Asked whether 10 fractal, multifractal and chaos features add information orthogonal to the live production panel. Ten per-candidate results came back, but the closing cross-candidate matrix the index promises never ran.
| Quantity | Value | Meaning |
|---|---|---|
| Candidates evaluated | 10 of 10 | every declared candidate has a page |
| Result split | 6 / 4 / 0 across the gate bands, 0 pending | |
| Valid cells per candidate | 207 (10 regime slices x 21 symbol-threshold cells) | |
| Max |rho| across the slate | 0.56 to 0.94 | bands: clear <= 0.85, watch 0.85-0.95 |
| Compute cap | 2 cores, 4 GB, read-only ClickHouse | spare capacity only |
| Firing | What it did |
|---|---|
| 01-10 | One candidate per overnight firing, each rendered with per-slice tables |
| 02 | First result, sitting exactly on the 0.85 band edge |
| 05 | Cleanest result of the slate at max |rho| 0.56 |
| 10 | Last candidate at 0.81; bar_petrosian_fd the worst competitor in 80 cells |
| Finish | Never fired โ no cross-candidate artifact exists in the folder |
The Finish cross-candidate matrix. The index defers two named within-slate clusters to it, so without it the campaign cannot say which of its clear candidates are mutually redundant, and no promotion decision was ever reached. Needs one Finish firing from the OPERATOR; the sibling CECP campaign's crosscheck page is the template. Correcting an earlier reading: the spike ledger's lone PENDING row is timestamped 20 hours BEFORE the last dashboard update โ a stale pre-run smoke artifact, not a contradiction of the 10 published results.
A declared exit artifact is missing, and its presence is exactly what distinguishes the settled CECP sibling from this campaign.
'Within-slate clusters (#45 vs #24, #35 vs #36) are resolved by the cross-candidate matrix at Finish.' (index.html) โ no in-campaign stop declaration exists.
Lifecycle status is process state, not a judgement of the findings. Results are stated as numbers with their uncertainty.
Evaluating 10 fractal / multifractal / chaos / complexity candidate features for whether each adds information orthogonal to the live production panel โ one candidate per overnight firing, read-only across 10 regime slices ร 21 (symbol, threshold) cells on ClickHouse. Gate: โฯโโค0.85 โง Rยฒโค0.85 โง h_normโฅ0.05 (worst-valid-cell). Scope: orthogonality only โ no null, no signal/target. Resource-capped to 2 cores @ 200% / 4 GB, spare capacity only.
| # | Candidate | Verdict | Max โฯโ | Valid cells | Cost |
|---|---|---|---|---|---|
| 01 | ๐ธ๏ธ Ordinal partition network โ global node entropy network-complexity ยท ordpy | WATCH | 0.94 | 207 | medium |
| 02 | ๐ชข Higuchi fractal dimension fractal-dimension ยท antropy | PASS | 0.85 | 207 | fast |
| 03 | ๐ Generalised Hurst H(q=2) hurst-multifractal ยท nolds | WATCH | 0.87 | 207 | medium |
| 04 | ๐ผ Permutation JensenโShannon distance to 1/fยฒ noise complexity-entropy ยท ordpy | WATCH | 0.90 | 207 | fast |
| 05 | ๐ MF-DFA multifractal spectrum width ฮฮฑ (price) multifractal ยท custom | PASS | 0.56 | 207 | slow |
| 06 | ๐ ฮh(q) on |returns| โ volatility multifractality multifractal-volatility ยท MFDFA | PASS | 0.68 | 207 | slow |
| 07 | ๐ ฮh(q) on returns multifractal ยท MFDFA | PASS | 0.65 | 207 | slow |
| 08 | โก ฮh(q) on volatility increments (ฮฯยฒ) multifractal-volatility ยท MFDFA | PASS | 0.71 | 207 | slow |
| 09 | ๐ Correlation dimension Dโ (chaos) chaos-dimension ยท nolds | WATCH | 0.89 | 207 | slow |
| 10 | ๐ฆ Largest Lyapunov exponent (Rosenstein) chaos ยท nolds | PASS | 0.81 | 207 | slow |