โ€บNavigation
Orthogonality measurement instruments ยท priority list
PHASE 2 ยท POST-SELECTION ฮพ VALIDATED (SELECTION-PARTIAL) CODEC SKIPPED ยท O-INFO REWORKED 3 FEASIBLE NEXT CANDIDATES

๐ŸŽฏ Priority list โ€” orthogonality measurement candidates

Updated 2026-06 after the CODEC/FOCI Selection sweep. Chatterjee ฮพ (A1 pairwise) is validated as the only certified sieve (S1/S2/S3/S5 PASS, Selection-partial โ€” value-add over Pearson/Spearman pending a non-circular S4). CODEC (A2) is skipped (inert on the panel + no faithful null) and O-information (A3) is reworked to a bias-centered permutation null. The three feasible candidates below close the conditional/synergy gap with cheap, faithful nulls. Already-spiked metrics (ฮพโ‚™, Aggregated HSIC, IAAFT) live in the spike showcase.

SELECTION-SWEEP OUTCOME ๐Ÿ”’ DEPLOYMENT BLOCKED ON TERRY

Where the cascade landed

A1 ยท Chatterjee ฮพ

โœ… VALIDATED โ€” Selection-partial

S1 (mutation kill 0.96) / S2-ฮพ / S3 (Jaccard 0.88) / S5-ฮพ PASS. The only certified sieve. Value-add over Pearson/Spearman unproven โ€” S4 deferred (labels are Pearson-derived โ†’ circular).

A2 ยท CODEC / FOCI

๐Ÿšซ SKIPPED

Nearly inert on the crypto panel (0.16 flags/cell, T(canary)โ‰ˆ0) and no faithful FOSS null. Replaced by GCM / RCoT below.

A3 ยท O-information

โ™ป๏ธ REWORKED โ€” watch only

Naive |ฮฉ|โ‰ฅ0.5 flag dropped (manufactures false synergy at small n). Kept as a cheap bias-centered permutation-null watch.

FEASIBLE NEXT ยท 3 CANDIDATES โ‰ค2 CORES ยท NUM_THREADS=1 FAITHFUL + CHEAP NULL

Close the conditional / synergy gap โ€” 3 feasible candidates

Filter: faithful + cheap null (the exact thing CODEC lacked), no O(nยณ) at grid scale (the runaway risk), FOSS, parameter-light. All three still cap at Selection-partial until a Pearson-independent S4 label exists, and all face the ZERO_LOCAL_COMPUTE adoption barrier โ€” these evaluate the question cheaply; they are not deployment candidates.

#1 ยท A2 conditional

๐Ÿงญ GCM (Generalised Covariance Measure)

Regress X~Z and Y~Z, test residual cross-covariance. Analytic N(0,1) null, model-agnostic. The principled CODEC replacement.

FOSS: GeneralisedCovarianceMeasure (CRAN) / pywhy-dodiscover
Cost: O(N) โ€” two regressions + scalar test
Closes: conditional redundancy (faithful cheap null)
#2 ยท A2 conditional

๐Ÿงญ RCoT (randomized conditional)

Linear-in-n KCI approximation (random Fourier features) โ€” fixes the O(nยณ) wall that caused the core runaway. Analytic Lindsay-Pilla-Basak null.

FOSS: causal-learn (RCIT, preliminary) / ericstrobl/RCIT (R)
Cost: O(N) โ€” linear-time
Closes: conditional redundancy at scale (cross-check on GCM)
#3 ยท A3 synergy

๐Ÿ”บ Bias-centered O-info permutation null

Keeps the synergy question: drop the uncalibrated naive flag, use the permutation null. Already calibrated here (Type-I 0.058, power 1.0, ~26s / 1 core).

FOSS: hoi (BSD-3)
Cost: closed-form ฮฉ + ~199 shuffles โ€” cheap
Closes: higher-order synergy (cheap watch)

Reference oracle, not a primary: KCI (causal-learn, analytic Gamma null) is O(nยณ) โ€” the instrument that caused the 15โ€“23-core runaway. Use it only subsampled (n โ‰ค 1000) to validate GCM / RCoT, never as the grid workhorse. Already-catalogued cheap alternates: Copula-Entropy CI test (copent.ci) and CMIknn (tigramite).

WORKFLOW PER-METRIC

Per-candidate testing workflow

Every metric in the priority list rides the same three-step rail. Gate 0.5 establishes the FOSS-implementation smoke test. Gate 3 (IAAFT) is currently prohibited โ€” see the IAAFT page for the prohibition notice and cascade impact. A replacement null-calibration approach is needed before any new spike can complete its evidence chain. Multi-symbol confirms the finding isn't BTC-specific.

Step 1 ยท Gate 0.5

FOSS spike (โ‰ค 1 hr)

Wire FOSS package against BTCUSDT @ 100 dbps with the 18-feature pair panel. Per-pair score + permutation null. Verdict: signal vs. noise.

Step 2 ยท Gate 3 ยท ๐Ÿšซ PROHIBITED

IAAFT calibration (blocked)

Phase-randomised surrogate methodology not approved by project policy. See prohibition notice. A replacement null-calibration step is needed before any new spike can complete its evidence chain.

Step 3 ยท Multi-symbol

Robustness

Re-run on ETH + ADA. Verifies the dependency pattern isn't BTC-only. Note: prior Chatterjee ฮพโ‚™ 3/3 replication relied on the now-prohibited Gate 3, so that verdict is also under review.

Parallel-fleet pattern: open four tmux panes on bigblack, one per Batch-A metric. Each pane runs an independent Claude Code session against its own metric, sharing only the read-only BTCUSDT data. Results land in findings/dashboard/instruments/spikes/candidates/ as new spike pages โ€” same template as the existing six.

All 30 candidates, ranked

P1 โ†’ P4 ยท 4 tiers ยท 6 categories

Tier coloring reflects how directly the metric measures orthogonality. P1 = top picks across the four catalog gaps. P2 = strong secondary picks (Tier A direct measurement, FOSS-ready). P3 = useful breadth (Tier A, non-top-pick or needs port). P4 = Tier B adjacent (causal direction, regime stability) โ€” doesn't directly score orthogonality, but adds context.

# Tier Metric Category FOSS package Effort Gap / role
โ˜…P1GCM (Generalised Covariance Measure)A2GeneralisedCovarianceMeasure (CRAN) / pywhyEasy ยท O(N)Conditional ยท analytic N(0,1) null ยท CODEC replacement
โ˜…P1RCoT / RCIT (randomized conditional)A2causal-learn (prelim) / RCIT (R)Moderate ยท O(N)Conditional at scale ยท linear-time KCI approx ยท LPB null
1P1Distance Correlation (dCor)A1dcor (MIT)Trivial (~5 min)A1 cross-method check on Chatterjee
โ€”SKIPCODEC / FOCIA2xicorpy (MIT)โ€”SKIPPED โ€” inert (0.16 flags/cell) + no faithful null
โ˜…P1ยทwatchO-information + permutation nullA3hoi (BSD-3)Easy ยท cheapSynergy ยท bias-centered shuffle null (naive |ฮฉ|โ‰ฅ0.5 dropped)
4P1Marchenko-Pastur eigenvalue clippingA4pyRMT (BSD-style)Easy (~30 min)Matrix-level / panel-wide
5P2CMIknnA2tigramite (GPLv3)Easyk-NN conditional MI ยท PCMCI core
6P2HHG (Heller-Heller-Gorfine)A1hyppo (Apache-2)EasyDistance-rank consistent independence test
7P2MIC / MICe / TICA1minepy (GPL)EasyEquitability framework (Reshef 2011)
8P2HSIC (median heuristic, gamma-approx)A1hyppo (Apache-2)EasyNon-aggregated HSIC ยท simpler than HSICAgg
9P2Copula EntropyA1copent (MIT)EasyRank-uniformised MI ยท scale-invariant
10P2BROJA-2PIDA3BROJA_2PID + dit (BSD-3)ModerateConvex-opt PID ยท unique / redundant / synergistic
11P2Tracy-Widom edge testA4TracyWidom (PyPI)EasyPer-eigenvalue p-value ยท pairs with MP
12P2RIE (Rotationally Invariant Estimator)A4pyRMT (BSD-style)EasyOptimal nonlinear shrinkage of correlation matrices
13P2Distance MultivarianceA3R multivariance (port)ModeratedCov extended to โ‰ฅ 3 vars
14P3Hoeffding's DA1R independence (port)ModerateU-shape dependence ยท pre-Chatterjee classic
15P3Bergsma-Dassios ฯ„*A1R TauStar (port)ModerateConsistent independence test ยท extends Kendall
16P3Schweizer-Wolff ฯƒA1R copBasic (port)ModerateCopula Lยน total dependence
17P3Williams-Beer PID (Imin)A3dit (BSD-3)EasyOriginal PID ยท classical (overestimates redundancy)
18P3Ince's Iccs PIDA3dit (BSD-3)EasyPointwise common-surprisal PID
19P3Graphical Lasso (GLasso)A2sklearn (Apache-2)EasyL1 inverse-covariance ยท Gaussian CI graph
20P3Copula Entropy CI testA2copent.ci() (MIT)EasyRank-shuffled CI test via copula entropy
21P3Ledoit-Wolf NLS (2020)A4R nlshrink (port)ModeratePolynomial nonlinear shrinkage
22P3Sparse PCA / Kernel PCA / Robust PCA / NMF / FastICAA4sklearn (Apache-2)Easy eachFive matrix decompositions for non-Gaussian / nonlinear panels
23P4Reduced Transfer Entropy (Kirkley 2025)B1Author GitHub (port)ModerateClosed-form significance ยท no surrogate needed
24P4Effective Transfer Entropy (Dimpfl-Peter)B1R RTransferEntropy or IDTxl (GPLv3)ModerateFinance-standard bias-corrected TE
25P4PCMCI / PCMCI+B1tigramite (GPLv3)EasyMultivariate causal-discovery graph
26P4Sliced Wasserstein dependencyB2POT (MIT)ModerateOT-based ยท outlier-robust
27P4CRQA (Cross-Recurrence Quantification)B2PyRQA (Apache-2)EasyDistribution-free coupling
28P4Robust BOCPD (Altamirano 2023)B2Author GitHubModerateHeavy-tail-robust Bayesian Online Changepoint
29P4Granger / Kernel Granger / Neural GrangerB1statsmodels + Neural-GC (MIT)EasyClassical + nonlinear extensions ยท baseline causal
30P4CCM + cCCMB1pyEDM (BSD)ModerateTakens-embedding causal coupling ยท cCCM fixes leakage

Alternative batches

documented for the record

Batch A is recommended because it closes four distinct catalog gaps simultaneously. Batches B and C are documented here so the choice is traceable โ€” neither is currently active.

BATCH B"Tier A all in"

Two A1 metrics + A2 + A3

Heavier on A1 confirmation โ€” two pairwise metrics (dCor for cross-method, HHG for distance-rank consistency) plus CODEC/FOCI and O-information. Trades the A4 matrix-level gap for stronger pairwise confidence.

  • ยท Session 1 โ€” dCor (A1 #1 cross-check)
  • ยท Session 2 โ€” HHG (A1 alternate)
  • ยท Session 3 โ€” CODEC / FOCI (A2)
  • ยท Session 4 โ€” O-information (A3)
BATCH C"5-gap sweep"

Tier A + Tier B mix

Broadest scope: conditional, synergy, causal direction, and regime stability all in one batch. Trades direct-measurement weight for adjacent-instrument breadth.

  • ยท Session 1 โ€” CODEC / FOCI (conditional)
  • ยท Session 2 โ€” O-information (synergy)
  • ยท Session 3 โ€” Reduced TE (causal direction)
  • ยท Session 4 โ€” Sliced Wasserstein (regime stability)

Cross-references