24 October research settlement
Research-only retrospective. Official certified selections remain at zero; no verified book execution or CLV.
Cases and concrete tests
Own goal, dismissal and late second Tottenham goal: measure card/set-piece tail sensitivity prospectively.
Eight lineup changes; four goals after half-time. Test rotation and rest effects.
Even a winning total can mask poor team-strength attribution. Evaluate 1X2 and totals separately.
These are tracked hypotheses and procedural safeguards, not proved causal effects. Model coefficients changed: zero.
Against historical prior probabilities
| Market | Poisson Brier | Baseline Brier |
|---|---|---|
| 1X2 | 0.6419 | 0.6702 |
| BTTS | 0.2502 | 0.2455 |
| Over 2.5 | 0.2585 | 0.2479 |
Lower is better. BTTS and totals are worse for this model; no live certification has been established.
Probability before return
Today's most promising quoted two-leg research scenario had reference odds 2.559, conservative joint p 32.9%, robust EV −15.7%; therefore no release. Historical market rejections remain in an immutable learning ledger.
What the data can and cannot show
Current import: 1,372 rows, 772 reported final scores, 600 open cases. Forward research: 38 rows and 9 reference-price joins; zero verified Greece-executable quotes. Preferred historical archive: 5,161 results across 15 files (overlaps separate datasets).
Official Track Record: zero certified selections. Private website CORE writer/readback not confirmed; earlier dated issues preserved.
Return to Daily COREDisagreement is evidence, not an inconvenience
The morning independent goal-rate baseline was challenged with publicly projected xG rates. Both model variants are uncalibrated; observed disagreements prevent us from preserving an overconfident conservative probability without a new approved assessment.
| Match | Market | Morning raw | New xG raw | Action |
|---|---|---|---|---|
| Liverpool–Brighton | Over 1.5 | 83.1% | 81.4% | Research watch |
| United–Bournemouth | Over 1.5 | 80.1% | 81.4% | Research watch |
| Palace–Newcastle | Over 2.5 | 68.8% | 57.9% | DOWNGRADE |
The three morning six-part Validation Scores are 66/100, 65/100, 62/100 respectively; no new score was certified after the challenger. The pre-match CORE Doubler release remains unqualified. The research snapshot and prior editorial audit are separate immutable versions; the official betting Track Record remains unchanged.