Verdict: Earnings-Move and IV-Crush Model (US) (M-026) is rated “FIT WITH CONDITIONS” in its validation review dated 19 Oct 2026.
Key result: 2025-26 MAE vs k x IV30 move: 3.8% lower, BH p<0.001.
Key result: 2024-26 MAE vs name 8-event median: 3.4% lower.
How it was tested: the record does not state separate training and test periods.
Horizon: one event: close on T-1 -> close on the first session able to price the release. Coverage: US universe names with primary EDGAR 8-K 2.02 releases of known session; C(T-1) >= $5; median 63-session dollar volume >= $2M; survivors only.
Stage: Validated (shadow output). Version v1; the hub record is as of 6 Nov 2026.
The verdict
FIT WITH CONDITIONS as a level forecast of the absolute reaction move and as descriptive crush bands; NOT a validated cheap/rich signal against traded straddles.
Development summary: Registered rule -> MOVE FORECAST, CRUSH DESCRIPTIVE. The stage-A champion A3 (gradient-boosting quantile on name history, chosen on 2021-23) beats the name's 8-event median by 3.4% MAE and the sector x size base rate by 5.1% (2024-26, n=22,044), and every IV30-implied benchmark in 2025-26 (k x IV30 move 3.8%, IV-excess 9.3%, post-hoc trading-day IV30 2.8%). The IV stack S1 passes the same gate but adds nothing over A3. Crush: the registered OLS (KO) does not beat the name's median past crush (K1); the challenger KG does by 17.1%.
What the review allows
exp_move_model_pct and its q25-q75 band as the expected size of the reaction move (2024-26 coverage 0.47, share above q50 0.53)
crush_ratio_expected as a descriptive typical crush for the name (K1)
ratio_model_vs_implied as a SCREEN for names worth pricing by hand
What the review rules out
Treat ratio_model_vs_implied as a measured trading edge: the benchmark is an IV30 proxy and most of its ranking power is mechanical (IC 0.22 vs 0.17 for the name's median alone)
Use the output for direction - it forecasts the size of the move only
Read rows with confidence=low (few releases, unknown session, illiquid) as forecasts
What it predicts
Item
Detail
Question it answers
Forecast each upcoming US earnings release's absolute reaction-day move and IV30 crush vs the implied move
Method
GBM quantile (A3) vs OLS (A1) vs IV stack (S1)
How it works
Gradient-boosting quantile bands; OLS and IV-stack rivals
Inputs
Name's past 8 reaction moves (median, dispersion, last), Normal-day volatility (63d, event days masked), Size, liquidity, sector, release session, Days since last report, pre-event 21d move, IV30 level, percentile, IV/HV, run-up (IV stack and crush only)
Output
Per upcoming report: expected |move| % (q50) with q25-q75 band, implied move, IV crush ratio
Horizon
one event: close on T-1 -> close on the first session able to price the release
Coverage
US universe names with primary EDGAR 8-K 2.02 releases of known session; C(T-1) >= $5; median 63-session dollar volume >= $2M; survivors only
Markets
US
Model type
Model · ML/Stat
Stage
Validated (shadow output)
Version
v1, first recorded 16 Oct 2026
Validation evidence
Headline results from the model's development record, as the hub publishes them.
Measure
Result
Compared with
Note
MAE vs k x IV30 move (2025-26)
3.57 pp
3.72 pp
3.8% lower (95% CI 2.5 to 5.4%, BH p 0)
MAE vs IV-excess event move
3.57 pp
3.94 pp
9.3% lower (95% CI 6.9 to 11.8%, BH p 0)
MAE vs name 8-event median (2024-26)
4.34 pp
4.49 pp
3.4% lower (95% CI 2.6 to 4.2%, BH p 0)
MAE vs sector x size rate (2024-26)
4.36 pp
4.60 pp
5.1% lower (95% CI 4.6 to 5.6%, BH p 0)
Crush MAE (ratio)
K1 0.0783
KO 0.0753, KG 0.0649
KO vs K1 BH p 0.098 (fails); KG vs K1 17.1% lower (95% CI 13.1 to 20.8%, BH p 0)
Limitations
IV benchmarks are IV30 proxies; no straddle history exists to test a traded edge.
Survivor universe and today's sector/share count (stated look-aheads).
Questions
Is the Earnings-Move and IV-Crush Model (US) validated?
Its validation review dated 19 Oct 2026 rated it FIT WITH CONDITIONS as a level forecast of the absolute reaction move and as descriptive crush bands; NOT a validated cheap/rich signal against traded straddles.
What does the Earnings-Move and IV-Crush Model (US) predict?
It answers: Forecast each upcoming US earnings release's absolute reaction-day move and IV30 crush vs the implied move. Output: Per upcoming report: expected |move| % (q50) with q25-q75 band, implied move, IV crush ratio. Horizon: one event: close on T-1 -> close on the first session able to price the release. Coverage: US universe names with primary EDGAR 8-K 2.02 releases of known session; C(T-1) >= $5; median 63-session dollar volume >= $2M; survivors only.
How was the Earnings-Move and IV-Crush Model (US) tested?
2025-26 MAE vs k x IV30 move: 3.8% lower, BH p<0.001; 2024-26 MAE vs name 8-event median: 3.4% lower; [q25,q75] coverage 2024-26: 0.47 (target 0.50).