Look-ahead-free Spearman IC at every monthly rebalance on a 371-ticker US universe (2022-01-19 → 2026-03-23, monthly forward window). Every score uses only data that existed on the rebalance date — price bars from Yahoo, fundamentals from EDGAR XBRL filings filtered by their filed timestamp. |IC| above 0.03 is the band professional hedge funds operate in; the headline is the strongest composite produced by the engine. Individual factor names and exact composition are proprietary — the IC numbers are real and reproducible from the published universe.
| Factor | n | cross-sec IC | cross-sec Sharpe | sec-neutral IC | sec-neutral Sharpe | hit rate |
|---|---|---|---|---|---|---|
| Composite A | 37 | +0.0516 | +1.03 | +0.0466 | +1.42 | 67.6% |
| Composite B | 37 | +0.0345 | +0.87 | +0.0347 | +1.26 | 73.0% |
| Composite C | 37 | +0.0257 |
Snapshot generated 2026-04-29. Cross-sectional IC ranks every ticker against every other; sector-neutral IC demeans each factor within its sector first, isolating within-sector alpha from sector rotation. Citations: George & Hwang 2004 (momentum), Lehmann 1990 (reversal), Ang-Hodrick-Xing-Zhang 2006 (low-vol), Amihud 2002 (illiquidity), Cooper-Gulen-Schill 2008 (asset growth). Production engine combines these with insider, options, NLP, pharma, and macro factors not covered in this price-only XBRL snapshot.
How much the engine beats a coin flip after adjusting for risk. The single most-cited number in investing — higher means returns are more consistent relative to volatility.
<0.5 = weak · 0.5–1.0 = average · 1.0–2.0 = good (where professional hedge funds live) · >2.0 = stellar (Renaissance, Citadel territory).
Like Sharpe, but only penalises losses — upside volatility no longer counts as risk. Sortino is almost always higher than Sharpe for a decent strategy.
<0.7 = weak · 0.7–1.5 = average · 1.5–3.0 = good · >3.0 = stellar. A Sortino far above Sharpe means the strategy has positive skew (rare big wins).
Annual return divided by the worst drawdown. A brutal metric — a +30%/year strategy with a −50% drawdown earns Calmar 0.6 (not worth it). Institutional allocators look at this before Sharpe.
<0.5 = poor · 0.5–1.0 = average · 1.0–3.0 = good · >3.0 = trader-of-the-year tier.
Percentage of signals that came true. For BUY signals: share of cases where the price rose over the next 30 days. For SELL: share where it fell.
Each individual factor on its own — how well it predicts 30-day forward returns by itself. Useful for spotting which factor actually carries the signal in this universe.
The snapshot Sharpe above aggregates across the entire signal history. Walk-forwardbelow is the honest simulation: the engine rebalances a top-scoring, balanced-risk portfolio on a fixed monthly cadence using only data available on each rebalance date — no look-ahead bias. This is the number a quant fund cites when raising capital from limited partners.
Sharpe computed in honest simulation — the engine rebalances on a fixed cadence using only data available on each rebalance date. This is the number a quant fund cites when raising capital from limited partners.
Same scale as the snapshot Sharpe, but walk-forward always reads slightly lower. ≥ 0.7 walk-forward Sharpe is already an investable strategy. ≥ 1.0 is exceptional.
Annual compound return the strategy earned in walk-forward simulation. Realistic — no look-ahead bias.
S&P 500 long-term ≈ +10% / year. > +15% is meaningful alpha. > +25% is institutional-tier.
Not an investable portfolio. Each trading day the bullish names form an equal-weight basket, and that basket’s realised 30-day return is chained onto the day before. Those 30-day holds overlap almost completely, so the same month of performance is counted many times over and the total grows far faster than a funded account could. Read it as signal quality aggregated over time, not as money made.
Reconstructs the engine on 1000 tickers across the 2021-05-14 → 2026-10-02 window (93 rebalance dates, 30-day forward returns). Different from the live track-record above — that reads our daily signal_history since 2026-04-09; this one reconstructs what the engine WOULD have published on every rebalance date across that window using only data available at the as-of moment. No look-ahead.
| +0.54 |
| +0.0327 |
| +1.14 |
| 62.2% |
| Factor I | 37 | +0.0390 | +0.89 | +0.0276 | +0.71 | 64.9% |
| Factor II | 37 | −0.0238 | −0.75 | +0.0059 | +0.28 | 45.9% |
| Factor III | 48 | −0.0249 | −0.56 | −0.0096 | −0.30 | 39.6% |
| Factor IV | 46 | −0.0209 | −0.50 | +0.0031 | +0.12 | 34.8% |
| Factor V | 37 | +0.0100 | +0.27 | +0.0048 | +0.27 | 56.8% |
| Factor VI | 37 | −0.0079 | −0.21 | +0.0069 | +0.27 | 45.9% |
| Factor VII | 46 | +0.0157 | +0.21 | +0.0206 | +0.38 | 52.2% |
| Factor VIII | 47 | +0.0013 | +0.02 | +0.0002 | +0.01 | 57.4% |
Correlation between what the engine predicted and what actually happened. 0 = guessing like a coin flip; higher means the score has stronger measured rank-correlation with future returns in this sample.
0.00 = random · 0.03–0.05 = professional hedge fund · 0.05–0.10 = top-tier · >0.10 = phenomenal.
The worst peak-to-trough fall of that same basket series — the deepest stretch where the bullish set gave ground back. It measures the signal, not a portfolio: the series is not investable.
0 to −10% = very calm · −10 to −20% = normal for equity · −20 to −40% = painful · worse than −40% = brutal.
Each trading day the bullish names form an equal-weight basket; that basket's realised 30-day return is chained onto the day before. The 30-day holds overlap almost completely, so the same month of performance is counted many times over.
Not comparable to buy-and-hold: no fixed capital, no position limit, no costs. Read it as signal quality aggregated over time, not as money made.
50% = coin flip · 55–60% = real edge · 60–70% = stellar · >70% = suspicious (likely overfit).
IC > 0.05 on a single factor = real signal · IC < 0.02 = noise · negative IC = factor works in reverse, invert the sign.
Deepest peak-to-trough loss on the simulated equity curve. If you had held the strategy from day one, this is how far you would have been underwater at the worst moment.
0 to −15% = excellent · −15 to −25% = ok for an equity strategy · −25 to −40% = painful · worse = red flag.