โ€บNavigation
Dashboard โ€บ Three-Axis Probes

axis 1 ยท empirical ยท operational

๐ŸŽฏ Orthogonal

Does this feature tell us something no existing feature already tells us? Established empirically on real bars, never from the estimator's paper. The cross-asset / cross-regime sweep lives HERE (Terry msg 429 step 3) โ€” it is the orthogonality substrate, not the agnostic test.

Machinery

layerwhat it measureswhere
fast Spearman |ฯ|monotone redundancy vs every shipped column, per cell; BAN >0.95 ยท WATCH 0.85โ€“0.95 ยท PASSorthogonality_probe.py
h_normmass-collapse: does the quintile encoding retain any information (<0.05 = constant-like)orthogonality_probe.py
regime aggregationmin_regime_pass_rate across 7 epochs โ€” the honest number (same-epoch slices correlate)aggregate_candidates()
W1 distribution (report-only)Wasserstein-1 to the zero ideal = mean per-cell score on [0,1]; the L1 complement of the worst-cell Lโˆž verdict. Never gates.campaign_summary_adapter.py
Chatterjee ฮพnonlinear / functional / asymmetric redundancy (ฮพ=1 โ‡” Y=f(X)); Linโ€“Han analytic nullcodec_foci_eval.py
CODEC T(a,b|rest)conditional redundancy given the rest of the panelcodec_foci_eval.py
O-informationhigher-order synergy/redundancy tripletscodec_foci_eval.py
MOWFO walk-forwardleakage-free discover-on-train / verify-on-test stability (purge 200, embargo 10)walk_forward()

Honest limits (why PASS โ‰  proof yet)

Verdict contract (what the gate reads)

Two accepted evidence dialects: (1) the probe summary โ€” PASS when min_regime_pass_rate โ‰ฅ 1.0 (holds in every regime epoch); (2) campaign/spike per-cell records via campaign_summary_adapter.py โ€” worst-valid-cell over โ€–ฯโ€– + panel-Rยฒ + h_norm, stored labels re-derived from raw numbers, WATCH honest non-promotable. No evidence โ‡’ PENDING, never a faked pass.

Distribution view โ€” (worst, W1) (2026-06-13, report-only)

The verdict collapses 208 cells to the single worst (an Lโˆž view: "redundant somewhere?"). Beside it we now report Wasserstein-1 to the zero-redundancy ideal โ€” the L1 complement: "redundant everywhere?". On [0,1] scores, W1 reduces exactly to the mean of the per-cell scores (Vallender 1974) โ€” auditable, computed through scipy and identity-asserted.

Same worst, different story (real entropy-spike data):
bar_svd_entropy โ€” worst |ฯ| 0.97 / W1 0.53 โ†’ a regime-local spike.
bar_permutation_entropy โ€” worst |ฯ| 0.97 / W1 0.77 โ†’ redundant broadly.
Worst-cell alone cannot separate them; W1 does. Both stay BAN โ€” W1 never gates, it only explains the worst.

Deep dive: AXIS-1-ORTHOGONAL.md ยท AXIS-1-W1-DISTRIBUTION.md

โ† HubParameterless โ†’