Walk-forward validation — model card

v1.0 · Tier 2Live Out-of-sample factor-weight evaluation.

Purpose

Validates that factor weights generalize beyond the training window. Splits historical bars into a sequence of rolling (in-sample → out-of-sample) windows. Fits weights on the IS, evaluates on the OOS. The OOS performance — not the IS fit — is what gets reported. No look-ahead by construction.

Inputs

Bars (≥ 252 days for daily, ≥ 1000 minutes for intraday)
Factor decomposition per ticker per timestamp
Window config: IS-length (default 90 days), OOS-length (default 30 days), step (30 days)

Outputs

{
  windows: [
    { isStart: '2024-01-01', isEnd: '2024-04-01', oosStart: '2024-04-01', oosEnd: '2024-05-01',
      isSharpe: 1.8, oosSharpe: 1.2, isHitRate: 0.62, oosHitRate: 0.55 },
    …
  ],
  aggregate: {
    avgOosSharpe: 1.1,
    oosToIsRatio: 0.66,        // closer to 1.0 = better generalization
    overfittingFlag: 'moderate'  // ratio < 0.5 → 'severe'
  }
}

Method

Anchored walk-forward (the IS window grows; OOS slides). Weight-fit via OLS on z-scored factors against forward returns; OOS evaluated by computing predicted return rank, top-decile minus bottom-decile (long-short proxy), and reporting Sharpe + hit rate.

Limitations

What this is NOT:

Not transaction-cost-aware (use Costs separately)
Not regime-conditional (factor weights regress globally — LS6 layer-conditional engine handles regime in production)
Not survivorship-bias-corrected — uses today's universe; doesn't include delisted tickers

Source

engine/walkforward.js

Reviewed: 2026-04-27 · Next: 2026-07-27 · Per Bible §15.61