Hypothesis — expB_artifact_baselines#

Pre-specified 2026-08-12, before any real-data measurement. Detectors passed the §8.1 synthetic gate first (11 tests, benchmarks/synthetic/).

Hypothesis
  The three mechanical artifacts identified in expA are MEASURABLE and
  MATERIAL in this dataset at 1-minute resolution:
  (i)   bid-ask bounce produces negative AC1 in 1min returns, larger for
        less liquid names (Roll-implied relative spread as the null level);
  (ii)  illiquid names are stale on a large fraction of RTH minutes;
  (iii) LOCF-gridding makes SPY spuriously "lead" stale tickers at +1 min,
        with the artifact magnitude increasing in staleness.

Falsification criterion
  The experiment fails if the nulls are unusable as baselines: bounce AC1
  indistinguishable from 0 across the liquidity spectrum (|median AC1| CI
  covering 0 for the bottom-liquidity tercile), or no monotone relation
  between staleness and the +1min SPY cross-correlation (Spearman rho <= 0
  across tickers).

Artifact null(s)
  This experiment BUILDS the artifact nulls; its own null is the synthetic
  ground truth (§8.1 gate) — detectors verified to read 0 on random walks.

Method
  Universe (pre-specified): LIQUID = {AAPL MSFT NVDA AMZN GOOGL META TSLA
  JPM XOM UNH} + {SPY QQQ}; RANDOM = 30 tickers drawn from the full 1min
  stock ticker list with numpy seed 42 (deterministic given the cached
  list). Window: 2024-01-02 → 2024-04-01, RTH only (09:30 ≤ t < 16:00),
  adjustment adj_split, all data via hf_client (cached).
  Per ticker: staleness ratio on the 390-min RTH grid; trade-to-trade 1min
  log-return AC1 with moving-block bootstrap CI (block = 1 day, n=300,
  seed=42); Roll implied relative spread; VR(5), VR(30).
  Cross: lagged xcorr (±3 min, NaN-aware, day-boundary safe) of LOCF-grid
  returns vs SPY; SPX-vs-SPY as the index-staleness case.
  This is a DESCRIPTIVE measurement of artifact levels, not an anomaly
  claim: no OOS split; every number is Level 0 by construction.

Result
  Run 20260812T055602Z (47 requests, 795 185 rows, 0 retries). 31/42
  tickers had enough data (>=30 days, >=2000 returns); 11 dropped, listed
  in results.json. Median by staleness tercile (staleness | AC1 | Roll rel
  spread | VR30 | SPY-leads-+1min):
    fresh  0.000 | -0.009 | 1.4 bp  | 0.972 | +0.005
    mid    0.107 | -0.051 | 3.8 bp  | 0.901 | +0.047
    stale  0.691 | -0.232 | 17.0 bp | 0.547 | +0.021
  Spearman(staleness, SPY-leads-+1) = +0.43 (> 0: monotone). Extreme case
  RITM.B: staleness 0.90, AC1 -0.253, VR30 0.35. Mega-caps: AC1 CI covers
  0 and the Roll estimator is undefined (positive autocov) — no measurable
  bounce at the top. SPX-vs-SPY: corr 0.965 at lag 0 and +0.065 with SPY
  leading by 1 min (index prints lag the tradable ETF).

Interpretation
  Both falsification criteria FAILED to trigger: the nulls are usable.
  Headline: an uncorrected VR/AC1 scan on mid/low-liquidity names is
  DOMINATED by artifacts — VR30 of 0.55 and AC1 of -0.23 arise with no
  economic mean reversion whatsoever. All numbers Level 0 (descriptive
  null levels), as pre-specified. Magnitudes recorded in
  research/artifact_taxonomy.md.

Next experiment
  expC (reversion scan) consumes these nulls: any reversion claim must
  exceed the bounce null for its liquidity bucket. Phase 1 literature
  sweep proceeds in parallel.

Analysis — expB_artifact_baselines#

Run: results/expB_artifact_baselines/20260812T055602Z/results.json (hardware manifest embedded; 47 network requests, 795 185 rows, protocol pre-specified in hypothesis.md; detectors passed the §8.1 synthetic gate — 11 tests — before touching this data).

The measured artifact nulls (Q1 2024, RTH 1min, 31 tickers)#

Staleness tercile	median staleness	median AC1	Roll rel. spread	VR(30)	SPY leads +1min
fresh	0.000	−0.009	1.4 bp	0.972	+0.005
mid	0.107	−0.051	3.8 bp	0.901	+0.047
stale	0.691	−0.232	17.0 bp	0.547	+0.021

• Bounce/staleness dominate naive reversion metrics. With zero planted economics, illiquid names show VR(30) = 0.55 and AC1 = −0.23 (extreme: RITM.B, 90 % stale minutes, VR30 = 0.35). Any reversion scan that does not clear these levels for its liquidity bucket is measuring market plumbing.
• Mega-caps show no measurable bounce at 1min: AAPL/SPY/NVDA AC1 CIs cover 0 and the Roll estimator is undefined (positive lag-1 autocov) — the bounce null is liquidity-dependent, not universal.
• The stale-price lead-lag artifact is real and monotone: SPY spuriously "leads" tickers by +1 min in proportion to their staleness (Spearman = +0.43). It peaks in the mid tercile (+0.047): the stalest names trade so rarely that even LOCF correlation collapses — the artifact is worst where it is least obvious.
• SPX-vs-SPY: contemporaneous corr 0.965, and a +0.065 cross-correlation with SPY leading by 1 minute. An "ETF price discovery leads the index" finding is manufactured by index print staleness — measured here so Q2 hypotheses must beat it.

Limitations#

Level 0 by construction (descriptive nulls; single quarter; one venue's bar convention). 11/42 tickers dropped for insufficient data — the null for ultra-illiquid names is therefore understated. Q1-2024-specific levels; expC should re-measure per period rather than reuse these constants blindly.

Verdict#

Complete — nulls established and usable. Both pre-specified falsification criteria failed to trigger. Numbers are recorded in research/artifact_taxonomy.md; expC (reversion scan) must report every effect net of the bucket-matched bounce null, and expD must run the synchronized-vs-raw timestamp comparison this experiment quantified.

expB_artifact_baselines#

Artifact baselines: null distributions of bounce, staleness, non-synchronous lead-lag

Status: completed 2026-08-12 — see analysis.md; null magnitudes recorded in research/artifact_taxonomy.md.