Generated 2026-10-05 | state-augmented vol targeting (walk-forward from 2008-01-01, holdout >= 2023-01-01, cost 5bps on |dpos|) and equity-curve feedback.
eqmom -- vol targeting (walk-forward)
| variant | ret | sharpe | maxDD | downvol | CVaR5 | calmar |
|---|---|---|---|---|---|---|
| BASE | +759.5% | 0.76 | -45.4% | 17.9% | -2.20% | 16.72 |
| VT_naive | +731.2% | 0.95 | -22.2% | 12.8% | -1.68% | 32.92 |
| VT_state | +373.5% | 0.70 | -33.6% | 15.4% | -1.58% | 11.13 |
eqmom -- vol targeting (HOLDOUT)
| variant | ret | sharpe | maxDD | downvol | CVaR5 | calmar |
|---|---|---|---|---|---|---|
| BASE | +236.6% | 1.49 | -26.0% | 18.3% | -2.48% | 9.08 |
| VT_naive | +156.6% | 1.49 | -19.4% | 13.6% | -1.89% | 8.07 |
| VT_state | +73.7% | 1.20 | -18.3% | 11.2% | -1.17% | 4.02 |
state-VT-vs-naive-VT block-placebo p = 0.950
flagship -- vol targeting (walk-forward)
| variant | ret | sharpe | maxDD | downvol | CVaR5 | calmar |
|---|---|---|---|---|---|---|
| BASE | +933.8% | 0.88 | -30.8% | 15.4% | -1.92% | 30.32 |
| VT_naive | +223.1% | 0.73 | -17.3% | 9.2% | -1.23% | 12.88 |
| VT_state | +210.9% | 0.73 | -16.8% | 8.8% | -1.17% | 12.58 |
flagship -- vol targeting (HOLDOUT)
| variant | ret | sharpe | maxDD | downvol | CVaR5 | calmar |
|---|---|---|---|---|---|---|
| BASE | +193.8% | 1.62 | -18.3% | 13.1% | -1.57% | 10.58 |
| VT_naive | +121.9% | 1.55 | -14.5% | 10.1% | -1.44% | 8.42 |
| VT_state | +115.7% | 1.54 | -14.2% | 9.7% | -1.38% | 8.15 |
state-VT-vs-naive-VT block-placebo p = 0.550
Equity-curve feedback (spec 11)
Lagged own-equity-curve stats vs own forward 21d return/vol/drawdown; partial controls the market state index. BH q across all tests.
| sleeve | feature | target | IC | IC OOS | partial vs state | q | status |
|---|---|---|---|---|---|---|---|
| eqmom | dvol | vol21 | +0.522 | +0.412 | +0.480 | 2.18e-19 | INTERESTING |
| eqmom | dd | vol21 | -0.224 | +0.273 | -0.138 | 0.00147 | WEAK |
| flagship | roll_sharpe | fwd21 | -0.039 | -0.265 | -0.041 | 0.907 | FAILED |
| eqmom | ac1 | fwd21 | -0.032 | -0.257 | -0.032 | 0.907 | FAILED |
| flagship | dd | fwd21 | -0.225 | -0.247 | -0.232 | 7.25e-05 | INTERESTING |
| flagship | hit63 | fwd21 | -0.018 | -0.244 | -0.017 | 0.908 | FAILED |
| eqmom | roll_sharpe | vol21 | +0.035 | +0.213 | +0.117 | 0.907 | FAILED |
| flagship | roll_sharpe | vol21 | +0.066 | -0.212 | +0.176 | 0.767 | FAILED |
| flagship | ret40 | vol21 | -0.002 | -0.189 | +0.073 | 0.981 | FAILED |
| eqmom | ret40 | vol21 | +0.126 | +0.186 | +0.168 | 0.152 | FAILED |
| flagship | dvol | vol21 | +0.622 | +0.184 | +0.590 | 9.85e-29 | INTERESTING |
| flagship | ac1 | fwd21 | +0.061 | -0.182 | +0.063 | 0.706 | FAILED |
| flagship | ret40 | fwd21 | -0.032 | -0.171 | -0.031 | 0.907 | FAILED |
| flagship | dvol | fwd21 | +0.225 | +0.166 | +0.223 | 7.25e-05 | INTERESTING |
| flagship | dd | vol21 | -0.324 | -0.145 | -0.258 | 1.54e-06 | INTERESTING |
| flagship | ac1 | vol21 | -0.052 | -0.134 | -0.033 | 0.907 | FAILED |
| flagship | ret20 | vol21 | +0.017 | -0.129 | +0.064 | 0.946 | FAILED |
| eqmom | hit63 | vol21 | +0.180 | +0.119 | +0.162 | 0.0177 | INTERESTING |
| eqmom | roll_sharpe | fwd21 | -0.026 | -0.109 | -0.027 | 0.907 | FAILED |
| eqmom | ret10 | vol21 | -0.011 | -0.108 | -0.017 | 0.946 | FAILED |
| flagship | ret20 | fwd21 | -0.056 | -0.105 | -0.057 | 0.765 | FAILED |
| eqmom | ret5 | vol21 | -0.041 | -0.103 | -0.051 | 0.907 | FAILED |
| eqmom | hit63 | fwd21 | +0.002 | +0.099 | -0.017 | 0.981 | FAILED |
| eqmom | dd | fwd21 | -0.123 | -0.073 | -0.110 | 0.0763 | WEAK |
| flagship | hit63 | vol21 | +0.086 | -0.052 | +0.150 | 0.581 | FAILED |
| flagship | ret10 | fwd21 | -0.032 | -0.051 | -0.033 | 0.907 | FAILED |
| flagship | ret5 | fwd21 | -0.022 | -0.045 | -0.024 | 0.907 | FAILED |
| eqmom | dvol | fwd21 | +0.127 | -0.040 | +0.119 | 0.0667 | WEAK |
| flagship | ret5 | vol21 | +0.014 | -0.040 | +0.020 | 0.946 | FAILED |
| eqmom | ret10 | fwd21 | -0.020 | +0.030 | -0.028 | 0.907 | FAILED |
Verdict
- eqmom walk-forward Sharpe: BASE 0.76, naive-VT 0.95, state-VT 0.70; placebo p=0.95.
- flagship walk-forward Sharpe: BASE 0.88, naive-VT 0.73, state-VT 0.73; placebo p=0.55.
- Equity-curve: 6 feature-target pairs pass the robust gate (q<0.05, sign-stable, |OOS IC|>=0.05) out of 36.
State-augmented vol targeting: REJECTED. The state index does not improve a vol-target overlay over a plain EWMA vol target (block-placebo p=0.95 eqmom, 0.55 flagship; state-VT Sharpe is actually lower for eqmom). The incremental forward-vol information from Phase 1/2 does not convert into a better overlay once a simple trailing vol estimate is in the model.
Byproduct (not an equity-state result): plain EWMA vol targeting on eqmom materially improves walk-forward risk (Sharpe 0.76 -> 0.95, maxDD -45% -> -22%), but the holdout Sharpe is flat (1.49) with a large return give-back - worth separate investigation, independent of market state.
Equity-curve feedback: REJECTED for returns. No own-curve feature robustly times forward returns incrementally over the market-state index; the surviving pairs are risk-clustering artefacts (downside vol -> future vol) or drawdown persistence (flagship deeper drawdown -> lower forward return) - diagnostics, not tradeable timing.
Recommendation: DO NOT PURSUE any equity-state overlay on the existing sleeves. Equity-state variables remain useful only as an observe-only fragility monitor.