12H return across fold schedules
Past returns + prior funding · 10 folds is frozen; 15/20 are sensitivity checks
Source: genuine X-Perp external-feature ablation report
OKX EEA BTC X-Perp · research snapshot
Crypto Research Control Room
candidate · target
on genuine X-Perp. The policy issued decisions on of rows. Funding, mark price, futures fees, and slippage are included at 1x notional.
May-forward net ROI
Genuine futures evidence
May-forward BTC X-Perp evaluation; every figure is research-only and selection-biased.
12H feature ablation
Net ROI under the frozen 10-fold schedule
Source: genuine X-Perp external-feature ablation report
2H feature ablation
The same variants do not generalize cleanly to the faster cycle
Source: genuine X-Perp external-feature ablation report
X-Perp ablation explorer
All 26 timeframe and feature-variant cells.
| Cycle | Variant | Features | All-row accuracy | Exposed accuracy | Net ROI | Sharpe | Max DD | Calmar | Round trips | + folds |
|---|
Frozen forward holdout
Registered candidates, one permitted opening, protected outcomes not loaded.
| Candidate ID | Variant | Cycle | Target | Feature groups | Prior funding | Training cutoff |
|---|
Two-year spot-proxy archive
Historical pretraining research only; these figures are not futures ROI.
Historical 12H return by outer fold
Spot-price proxy · 10% target · five outer test periods
Source: active four-model spot-proxy coverage report
Best historical Calmar by timeframe
Best canonical confidence policy in each cycle
Source: active four-model spot-proxy coverage report
Historical 12H return by confidence
100% trades every decision; 10% requests the most confident set
Source: active four-model spot-proxy coverage report
| Rank | Cycle | Target | Actual rate | Net ROI | Annualized | Sharpe | Max DD | Calmar | Round trips | + folds |
|---|
Experiment decisions
What is active, rejected, or still exploratory.
Historical spot-proxy fold audit
The earlier 12H/10% result, preserved for traceability.
| Fold | Signal period | Actual rate | Threshold | Net ROI | Sharpe | Max DD | Calmar | Round trips | Winning blend |
|---|
Evaluation integrity
Futures accounting first; spot proxy is shown separately to prevent accidental comparison.
| Instrument | Evaluation window | Rows | Folds | Fee bps / side | Slippage bps / side | Funding events | Funding coverage | Mark price | Leverage | Status |
|---|
Historical spot-price proxy
Funding and futures basis were excluded from this older evaluation.
| Instrument | Evaluation window | Grain | Rows | Missing | Folds | Fee / side | Slippage / side | Funding | Leverage |
|---|