---
project: anomaly-atlas
document: Research gaps
author: Simon-Pierre Boucher
contact: contact@spboucher.ai
data_source: hfmarketdata.io
created: 2026-08-12
modified: 2026-08-12
status: reviewed
---

# Research gaps (Phase 3) — 22 testable hypotheses

Each hypothesis: **(a)** what existing work shows, **(b)** what is untested
on open HF data at this granularity, **(c)** why it matters, **(d)** plausible
mechanism, **(e)** biggest artifact/statistical reason it would be fake,
**(f)** smallest Mac-runnable falsifying experiment. Every hypothesis has an
explicitly constructable artifact null (charter §6 requirement). Format
compresses (a)–(f) into labeled lines.

The hypothesis *count* here is the multiple-testing budget input for expF —
expanding this list later requires a LOG entry and re-budgeting (§12).

## Family R — Mean-reversion (Q1)

### H01 — Liquid-name 1–60 min reversion is zero net of bounce
(a) Chordia et al. 2005: arbitraged within minutes by 2005. (b) Never
measured net of a *measured* bounce null on open 1min data 2000–2026.
(c) Anchors the whole survival curve; expected-negative. (d) None expected —
that is the point. (e) T1 bounce; T2 staleness. (f) VR/AC1 on 20 pre-specified
liquid names × 4 sub-periods vs expB bucket nulls; falsified if excess
reversion CI excludes 0 anywhere.

### H02 — Daily individual-stock reversal decay curve 2000–2026
(a) Lehmann/Jegadeesh gross reversal; Blume–Stambaugh: bounce halves it;
McLean–Pontiff decay. (b) The *net-of-measured-bounce* decay path on open
data. (c) Cleanest decay re-measurement. (d) Liquidity-provision compensation
shrinking with spreads. (e) T1 (dominant), T4. (f) Weekly-rebalanced reversal
deciles, both close conventions, net of EDGE spread; falsified (as decay
story) if net effect is stable across sub-periods.

### H03 — Index VR momentum has converged to 1; residual = Fisher artifact
(a) Lo–MacKinlay VR>1 (1988); Fisher 1966 stale-index bias. (b) SPX-vs-SPY VR
divergence as a direct artifact quantification. (c) Methodological: measures
the Fisher share. (d) Stale index prints. (e) T2/T3 — deliberately the object
of study. (f) VR(q∈{2,5,10,30}) on SPX vs SPY 1day+30min, 2008–2026; the
artifact claim is falsified if SPX VR ≈ SPY VR.

### H04 — Post-jump 1min overreaction reversal
(a) Overreaction literature is daily+; intraday jump reversion is mostly HFT
lore. (b) Reaction after >5σ 1min idiosyncratic moves on open data, net of
bounce (jumps widen spreads — the null must be *conditional* on the jump).
(c) Tests overreaction where attention is scarcest. (d) Liquidity vacuum then
refill. (e) Conditional bounce inflation (spread doubles at jumps → mechanical
reversal). (f) Event study on pre-specified jump definition, conditional Roll
null from post-jump spread estimates; ±CI on excess reversal.

### H05 — Crypto 1min reversion (no auction, 24/7, younger market)
(a) Crypto microstructure literature is thin and exchange-specific. (b) Open
1min BTC/ETH reversion with a constructable bounce null, 2013–2026 maturity
sweep. (c) Tests the "young markets are less efficient" prior cleanly.
(d) Retail flow + fragmented liquidity. (e) T1 (vendor aggregation may smooth
prints); volume field semantics unverified. (f) Same battery as H01 on
BTC/ETH/top-10, by year; falsified if excess reversion ≈ 0 across all years.

### H06 — FX 1min reversion at session boundaries
(a) FX microstructure well-studied on proprietary data. (b) Open 1min FX
around the 17:00 ET roll and fixings. (c) Session-boundary liquidity is a
mechanism magnet. (d) Dealer inventory at rollover. (e) T7 session semantics;
sparse Sunday bars (expA: single 00:xx bar). (f) AC1/VR by hour-of-week grid
(pre-counted), permuted hour-labels null.

## Family L — Lead-lag (Q2)

### H07 — Large→small lead-lag: decay + artifact share
(a) Lo–MacKinlay 1990; Scholes–Williams artifact. (b) Joint decay curve AND
staleness-share decomposition on open data 2000–2026. (c) Settles how much of
a textbook effect was ever real. (d) Slow information diffusion. (e) T3 —
measured in expB (+0.047 mid-tercile). (f) Size-decile xcorr on both-fresh
subsamples vs staleness-matched null, per sub-period.

### H08 — SPY → sector ETFs at 1min is zero on fresh pairs
(a) Index→constituent diffusion literature (daily, older). (b) ETF↔ETF at
1min on open data. (c) Expected-negative control extending H07 to the most
liquid pairs. (d) None expected. (e) T3 residual staleness in low-volume
sector ETFs. (f) xcorr SPY↔{XLF,XLE,XLK,…} both-fresh, staleness null;
falsified if any |lead| beats null after FDR.

### H09 — ES futures → SPY at 1min: does any ordering survive the floor?
(a) Sub-second futures lead is established (proprietary data). (b) Whether
ANY of it is visible at the 1min floor on open data, and whether the futures
splice variant changes the answer. (c) Structural price-discovery question
our data can just barely reach. (d) Futures-first discovery. (e) T3/T7
session mismatch; splice artifacts (3 variants = built-in robustness check).
(f) ES↔SPY xcorr 09:31–15:59 both-fresh, all 3 adjustments, sub-periods
2008–2026.

### H10 — Crypto weekend return predicts Monday open gap of crypto-exposed equities
(a) Effectively untested on open data; related overnight-information papers.
(b) The full 24/7→session-open information relay. (c) One of the few genuinely
NEW questions this dataset can pose. (d) Equity prices cannot move while
crypto does. (e) Selection of "exposed" set post-hoc (must pre-register:
e.g. COIN, MSTR, RIOT, MARA, HUT + placebo set); overlapping market-wide
weekend news (control: SPY gap). (f) Regress Monday open gap on BTC weekend
return vs placebo set + permuted-weekend null; ~400 weekends.

### H11 — SPX→SPY "lead" is 100 % staleness artifact
(a) Fisher 1966. (b) Direct decomposition at 1min on open data (expB seed:
+0.065). (c) Publishable artifact demonstration (charter result-type C/D).
(d) Index print lag. (e) None — the artifact IS the claim. (f) Show the lead
disappears when SPX grid is lagged by its print-latency estimate; falsified
if a residual lead survives the correction.

### H12 — Daily options-chain activity → next-day underlying volatility
(a) Options volume→volatility literature (mixed, mostly intraday quotes).
(b) With *daily* chains only: does aggregate chain volume/OI change add to a
HAR-RV vol forecast? (c) Uses the dataset's unique options depth; forecasting
vol (not returns) dodges the efficiency prior. (d) Informed vol trading.
(e) Mechanical volume–vol contemporaneity (must be strictly lagged);
HAR benchmark must be strong. (f) HAR-RV ± options features, Diebold–Mariano
with block bootstrap, pre-2020 train / post-2020 validation.

## Family C — Calendar (Q3)

### H13 — Turn-of-month survived 2006–2026
(a) Ariel; Lakonishok–Smidt; Marquering et al.: last survivor as of 2006.
(b) Its post-2006 fate on open data. (c) THE calendar re-test. (d) Pension/
payroll flows. (e) Calendar p-hacking (window choice) — ONE pre-registered
window (−1..+3); OpEx overlap. (f) SPY daily 2000–2026, permuted-calendar
null, SPA vs full window universe, sub-periods.

### H14 — Monday effect stays dead (negative control)
(a) French 1980; decayed per Schwert/Marquering. (b) 2000–2026 open-data
confirmation. (c) Calibrates the calendar pipeline on a known corpse — if we
"find" it, the pipeline is broken. (d) None. (e) T4/T7 conventions. (f)
Day-of-week means + permuted null; expected inside null bands.

### H15 — Intraday momentum (Gao) post-publication decay
(a) Gao et al. 2018 (in-sample 1993–2013); Baltussen 2021 mechanism. (b) The
2018–2026 out-of-sample fate — untouched by the original authors, perfectly
measurable here. (c) Textbook McLean–Pontiff test at intraday scale. (d)
Gamma hedging → should persist when dealer gamma is short. (e) Overnight-gap
spec choice; U-shape spread costs at both trade times. (f) Frozen published
spec on SPY 30min bars, split pre/post-2018, gamma-state proxy from our
options chains; DSR-corrected.

### H16 — Overnight premium is (partly) a close-convention artifact
(a) Overnight-vs-intraday split literature. (b) Sensitivity of the premium to
close/open conventions quantifiable with both series (T4: 8bp/day discrepancy
seed). (c) Convention share of a famous stylized fact. (d) Auction dynamics.
(e) The artifact is the object. (f) Recompute premium under 4 convention
combos on 20 liquid names + SPY; report the spread of estimates.

### H17 — Crypto "weekend effect" with permuted-calendar null
(a) Anecdotal/thin literature. (b) Systematic day-of-week/hour-of-week on
open 24/7 1min data 2013–2026 with honest correction. (c) Only market with
no session artifacts — the cleanest calendar laboratory. (d) Retail weekend
flow. (e) Pure MT risk (168-hour grid!) — strict pre-count + FDR + SPA.
(f) Hour-of-week grid on BTC/ETH, permuted labels, FDR; survivors → SPA.

### H18 — Holiday-adjacent effects are dead
(a) Lakonishok–Smidt documented; decayed per Marquering. (b) 2000–2026
confirmation, pre-holiday and post-holiday, exchange calendar from data
gaps. (c) Completes the calendar corpse count. (d) None expected. (e) Small-n
(≈9 holidays/yr) → low power, wide CIs; report power explicitly. (f) Event
windows vs permuted-date null.

### H19 — DST-transition week distortions
(a) Sparse literature (sleep/mood papers, disputed). (b) At 1min: does the
hour-of-day return/vol profile mis-align in DST weeks (a T7 check that could
also be a mild real effect)? (c) Doubles as a data-integrity test. (d)
Clock-relative order flow. (e) OUR OWN timestamp handling (T7) — the test
must first prove the pipeline handles DST. (f) Compare DST-week intraday
profiles vs adjacent weeks, permuted-week null.

## Family V — Volatility & structure (supporting)

### H20 — Intraday bounce/staleness profile (U-shape of the artifact)
(a) Wood et al. U-shape. (b) The intraday profile of OUR bounce null —
taxonomy open item. (c) Input required by expE/expG; positive control.
(d) Known microstructure. (e) None. (f) Roll/EDGE spread by half-hour bucket
on the expB universe.

### H21 — Realized-vol persistence (positive control)
(a) HAR-RV: overwhelming evidence of vol persistence. (b) n/a — we USE it as
the pipeline's positive control: a detector that misses vol persistence is
broken. (c) Guards against over-aggressive nulls (Type II). (d) Volatility
clustering. (e) Noise-biased RV (use 5min subsampling per Zhang et al.).
(f) HAR-RV R² on SPY must be strongly positive OOS; failure = pipeline bug.

### H22 — OpEx-week patterns in volume/vol (options-expiration calendar)
(a) Practitioner lore + some evidence on pinning. (b) With actual expiration
calendars from our chains: volume/vol/return patterns on OpEx Fridays vs
matched Fridays. (c) Mechanism-rich calendar effect with a natural control
group. (d) Hedge unwinds. (e) Confound with turn-of-month (H13) and
quarter-end — joint design needed. (f) OpEx vs non-OpEx Friday panel,
permuted-Friday null.

---

**Budget declared: 22 hypotheses** (+ their pre-specified parameter grids,
enumerated in each experiment's hypothesis.md). Preference for prototyping
goes to hypotheses with the cleanest nulls and highest information value per
CPU-hour (Phase 4 ranking → `candidate_ranking.md`).