--- project: anomaly-atlas document: expE_calendar_scan/hypothesis author: Simon-Pierre Boucher contact: contact@spboucher.ai data_source: hfmarketdata.io created: 2026-08-12 modified: 2026-08-12 status: final --- # Hypothesis — expE_calendar_scan *Pre-specified 2026-08-12 before the scan ran. The calendar space is where p-hacking is easiest (STW 2001), so the hypothesis budget below is a hard cap: any test not listed here is not run.* ```text Hypothesis Tests H13/H14/H18 (scan half) + measures H20. Priors from the verified literature: the Monday effect is dead (French 1980 -> decayed); holiday effects are dead; turn-of-month was the last survivor as of 2006 (Marquering et al.) and its TRAIN-period status is the open question. H20 (intraday U-shape of vol and spread) is expected to be strongly present — it is a POSITIVE CONTROL and a cost-model input, not an anomaly. Falsification criterion H14 (negative control): the pipeline is BROKEN if any weekday passes the family-wise permutation test — investigate the pipeline, not the market. H13: "turn-of-month survived in train" is falsified if the ToM window mean fails the family-wise permutation test. H20: the U-shape control FAILS the pipeline if flat (spread/vol profile must peak at open/close). Artifact null(s) Permuted-calendar null (MANDATORY per charter §8.5): within-YEAR permutation of daily returns (preserves annual regimes, destroys calendar alignment), n = 2000, seed = 42. Family-wise: the max-|statistic| across all 8 tests is recomputed on every permutation (mini reality check) — each observed statistic gets both a marginal and a family-wise p. T4 guard: daily bars only (one close convention throughout). Method (budget pre-counted: 8 tests) Instrument: SPY 1day, adj_splitdiv, TRAIN 2000-2016 (single pre-declared instrument — no cross-sectional expansion). Statistics, each = mean daily return in the class minus overall mean: T1..T5 day-of-week (Mon..Fri) [H14, 5 tests] T6 turn-of-month window: last trading day through +3 first trading days of the next month [H13, 1 test] T7 pre-holiday day (day before a non-weekend gap) [H18] T8 post-holiday day [H18] Corrections: BH-FDR over the 8 marginal permutation p-values AND the family-wise max-stat p. Sub-period stability (2000-2007 / 2008-2015) reported descriptively for anything that passes. H20 measurement (no hypothesis test): per half-hour bucket on the liquid 12, 1min, 2014-2015 (cached): median |1min return|, EDGE spread per bucket, staleness per bucket -> taxonomy T1/T2 intraday profile. Result Run 20260812T065222Z (1 new request — everything else cache-served; 4 024 trading days). NOTHING SURVIVES: all 8 observed effects sit inside their permuted-calendar 95% bands (family-wise p >= 0.93 everywhere; best marginal p = 0.24, turn-of-month). Monday: -1.4 bp, p_marg 0.73 — the negative control held. Turn-of-month: +4.7 bp overall, and DECAYING inside train (+7.9 bp 2000-2007 -> +1.6 bp 2008-2015). Holiday effects: wide bands (n=144), nothing. H20 control: PASSED — volatility U-shape (6.6 bp open / 2.4 midday / 2.9 close); spread declines monotonically (2.8 -> 1.2 bp), i.e. an L-shape, not a U — recorded as measured. Interpretation (Level 0.) The cleanest possible calendar answer on this data: the pre-declared 8-test budget produces zero candidates. The 2006-era "last survivor" (turn-of-month) does not clear the permuted-calendar bar in train and decays within it — candidate 02 (H13) now carries a strong dead-on-arrival prior into its validation-split protocol. Both pipeline controls behaved (Monday stayed dead; U-shape present), so this negative is a *finding*, not a power failure — though holiday-class power is low (wide bands) and is reported as such. H20 numbers feed taxonomy T1/T2: first-30-minutes claims face 2-3x the midday artifact level. Next experiment expF: correction battery (White RC/SPA + DSR with the 22-hypothesis budget) over everything C/D/E surfaced. ```