Every hypothesis tested on the research split, logged against the project's 2,000-hypothesis budget. Scans never touch Validation or the Vault — they are layer-3 measurement, not execution. A null result logged here is a completed experiment, not a failure of the machinery: it retires a hypothesis and spends one line of the budget either way.
| Date | Name | Split | Hypotheses | Verdict | Links |
|---|---|---|---|---|---|
| 2026-07-27 | htf-level-proximity | research 2017-08-17..2022-12-31 | 12 (cum 273) | Q2 PARTIAL: efficiency NEAR | doc ↗ · results ↗ |
| 2026-07-27 | failed-break-reversion | research 2017-08-17..2022-12-31 | 10 (cum 261) | REVERSION scan: see mechanical checks | doc ↗ · results ↗ |
| 2026-07-27 | mentor-tde-definition | research 2017-08-17..2022-12-31 | 28 (cum 251) | see engine/scans/2026-07-27-mentor-tde-definition-scan.md Results | doc ↗ · results ↗ |
| 2026-07-25 | streak-regime-ethusdt | research 2017-08-17..2022-12-31 | 12 (cum 223) | see results.json | results ↗ |
| 2026-07-25 | streak-regime-btcusd | research 2017-08-17..2022-12-31 | 12 (cum 211) | see results.json | results ↗ |
| 2026-07-25 | universality-xrpusdt | research 2017-08-17..2022-12-31 | 22 (cum 199) | see engine/scans/2026-07-25-crypto-universality-sweep.md Results section | results ↗ |
| 2026-07-25 | universality-bnbusdt | research 2017-08-17..2022-12-31 | 22 (cum 177) | see engine/scans/2026-07-25-crypto-universality-sweep.md Results section | results ↗ |
| 2026-07-25 | universality-solusdt | research 2017-08-17..2022-12-31 | 22 (cum 155) | see engine/scans/2026-07-25-crypto-universality-sweep.md Results section | results ↗ |
| 2026-07-24 | 2026-07-24-streak-follow-btc-v1 | research (execution prereg) | 4 (cum 148) | H1-H3 HELD, pre-declared bar MET: A1 gross > 0 and > C1 (H1), net(Vantage) > 0 (H2), gross positive in >=4/6 year blocks (H3), n=487 traded. H4 (drag law) did NOT hold within tolerance -- measured drag ran ~46% above the ex-ante cost/stop prediction. Registry run 21. | — |
| 2026-07-24 | upstreak-eth | research 2017-08-17..2022-12-31 | 16 (cum 133) | see engine/scans/2026-07-24-upstreak-eth-replication-scan.md Results section | doc ↗ · results ↗ |
| 2026-07-24 | session-structure | research 2017-08-17..2022-12-31 | 24 (cum 117) | Base session structure (1-4) CONFIRMED as pre-declared honest prior: europe_block (08-15 UTC) significantly elevated abs move vs unconditional at all horizons (rand-p=0.000, +3.8/4.1/4.5/6.4% at 15m/30m/1h/4h); asia_block and us_block both suppressed (rand-p=1.000 all horizons); weekend flat (marginal suppression 15m only, rand-p=.159, fading by 1h). Depth-2 interaction (5-6) FAILS the pre-declared combination rule at all 4 horizons: disp4_in_window (n=191, signed +0.227/+0.511/+0.439/+0.032 ATR) never beats BOTH disp4_out_window (n=326, +0.324/+0.424/+0.346/+0.358) AND the unconditioned disp4 record (n=517, +0.288/+0.456/+0.381/+0.238, verified against results-btcusd.json) by more than bootstrap CI overlap -- CIs fully overlap at every horizon, and in_window point estimate is the LOWEST of the three at 15m and 4h. Logged as a filter shrinking n, not an interaction, per the pre-declared alternative reading. Year-blocks: in_window 3-5/6 sign-stable per horizon vs out_window 4-5/6 -- out_window if anything more stable. Session filters remain a tail-risk tool only (twice-observed role, run17/18 H3); does not resurrect disp_follow_btc_v1 (Validation FAIL final). | doc ↗ · results ↗ |
| 2026-07-24 | rhythm | research 2017-08-17..2022-12-31 | 16 (cum 93) | SURVIVED (not retired) per the pre-declared rule: streak_up_ge8 clears all three criteria at both 1h (+0.340 ATR, boot90 CI [0.088,0.590], 6/6 year-blocks positive) and 4h (+0.603 ATR, CI [0.155,1.051], 6/6 positive), monotone vs streak_up_ge5 at the same horizons (+0.039 CI crossing zero / +0.096 CI crossing zero -- ge8 both stronger and CI-clean where ge5 is neither). Asymmetric: streak_down_ge8 fails the rule at every horizon (sign flips vs streak_down_ge5 at 1h/4h, year-blocks only 3-4/6 stable, CI always crosses zero); 15m/30m show no continuation on either side (mild reversal or CI-crossing-zero). Shuffle-p degenerate (=1.000) throughout by construction (constant per-variant direction -- permuting an array of identical values is a no-op), exactly as declared and as noted in the level-reaction precedent; the other three nulls carry the weight. Frequency-matched random null confirms streak_up_ge8 events are genuinely high-vol (rand-p 0.007-0.028 at 1h/4h) -- consistent with, not separate evidence for, the directional finding. n=498 for streak_up_ge8 (>min_events=200, not flagged underpowered, but the smallest sample of the four conditions and the survival rests on it alone). | doc ↗ · results ↗ |
| 2026-07-24 | h4-ignition | research 2017-08-17..2022-12-31 | 20 (cum 77) | FAIL as pre-registered: (a) ignition-continuation REFUTED -- signed effect null-to-reversal at primary threshold (15m -0.032/30m -0.106/1h -0.095/4h -0.101 ATR, shuffle-p 0.305/0.035/0.156/0.410), non-monotone with magnitude (1.25:-0.055/-0.154/-0.151/-0.139 vs 1.75:+0.006/-0.075/-0.080/-0.064 -- effect WEAKENS toward zero as threshold rises, opposite of a plateau), and opposite in sign+dose-direction to the generic displacement scan's positive monotone continuation (disp3-5 at same horizons, see report table). (b) H4-boundary time-of-day structure weakly CONFIRMED: h4_open_any abs move elevated vs unconditional through 1h (rand-p 0.031/0.045/0.088), non-boundary control at-or-below baseline (rand-p 1.000/0.826/0.058); fades to noise by 4h both sides. Matches pre-declared honest prior almost exactly (a reduces to nothing, not even generic displacement; b shows modest time-of-day structure). | results ↗ |
| 2026-07-23 | 2026-07-23-disp-follow-btc-v1 | research (execution prereg) | 4 (cum 51) | ALL 4 HYPOTHESES HELD, pre-declared bar MET -- first candidate with positive net expectancy under realistic costs. A2 disp5 gross +0.4101R, net(Vantage) +0.2800R (n=282), 3.15x measured drag; positive even at the Dukascopy cost bound (+0.0608R). Registry run 19. (Separately: the Validation-split look on this spec, 2026-07-23-disp-follow-btc-v1-validation.md, FAILED 3/4 criteria and burns Validation look 1/10, not a research hypothesis -- not counted here.) | — |
| 2026-07-23 | level-reaction | research 2017-08-17..2022-12-31 | 16 (cum 57) | NULL all 16 cells: touch-bounce ~0 everywhere; break-continuation dead, mildly inverted (failed-break reversion -0.13 ATR/4h logged, not promoted). Shuffle-null degenerate for one-directional conditions (on record). | doc ↗ · results ↗ |
| 2026-07-23 | b4-fit-mentor-stops-vs-r7-levels | Track M descriptive (mentor-trades join, n=20) | 1 (cum 41) | B4-fit NOT CONFIRMED: median 0.5309 ATR (null p5 0.2808) | — |
| 2026-07-23 | 2026-07-23-displacement-event-scan | research 2017-08-17..2022-12-31 | 6 (cum 40) | ETHUSDT REPLICATION CONFIRMED: disp5 +0.720 ATR/+19.4bp vs BTC +0.670/+19.2, 5/6 years, boot CI>0 - crypto-class structure (gold still inverts). | doc ↗ |
| 2026-07-23 | 2026-07-23-displacement-event-scan | research 2017-08-17..2022-12-31 | 6 (cum 34) | REPLICATION REFUTED: signed 1h NEGATIVE all disp levels (-0.22..-0.54 ATR, n=61/30 at disp4/5) on 10y gold — BTC effect is instrument-specific; NOT promoted, no Validation look | doc ↗ |
| 2026-07-23 | 2026-07-23-displacement-event-scan | gbpusd 2024-01..2026-07 | 6 (cum 28) | NOT COMPARABLE | doc ↗ |
| 2026-07-23 | 2026-07-23-displacement-event-scan | research 2017-08-17..2022-12-31 | 6 (cum 22) | M1: underpowered null (n=36/22, 2.5y) - replication NOT met, not refuted. M2 abs: REPLICATES (rand-p<=0.015). | doc ↗ |
| 2026-07-23 | 2026-07-23-displacement-event-scan | research 2017-08-17..2022-12-31 | 6 (cum 16) | M1 signed continuation: STRONG disp4/5 15m-1h, all 4 nulls passed, 5/6 yr stable, plateau, n=517/282. M2 vol elevation: confirmed. | doc ↗ |
| 2026-07-22 | 2026-07-22-fable-mentor-internal-btc | research (execution prereg) | 4 (cum 17) | FAIL -- gold signal INVERTS on BTC. n=165, gross -0.0706R vs coin-flip baseline gross +0.0526R (n=403). H1 REFUTED, H4 REFUTED (spec underperforms coin-flip by -0.123R gross). H2 HELD (drag 0.152R), H3 HELD (no trade <= -5R). Mentor branch 0-for-2 on BTC. | — |
| 2026-07-22 | 2026-07-22-mentor-btc-v1 | research (execution prereg) | 6 (cum 13) | Research-split FAIL. H1 REFUTED decisively (gross -0.0166R at n=11,836, at/below frequency-matched random null -0.0143R). H2 REFUTED (trail did not beat fixed TP). H3 HELD (session filter net +0.024R). H4 confirmed-by-construction only (costs deducted not paid in prices). Sign-stable only 3/6 regime blocks. | — |
| 2026-07-21 | 2026-07-21-fable-mentor-internal-xauusd | research (execution prereg) | 3 (cum 7) | XAUUSD FAIL on criteria 1 & 4 (net -0.0404R, n=43) but ALL 3 registered hypotheses HELD: gross +0.1002R vs coin-flip -0.1306R (H1), drag 0.141R within prediction (H2), no trade <= -5R (H3). First real gross signal evidence in the project. Result filed 2026-07-22. | — |
| 2026-07-21 | 2026-07-21-expansion-v2-costfix | research (execution prereg) | 3 (cum 4) | GBPUSD FAIL (net -0.472R, n=26). H1 CONFIRMED (drag law, 3rd replication, 24/24 arms), H2 REFUTED (catastrophic tail was a Christmas-Eve wide-spread state, not the rollover clock window), H3 CONFIRMED (net improved +2.03R vs v1, threshold +1.0R). | — |
| 2026-07-21 | 2026-07-21-donchian-external-dryrun | research (execution prereg) | 1 (cum 1) | H1 CONFIRMED: stop floor cut drag in 12/12 matched arms (e.g. 0.70R->0.15R). Overall FAIL (net -0.151R, gross -0.045R, GBPUSD) -- no signal to protect, consistent with Neely-Weller-Ulrich decay. | — |