The calibration window opened 16 May 2026 — first measured factor weights persisted 17 May 2026 on 1d and 7d horizons (3 regimes). The 30-day walk-forward horizon needs several non-overlapping 30-day periods before it can produce a stable Sharpe; those are still accumulating, so that number is not published yet. Meanwhile, here's a proof the math is honest: we make up a fake universe where we KNOW the answer ahead of time, then run cross-validation on it. If the math finds the fake signal at the right strength, you can trust it'll find a real one too.
The calibration window opened 16 May 2026 and the first calibrated factor weights landed in apex_factor_weights on 17 May 2026 (1d and 7d horizons across the 'all', 'risk_on', and 'transition' regimes — measured in-sample IC 0.055-0.10, sample 1091). The 30-day walk-forward cross-validation needs enough non-overlapping 30-day windows to produce a stable Sharpe; those are still maturing (its out-of-sample IC currently reads the 'no OOS windows yet' sentinel), so the number is not published yet. Until then this page demonstrates the math is provably correct: a known alpha planted into a synthetic universe, verified recovered at the right strength via the same IC machinery.
Three configurations: a strong-edge case (planted ρ ≈ 0.40), a modest-edge case (ρ ≈ 0.15 — typical hedge-fund factor), and a pure-noise control (ρ = 0). The recovered mean OOS IC should bracket each planted value, with the noise control cleanly near zero.
Three rounds: strong fake signal, modest fake signal, and pure noise (no signal). The math should find each fake at roughly the strength we planted, and find nothing in the noise. If it does, your real-money calls are running on the same correct math.
Three independent factors, each carrying a modest planted edge. The composite blend (equal-weight) harvests the union of edges — you should see composite IC strictly above the average of the three single-factor ICs and at least competitive with the best individual factor.
Momentum IC
0.201
Sharpe 1.3
Value IC
0.181
Sharpe 0.9
Sector IC
0.159
Sharpe 0.9
Composite IC
0.313
Sharpe 1.8 · Sortino 9.5
Composite cumret
81.4%
long-short compounded
Max drawdown
-0.8%
worst peak-to-trough
Lift: composite vs avg-single = 73% — the diversification benefit is real, not an arithmetic artefact: independent edges combine sub-additively in variance but additively in signal (Asness-Frazzini-Pedersen 2014).
Pairwise Spearman ρ — orthogonality of edges
momentum
value
sector
momentum
1.00
-0.01
-0.05
value
-0.01
1.00
0.01
sector
-0.05
0.01
1.00
Mean |ρ| off-diagonal = 0.024 — closer to 0 means more independent edges, more diversification benefit. < 0.20 = excellent, > 0.50 = significant redundancy.
What this proves
The walk-forward runner recovers the planted Spearman correlation within statistical sampling error across all three regimes — strong, modest, and zero. That demonstrates (a) no look-ahead bias, (b) correct rank-correlation math, (c) proper window aggregation. The same code path is now producing calibrated IC numbers on the live universe — see today's 1d / 7d horizon results in apex_factor_weights; /backtest will publish a 30-day walk-forward Sharpe once enough non-overlapping 30-day windows have matured — not yet the case.
Source: src/lib/data/historical-reconstructor.ts · Tests: scripts/test-historical-reconstructor.mjs(70 assertions, all green).