Four correct markets; one decisive tail event
The five 24 October football forecasts were frozen as research only. Arsenal team under 2.5, Villa–City under 4.5, Dortmund–Frankfurt under 3.5 and Leipzig–Elversberg over 2.5 won; Tottenham team under 1.5 lost in Chelsea 0–2 Tottenham. The A+B research Doubler therefore lost. No item is added to the official Track Record.
Chelsea 0–2 Tottenham
Facts: an own goal and an 80th-minute Chelsea red card preceded the second goal; xG was low. Hypothesis: team-total tail risk from set pieces, cards and keeper errors was underweighted. Test: prospective tail overlays—not one-match retuning.
Leipzig 4–0 Elversberg
Facts: eight starting changes and four second-half goals. Lesson: the winning over still carried rotation and late-game tail risk; outcome alone does not validate the forecast.
Villa 2–1 Manchester City
Facts: City led xG and shots on target, yet the under won. Lesson: correct total and misleading 1X2 game state can coexist; market-level attribution matters.
Descriptive research metrics only: Brier 0.1667; LogLoss 0.5119. Equal-unit arithmetic at the archived reference prices was +0.84u (+16.8% on five hypothetical units), but those prices were not proven Greece-executable and this is not realized ROI or official performance. Open the full immutable audit →
Three provisional markets; zero certified bets
Liverpool–Brighton · Over 1.5 goals
Poisson diagnostic p 83.1% · conservative 75.0%
Reference 1.17 · break-even 85.5% · robust EV −12.3% · VS 66/100.
Highest estimated hit probability among the price-linked candidates, but the price demands more than the conservative estimate. Storm, injuries and a simple uncalibrated baseline prevent release.
Crystal Palace–Newcastle · Over 2.5 goals
Poisson diagnostic p 68.8% · conservative 61.0%
Reference 1.63 · break-even 61.3% · robust EV −0.6% · VS 62/100.
Best price/probability balance, but even a small uncertainty haircut removes the apparent edge. It remains a research watch, not a recommendation.
Safest price-qualified combination: two legs, not three
Liverpool–Brighton Over 2.5 @1.57 × Palace–Newcastle Over 2.5 @1.63 · combined reference 2.559 · independence-scenario p 42.9% · conservative joint p 32.9% · break-even 39.1% · robust EV −15.7%.
The 3-leg higher-hit-rate-market alternative (Liverpool O1.5 × Man United O1.5 × Palace O2.5) reaches 2.289 but only 32.0% conservative joint probability and −26.7% robust EV. Four- and five-leg versions are inadmissible: there are not enough separately modelled, price-linked, distinct-event legs. No Builder is shown because there is no validated same-game joint simulation or Builder Tax.
Three admitted research watches; two places left empty
| # | Event / optimized market | p / conservative | Reference | Validation Score |
|---|---|---|---|---|
| 1 | Liverpool–Brighton · Over 1.5 16:00 local · last five: LIV W-D-W-L-W; BHA D-W-W-W-W · major absences both sides; wind/rain risk | 83.1% / 75.0% | 1.17 · public comparison · checked 08:18 UTC · not Greece-verified | 66 Data 19 · Lineup 13 · Model 11 · Market 10 · Form 6 · Uncertainty 7 |
| 2 | Man United–Bournemouth · Over 1.5 16:00 local · last five: United D-D-L-D-L; Bournemouth L-L-W-W-W · multiple United absences | 80.1% / 70.0% | 1.20 · public comparison · checked 08:18 UTC · not Greece-verified | 65 Data 18 · Lineup 12 · Model 11 · Market 12 · Form 6 · Uncertainty 6 |
| 3 | Crystal Palace–Newcastle · Over 2.5 16:00 local · last five scorelines and venue split reviewed · availability doubts remain | 68.8% / 61.0% | 1.63 · public comparison · checked 08:18 UTC · not Greece-verified | 62 Data 18 · Lineup 11 · Model 11 · Market 10 · Form 6 · Uncertainty 6 |
| 4 | Unfilled: no fourth distinct event cleared VS>60 with an independent numeric diagnostic and a matched eligible reference price. | |||
| 5 | Unfilled: no padding. Hull–Brentford and Sunderland–Leeds remain rejection/monitoring cases. | |||
The diagnostic is a transparent independent-Poisson calculation from currently reported scoring/conceding rates, not AIONORA’s approved OOS model. Last-ten evidence is unavailable because the league sample is only seven to nine games and the cross-competition join is incomplete; the Form component is capped accordingly. No p_model is certified.
Broad discovery, narrow evidence join
| Sport / scope | Discovered | Screened | Numeric model | Price-linked | Release-qualified |
|---|---|---|---|---|---|
| Football · official PL/Bundesliga + selected Serie A/La Liga | 13 explicitly verified fixtures | 7 collector-tracked + 6 manual schedule checks | 38 research-only 1X2 rows in source pipeline; 3 independent totals diagnostics here | 9 reference joins in pipeline; 3 editorial matches | 0 certified |
| NBA | 7 official games | schedule/status only | 0 rights-cleared pace/rotation models | 0 | 0 |
| EuroLeague / EuroCup | 0 senior games confirmed in today’s sampled window | calendar check | 0 | 0 | 0 |
| ATP / WTA | Basel/Vienna qualifying and multiple tour finals identified | tournament level only | 0 surface serve/return models | 0 rights-cleared executable joins | 0 |
Football market variants compared where inputs existed: 1X2, DC, DNB, BTTS, Over/Under 1.5/2.5/3.5 and team totals; no Asian Handicap. The current source refresh at latest readback reports 1,372 rows, 772 final results, 600 open cases, 38 research model rows and 9 reference-price joins. None is a verified Greece-executable quote.
Five evidence-led passes
- Hull–Brentford · Brentford 1X2 @1.65: public model references disagree sharply and the independent current-form/venue model is incomplete; favourite status is not safety.
- Liverpool–Brighton · Liverpool 1X2 @2.15: Brighton’s 5–0–0 aggregate across five league matches, Liverpool absences and adverse weather create material uncertainty.
- Man United–Bournemouth · BTTS @1.50: raw Poisson is about 59.9%, below the 66.7% break-even even before uncertainty stress.
- Barcelona–Real Madrid result/corners combinations: no joined current lineups, independent calibrated model or Greece-executable quote; headline fixture is not an evidence shortcut.
- NBA or tennis props/accumulators: no rights-cleared current rotation/minutes or surface serve-return pipeline and no executable price verification.
A passed market may later win. That would be a false-negative research observation, not a retroactive bet or profit claim.
Improvement not yet measurable
History
5,161 source-reported results in the preferred 15-file archive; the separate current four-file import contains 772 completed results. These overlap and are never summed.
Prospective evidence
The current pipeline still reports zero matchable evaluation events. One 4/5 day cannot establish calibration or justify a coefficient change.
Release integrity
0 official certified selections. Private writer initialization and website sync remain unverified; research is not backfilled.
Completed today
Canonical five-event audit; outcome cross-check; one loss decomposition; Brier/LogLoss labelled descriptive; market-variant stress; all 2–3 leg price-qualified combinations enumerated; 4–5 leg infeasibility documented; source/API readbacks repeated; no model weights changed.
Tests registered, not executed as model changes
Prospective red-card/own-goal/set-piece team-total tail overlay; rotation × rest interaction; closing-line capture only after an executable quote source exists; Poisson versus Dixon–Coles comparison only on untouched future cohorts.
Open gates: approved market-specific OOS model passport; immutable event/market identifiers for every Morning and private PRO row; confirmed critical lineups; current Greece-executable operator prices; basketball and tennis calibration/data rights; positive robust EV; verified private CORE writer readback.
Research and official record stay separate
Readbacks: official Track Record · Historical Data. Source scores are reported by the ingestion sources; they are not independent official score confirmation. Research worksheets never mutate the official record.