Iteration 04 — `bar_approximate_entropy` PASS

Pincus 1991 (PNAS 88:2297) · rolling 200-bar window · updated 2026-06-02 04:05 UTC

1 · Context of work

Candidate

bar_approximate_entropy (antropy)

Frozen test constants

m=2 · r=0.2*std · metric=chebyshev

Axis-2 status

defensible

Window

rolling 200 bar closes (project-wide LOOKBACK)

Matrix version

CH_FEATURE_BANLIST.md @ d74f75e1 · FRAMEWORK.md Axis-1 (|ρ|≤0.85, R²≤0.85, h_norm≥0.05)

Substrate

opendeviationbar_cache.open_deviation_bars · 21 (symbol,threshold) × 10 slices · read-only

2 · Grounded this iteration

Valid cells

207

Slices completed

10 / 10

PASS / WATCH / BAN

207 / 0 / 0

Max |ρ| vs existing features

0.7378

Max combinatorial R²

0.7343

Min h_norm (mass-collapse guard)

1.0000

Gate: PASS = |ρ|≤0.85 AND R²≤0.85 AND h_norm≥0.05 across every valid bar set. Grounded — all valid cells PASS.

Most frequent worst competitors:

trade_intensity — appeared as worst competitor in 82 cell(s)
price_impact — appeared as worst competitor in 25 cell(s)
volume_per_trade — appeared as worst competitor in 16 cell(s)

3 · Next steps

Iteration 5: bar_sample_entropy (tested at m=2 · r=0.2*std · metric=chebyshev).

4 · Pending tasks

3 bar set(s) skipped (insufficient bars / empty cell).
Axis-2 final status deferred to Stage 2 (project-wide constant ratification, after the full family is evaluated).
Cross-candidate ρ/LOO-R² matrix (run_cross_candidate.py) must run before any PROMOTE — catches sibling clusters.

Generated by emit_iteration_html.py · spike_2026-06-02-entropy · raw per-cell verdicts in results/bar_approximate_entropy_slice*.jsonl (bigblack-only, not committed).

Iteration 04 — bar_approximate_entropy PASS

1 · Context of work

2 · Grounded this iteration

3 · Next steps

4 · Pending tasks

Iteration 04 — `bar_approximate_entropy` PASS