Hypothesis — expF_multiple_testing#

Pre-specified 2026-08-12 before the battery ran. RC/SPA/DSR passed the §8.1 gate first (4 tests: pure noise never survives; a planted profitable rule always does).

Hypothesis
  The survival curve (charter result-type E): after honest correction for
  the FULL searched universe, few or none of the C/D scan leads survive.
  Prior: the daily 2008-2015 reversal family and possibly ES->SPY have the
  best odds; the tiny 2014-2015 lead-lag residuals and most intraday
  reversion cells should die.

Falsification criterion
  Not falsifiable as a directional claim — the DELIVERABLE is the measured
  survival rate at each layer. The pipeline is broken (investigate, not
  publish) if a rule from the expE calendar family survives SPA (expE
  already showed all 8 inside the permutation band).

Artifact null(s)
  The searched-universe null itself: RC/SPA bootstrap under H0 "no rule
  beats zero", universe = EVERYTHING the scans looked at (not only the
  FDR survivors) — 2 signed variants of every cell/pair/class.

Method (pre-declared)
  Rule construction (mechanical, no tuning):
    R-family (expC, 127 cells x2 signs): contrarian rule at the cell's
      timeframe on RTH trade-time returns, pos_t = -sign(r_{t-1});
      1day cells use the previous daily return. Daily aggregation; days
      without data = 0 (idle).
    L-family (expD, 49 pairs x2): follower timed by leader's previous
      1min return on both-fresh minutes, daily aggregation.
    C-family (expE, 8 classes x2): +/-(r_t - unconditional mean) on class
      days, 0 elsewhere (drift-adjusted so "long Mondays" cannot free-ride
      the equity premium).
  Blocks (common day calendars): expC 2000-2007, expC 2008-2015, expC 1min
  2014-2015, expD 2006-2007, expD 2014-2015, expE train.
  Battery per block: (1) naive |t|>1.96 count; (2) BH-FDR on two-sided
  rule p-values across ALL blocks jointly; (3) Hansen SPA (500 stationary
  bootstraps, mean block 5 days, seed 42) + StepM-style step-1 survivor
  count (rule t >= bootstrap max-stat 95th pct); (4) DSR of each block's
  best rule, n_trials = total universe size, sr_variance across the
  universe. White RC reported alongside SPA.
  IMPORTANT honesty note: rules are evaluated on the SAME train data the
  scans ran on — expF measures survival of the in-sample search under
  correction. Out-of-sample survival is expH's job on the validation split.

Result
  Run 20260812T072749Z — 0 network requests (582 cache hits; the frozen
  cache carried the whole battery). Universe: 372 signed rules, 6 blocks.
  Survival funnel (GROSS, frictionless): naive |t|>1.96 = 232 (62%) ->
  BH-FDR = 226 (61%) -> SPA step-1 = 68 (18%). All expC/expD blocks reject
  at the bootstrap floor (RC and SPA p = 0.002); the expE calendar block
  survives nothing (naive 0, spa_p 0.87) — the pipeline-broken tripwire did
  NOT fire. Best rules carry annualized Sharpe 10-31 — physically absurd,
  the §12 red flag. Intersection with the artifact-adjusted expC triage:
  12/14 cells also pass SPA. L-family step-1 includes -L:ES->SPY (the
  splice-invariant basis effect) and -L:SPX->SPY (the KNOWN index-staleness
  artifact, deliberately kept in the universe as a canary — it survives
  statistical correction, which proves the point below).

Interpretation
  (Level 0.) The survival curve's headline is METHODOLOGICAL and it is the
  strongest result of the project so far: statistical correction corrects
  for SEARCH, not for MECHANISM. 18% of gross rules survive Hansen SPA —
  and the survivors are dominated by bounce harvesting (a contrarian rule
  earns -autocov1 > 0 on paper and pays the spread in reality) plus
  frictionless lead-lag timing; the known artifact (SPX->SPY) sails through
  SPA unharmed. Honest validation therefore REQUIRES all three layers:
  artifact nulls (scans) AND search correction (expF) AND costs (expG).
  The double-filtered pool going to expG: 12 reversion cells + ES->SPY +
  the expD 2014-2015 FDR set. DSR by block is reported but is mostly a
  universe-heterogeneity diagnostic here (bounce-inflated Sharpe variance);
  documented, not over-read.

Next experiment
  expG (cost frontier) on the double-filtered pool; expH (validation split)
  for whatever survives costs.

Analysis — expF_multiple_testing#

Run: results/expF_multiple_testing/20260812T072749Z/results.json. Battery: 372 signed rules (every cell/pair/class the C/D/E scans searched), 6 blocks, White RC + Hansen SPA (500 stationary bootstraps) + BH-FDR + DSR. RC/SPA/DSR §8.1-gated first. 0 network requests — the entire experiment ran from the frozen cache (the reproducibility anchor doing its job).

The survival curve (charter result-type E) — gross, frictionless#

layer	survivors	rate
universe (searched)	372	100 %
naive |t| > 1.96	232	62 %
BH-FDR 5 %	226	61 %
Hansen SPA step-1	68	18 %

The headline is methodological#

The SPA survivors carry annualized Sharpes of 10–31 — physically absurd, which is the charter-§12 red flag, and the diagnosis is clean:

1. Statistical correction corrects for search, not for mechanism. A contrarian rule mechanically earns −autocov₁ > 0 on paper wherever bid-ask bounce exists — frictionless, that is "profit"; in reality it is the spread you would pay. SPA has no way to know that.
2. The canary proves it: −L:SPX→SPY — the index-staleness artifact we know is fake (T3, measured twice) — survives SPA comfortably.
3. The expE calendar block survives nothing anywhere (naive 0, SPA p 0.87): the pre-registered pipeline-broken tripwire did not fire.

Honest validation therefore requires ALL THREE independent layers — artifact nulls (the scans), search correction (this experiment), and costs (expG) — and no one of them substitutes for another. This goes into methodology.md as a design axiom, with this experiment as the demonstration.

The double-filtered pool (artifact-adjusted ∩ search-corrected, gross)#

12 of the 14 expC artifact-adjusted triage cells also clear SPA: AAPL/ATRO/BKE/HTD/ICUI/MSFT/SLF/XOM 5min, AXDX 30min, CKX 1day, JPM/NVDA 1min. Lead-lag: −L:ES→SPY (splice-invariant basis reversion) plus the expD 2014-2015 FDR residuals; −L:SPX→SPY is excluded from the tradable pool (routed to candidate 03 as the artifact demonstration). This pool — and nothing else — proceeds to expG.

Caveats#

In-sample by design (rules evaluated on the data that surfaced them; validation split untouched until expH). DSR values are reported per block but the bounce-driven Sharpe heterogeneity inflates sr0; treat them as universe diagnostics, not verdicts.

expF_multiple_testing#

Multiple-testing survival: White RC, SPA, FDR, deflated Sharpe on C-E output

Status: completed 2026-08-12 — survival curve measured: 372 -> 232 (naive) -> 226 (FDR) -> 68 (SPA, 18%), all gross; survivors dominated by bounce; the known SPX artifact survives SPA (statistical correction != artifact correction). Double-filtered pool -> expG.