Hy7 Panic Size WF compound +115.5%
Multi-factor IC-weighted composite scoring · Walk-forward OOS 2024-07 ~ 2026-08 · 25 OOS windows
Walk-forward OOS equity only; training windows excluded. Compare excess and drawdown vs BTC buy-and-hold; aligned start aids path comparison, not a live promise.
Peak-relative drawdown %; no extra factor or parameter leakage.
Walk-forward (rolling) is the industry gold standard for robustness. History is split into non-overlapping train→test windows: fit only on train, trade and record returns on the following out-of-sample (OOS) segment, then roll the whole window forward.
Discipline: every metric is OOS-only — training never enters evaluation; features are causally shifted bar-by-bar with an embargo between train/test to block look-ahead and overfitting. This page plots 25 OOS windows compounded into true equity.
Backtests deduct exchange-realistic costs trade by trade: taker fees + slippage + funding (overnight).
| One-way taker fee | 4.0 bps (0.040%) |
|---|---|
| One-way slippage assumption | 0.20 bps (0.0020%) |
| Funding treatment | Uses historical funding-rate series (×1) |
| When costs are applied | Per-fill cost deduction; all returns are after fees |
All shown returns (compound / annualized / Sharpe / drawdown) are after fees, not gross. Higher turnover and shorter holds mean larger cost drag, already in the curve.
Past backtests do not guarantee future results. Crypto is highly volatile; live performance can diverge materially due to liquidity, slippage, extremes, and regime/rule changes.
This page is a sanitized research view (factors, parameters, and precise internals hidden). For research only — not investment advice. Judge risks independently.