Verdict: Pre-open Gap Board (M-013) is rated “FIT FOR LIMITED USE” in its validation review dated 3 Sep 2026.
Key result: Reversal band alpha vs SPY, 10d: +1.14% (n=8,642).
Key result: Naive gap-up >+15%, same-day O->C: -1.18% (n=1,382) — the design this board avoids.
How it was tested: training data 2018→2026, test data 2018→2026; out-of-sample design (how it was checked on data it never saw while being built): none.
Horizon: same-day open through ~10 sessions for the reversal thesis. Coverage: US names >=$50M/day (~1,500 liquid names).
Stage: Production. Version v1; the hub record is as of 6 Nov 2026.
The verdict
FIT FOR LIMITED USE as context: the reversal-band finding is a genuine, evidenced 2-s.e. result and the fade demotion is well-supported, but the live premarket read has not been validated against the backtest open-gap definition.
Development summary: Built after measuring that the obvious version — buy the biggest gap-ups — loses money on 2.1M name-days; the board flags a real reversal band and demotes gap-ups to fade/avoid, but the premarket-vs-open drift it depends on is unmeasured.
What the review allows
The reversal bucket (-3..-8% gap, trend intact) as a real, 2-s.e.-clearing setup for same-day discretionary review.
Big gap-ups' fade/avoid labelling — the monotonic loss pattern is well evidenced.
gap_atr, not gap_usd, as the read on setup quality.
What the review rules out
Treat the board's live premarket gap as identical to the backtested open gap — the drift is unmeasured.
Auto-trade or size off this board; it has no forward scoreboard of its own.
What it predicts
Item
Detail
Question it answers
Which US names are gapping before the open, and does the evidence say the gap should be faded, read as a reversal candidate, or treated as context — given a naive gap-up-buy list is a measured loser (2.1M name-days)?
Method
Gap/ATR% threshold rules
How it works
Rule-threshold board: reversal band (-8%..-3% gap, above 200-DMA, RS6>=60, ATR%>=4) vs fade (>=+5% gap) vs context; ranked on gap/ATR% (gap_atr), the only version of the gap that predicts anything
Inputs
Premarket gap %, Gap in ATR units, Above 200-day average, 6-month relative strength, ATR% (volatility)
Output
Per-name bucket (reversal / fade / too_deep / in_play) with gap/gap_atr/atr_usd/gap_usd and the session-state flag (thin/live/closed)
Horizon
same-day open through ~10 sessions for the reversal thesis
Coverage
US names >=$50M/day (~1,500 liquid names)
Markets
US
Model type
Model · Rule
Stage
Production
Version
v1, first recorded 15 Aug 2026
Validation evidence
Training data: 2018→2026. Test data: 2018→2026. Out-of-sample design: none.
Headline results from the model's development record, as the hub publishes them.
Measure
Result
Compared with
Note
Gap up >+15%, same-day O->C
-1.18%
+0.01% all-liquid baseline
n=1,382; monotonically worse as the setup looks better beforehand
Reversal band (-3..-8%, >200DMA, RS6>=60, ATR%>=4)
+1.14% alpha/10d
-0.11% all-liquid baseline
n=8,642, +-0.44 (2 s.e.), the one setup that clears 2 s.e. in this platform
ATR% vs ATR$ separation
ATR%<2.5: 3.8% touch / ATR%>6: 51.5%
ATR$<$3: 15.0% / ATR$>$10: 23.4%
percent ATR is the version that predicts; dollar ATR is mostly a price filter
The hub's standard metrics for this model class (only those the record measured).
Metric
Sample / window
Compared with
Value
Range (CI)
Excess return at horizon (headline)
2018–26 · reversal band
all-liquid −0.11
+1.14 pp
[0.7, 1.58] pp
Observations
2018–26 · reversal-band events
n/a
8,642
n/a
What the validation review found
Is the design sound? Strong: the design was explicitly built AFTER measuring the naive version loses, and the board is structured around that finding (leads with reversal, demotes fade) rather than around intuition.
Data quality. Large backtest sample (2.1M name-days) but on the open gap, not the premarket indication the live board reads — see M-1.
Backtest. One regime (2018-2026, US), one board design; the reversal-band result (+1.14%, n=8,642) is the strongest 2-s.e. finding in this platform's research per its own framing.
Monitoring. No forward scoreboard for the live board; dated snapshots exist to build one but have not accrued enough history yet.
Limitations
Backtest-vs-live gap: the historical edge is measured on the OPEN gap; the board reads the PREMARKET indication 15 minutes ahead, and the drift between the two is not yet quantified.
Single regime, one market (US 2018-2026); no held-out test period for this specific board design.
No forward scoreboard exists yet for the live board itself — the backtest is historical, not a tracked live record.
Questions
Is the Pre-open Gap Board validated?
Its validation review dated 3 Sep 2026 rated it FIT FOR LIMITED USE as context: the reversal-band finding is a genuine, evidenced 2-s.e. result and the fade demotion is well-supported, but the live premarket read has not been validated against the backtest open-gap definition.
What does the Pre-open Gap Board predict?
It answers: Which US names are gapping before the open, and does the evidence say the gap should be faded, read as a reversal candidate, or treated as context — given a naive gap-up-buy list is a measured loser (2.1M name-days)? Output: Per-name bucket (reversal / fade / too_deep / in_play) with gap/gap_atr/atr_usd/gap_usd and the session-state flag (thin/live/closed). Horizon: same-day open through ~10 sessions for the reversal thesis. Coverage: US names >=$50M/day (~1,500 liquid names).
How was the Pre-open Gap Board tested?
Training data 2018→2026, test data 2018→2026, out-of-sample design: none. Reversal band alpha vs SPY, 10d: +1.14% (n=8,642); Naive gap-up >+15%, same-day O->C: -1.18% (n=1,382) — the design this board avoids; ATR%<2.5 vs ATR%>6, +10%/10d touch odds: 3.8% vs 51.5%.