily_rets = data.pct_change().fillna(0.0)
port_rets = (signals * daily_rets).sum(axis=1)
# NB: scales are RE-EVALUATED only on rebalance days. Daily flips of
# asym/dd scales would force half-in/half-out trades each session,
# burning the account through fixed per-trade fees ($2 US / $5 CN).
# Use cross-sectional mean of non-masked returns so PIT-masked
# NaN→0 fills don't dilute the market signal.
daily_rets = data.pct_change()
port_rets = (signals * daily_rets).sum(axis=1, min_count=1).fillna(0.0)
short_vol = port_rets.rolling(20, min_periods=10).std() * np.sqrt(252)
vol_median = short_vol.rolling(252, min_periods=126).median()
recent_ret = port_rets.rolling(20, min_periods=10).sum()
high_vol_neg = (short_vol > vol_median * 1.5) & (recent_ret < 0)
asym_scale = pd.Series(1.0, index=data.index)
asym_scale[high_vol_neg] = self.asym_vol_floor
signals = signals.mul(asym_scale.shift(1).fillna(1.0), axis=0) # PIT
asym_scale_raw = pd.Series(1.0, index=data.index)
asym_scale_raw[high_vol_neg] = self.asym_vol_floor
# === Light market-DD dampener ===
# Uses market (not strategy) drawdown to avoid negative feedback loop
mkt_rets = daily_rets.mean(axis=1)
mkt_eq = (1 + mkt_rets).cumprod()
# === Market-DD dampener (rebal-gated, NaN-aware market mean) ===
mkt_rets = daily_rets.mean(axis=1, skipna=True)
mkt_eq = (1 + mkt_rets.fillna(0.0)).cumprod()
mkt_dd = mkt_eq / mkt_eq.cummax() - 1
dd_scale = (1.0 + mkt_dd / self.dd_denom).clip(
dd_scale_raw = (1.0 + mkt_dd / self.dd_denom).clip(
lower=self.dd_floor, upper=1.0
)
signals = signals.mul(dd_scale.shift(1).fillna(1.0), axis=0) # PIT
# Sample scales at rebal points only, then step-hold between rebals.
combined = (asym_scale_raw * dd_scale_raw).shift(1).fillna(1.0)
rebal_scale = combined.where(rebal_mask, np.nan).ffill().fillna(1.0)
signals = signals.mul(rebal_scale, axis=0)
return signals