Outcome ≠ validated prediction
Ten reconstructed football markets covered eight distinct fixtures; correlated markets on Torino–Monza and Go Ahead–Telstar are not independent trials. Original immutable pre-match freezing, executable quotes and stakes are not verified. Official Track Record: 0 selections / 0 settled.
Vitória 5–0 Marítimo
Fact: U3.5 lost; free kick and own goal featured. Hypothesis: defensive weakness and extreme-score tail underweighted. Test: held-out tail calibration.
Stuttgart 2–3 Gladbach
Fact: six Stuttgart changes after Europe. Hypothesis: rotation × short-rest volatility underweighted. Test: prospective interaction study.
Bolton 1–5 Middlesbrough
Fact: BTTS won while the rejected away win also occurred. Lesson: a winning market may conceal opponent-strength and market-selection error.
Illustrative only: Brier 0.2834, LogLoss 0.7792, equal-unit reference arithmetic −1.52u. These are not official calibration, ROI or realized returns. No model weight was changed.
Open every recorded market, combination and forensic case →Quantified, partial and source-bounded
| Sport / scope | Discovered | Screened | Modelled | Price joined | Qualified |
|---|---|---|---|---|---|
| Football collector · EPL/Bundesliga | 42 research candidates | 42 source-level | 42 research-only 1X2 | 0 current joins | 0 |
| Official Saturday fixtures · PL/Bundesliga | 11 (5 + 6) | schedule/status | 0 release models | 0 Greece-executable | 0 |
| NBA · US date 24 Oct | 8 games | schedule only | 0 | 0 | 0 |
| Senior EuroLeague/EuroCup | 0 confirmed today | Round 7 ended 23 Oct | 0 | 0 | 0 |
| ATP/WTA accessible boards | Basel/Vienna qualifying; Ningbo SF; Tokyo qualifying | tournament level | 0 | 0 | 0 |
The collector also froze 10 source-tracked football fixtures for 24 October with immutable readback. Coverage does not equal validation. Serie A and other domestic/global competitions were spot-checked but not fully enumerated; basketball/tennis commercial-use feeds, last-5/last-10 joins and player-status pipelines remain absent. Therefore this is a PARTIAL SCAN, not a claim of worldwide completion.
Verified fixture windows
Premier League: Villa–City; Arsenal–Everton; Coventry–Fulham; Liverpool–Brighton; Chelsea–Tottenham. Bundesliga: Leipzig–Elversberg; Augsburg–Union; Mainz–Bremen; Köln–Schalke; Paderborn–Hamburg; Dortmund–Frankfurt.
Data status at 06:51 UTC
1,372 source rows; 772 final results; 600 open cases; 4/4 datasets; 0 duplicates/failures. Reference odds pipeline: 390 lines, but 0 current joined or Greece-executable quotes.
Evidence gate: no admission
Best Overall
No candidate has the required last-5/last-10 evidence join, lineup gate, approved calibrated model, current executable quote and robust EV.
Balanced CORE
1X2, DC, DNB, BTTS, totals and team totals cannot be ranked honestly without independent market probabilities and matched prices.
CORE Top 5
The list is intentionally empty. No numerical VS is displayed because the six component inputs are incomplete.
Doubler / Bet Builder
No separately qualified legs, adjusted joint probability, correlation simulation or Builder Tax. A 2.00+ price is not sufficient.
The public collector’s 4.22.2 model is unapproved, limited to EPL/Bundesliga 1X2 and has zero prospective evaluation events. Private PRO Top 10 remains private; no qualifying list was manufactured.
Five evidence-based rejections
- Arsenal–Everton favourite shortcut: no independent current market probability, confirmed lineup or executable Greek quote.
- Leipzig–Elversberg straight win: probable lineups show material absences/doubts; last-10 opponent-adjusted model and price join are missing.
- Villa–Manchester City goal total: no dated form/xG distribution or lineup-sensitive tail simulation.
- NBA opening-week props: no licensed rotation/minutes, pace/efficiency or player-market calibration pipeline.
- Tennis qualifying favourite / accumulator: no surface-specific serve-return model, verified fitness status or rights-cleared executable prices.
These are process rejections, not predictions that the teams or markets will lose. No Asian Handicap was considered.
Improvement not yet measurable
The historical source refresh completed and the 24 October research freeze was read back. However, zero unique prospective events are eligible for Poisson/Dixon–Coles evaluation, so Brier, LogLoss, reliability, ROI and CLV remain unavailable for the production model. Poisson-versus-Dixon–Coles remains a challenger test, not a promoted change.
Open gates: approved OOS model passport; match-level immutable Morning/PRO event IDs; current Greece-executable prices; lineup/player evidence; basketball and tennis data rights/calibration; private CORE writer initialization; verified website-sync status.