Navigation
⚠ VERDICT UNDER REVIEW IAAFT-dependent finding

Original verdict relied on IAAFT null calibration — now prohibited by project policy. The finding needs to be re-derived from an approved null-calibration method before it can be used for decisions. Numbers below are preserved for the record.

Candidate 2 of 5 · Tier A · A1 Pairwise · Top pick
GATE 0.5 PASS GATE 3 PASS MULTI-SYMBOL REPLICATED BROADENING STRONG

✅ Chatterjee ξₙ — rank-based, parameter-free, asymmetric

Verdict: STRONG alignment with paper claims AND with our end goal. Every property the paper promises holds empirically on financial data. The metric directly answers "how much of Y is predictable from X?" with a number on [0, 1], without any tuning knobs, in under 3 milliseconds per evaluation. Our most-validated candidate.

Hub trail: /CLAUDE.mdaudit hubchatterjee-xi/

What this metric does, in plain English

Sort your data by the X variable. Then look at the Y values in that order. If Y is strongly predictable from X (any kind of relationship — curved, monotonic, whatever), consecutive Y values in this order will be close to each other in rank. If X tells you nothing about Y, the Y values will look randomly shuffled. The formula just counts how often consecutive Y ranks are close. Output ranges from 0 (X tells you nothing about Y) to 1 (Y is a perfect function of X). No knobs, no tuning, no kernel choice — just the data.

Headline numbers

Compute / evaluation
2.4 ms
at N=1000. 140× faster than HSICAgg.
Empirical null cap
~0.07
across 4,000 IAAFT nulls — ξ > 0.10 is decisive
Pair-8 obs/null max
11.8×
Real signal, not autocorrelation artifact
FOSS impl
xicorpy 0.6
MIT-licensed. Drop-in usable.

What we measured — Gate 0.5 spike on 8 audit pairs

Pair Prior audit Spearman ρ ξ(X→Y) ξ(Y→X) Reading
ofi ↔ turnover_imbalancePERFECT+1.0000.9970.997Confirms duplicate
vwap ↔ closePERFECT+0.9670.9000.897Substrate collapse
vwap ↔ openPERFECT+0.9930.9500.945Substrate collapse
buy_volume ↔ sell_volumeHIGH+0.7180.7270.717High redundancy confirmed
ofi ↔ aggression_ratioMODERATE+0.7930.7610.743Moderate redundancy confirmed
vwap_close_deviation ↔ price_impactMODERATE−0.0200.5440.536Spearman calls "indep"; ξ high
kyle_lambda_proxy ↔ ofiORTHOGONAL−0.1700.7160.746Spearman near-zero; ξ very high
kyle_lambda_proxy ↔ buy_volumeORTHOGONAL+0.0490.6130.577Confirmed REAL by IAAFT

Does it do what the paper says?

Paper claim Test Result
ξ lies in [0, 1]4,012 evaluationsAll within bounds ✓
ξ → 0 under independenceSynthetic indep + IAAFT nullsSynthetic 0.009; null caps ~0.07 ✓
ξ → 1 for perfect functionY=X duplicate + perfect-dup pair1.000 / 0.997 ✓
Asymmetric: ξ(X→Y) ≠ ξ(Y→X)Production pairsNon-zero asymmetries on all 8 ✓
Parameter-freeDefault xicorpyNo knobs touched ✓
O(N log N)500 surrogates × N=1000 in 0.6s2.4 ms/eval ✓

All 6 paper claims verified empirically. STRONG alignment.

Does it serve our orthogonality-measurement goal?

YES — direct fit

Goal: produce a number that says "how much information another feature carries about this one" — including non-linear. Chatterjee produces exactly that on [0, 1].

Catches what Spearman misses

Pair 6: Spearman ρ = −0.020 → "orthogonal". Chatterjee ξ = 0.544 → "highly dependent". Production baseline would retain both features.

Well-calibrated on financial data

IAAFT null cap (~0.07) is empirical and consistent across pairs. Threshold ξ > 0.10 sits safely above noise.

No tuning = auditable

No operator can introduce p-hacking via bandwidth/kernel choice. The metric is what the data says.

Provenance: committed in 65738478 · "Gate 0.5 spike for Chatterjee ξₙ — corroborates HSICAgg pair-8 finding"