27 OCTOBER 2026 · RESEARCH UPDATE
Evidence first. Learning without hindsight.
AIONORA 5.5.5 is a partial research audit, not a validated model upgrade or a new certified betting record.
What we learned
Two Under 4.5 research forecasts missed on 26 October: Rizespor–Trabzonspor finished 3–4, and Walsall–Salford City finished 2–4. These results raise an important question about how accurately our models capture high-scoring outcomes and match volatility. We will evaluate alternative goal-distribution models against historical and prospectively recorded data, while reviewing the original pre-match evidence wherever available. No model improvement has yet been validated. Our research results remain separate from the official Track Record.
Keep cohorts separate
| Cohort | Research outcomes |
|---|---|
| 26 October CORE | 5 of 5 favourable |
| 26 October owner-facing PRO | 6 of 8 favourable |
| 25 October external PRO-related | 8 of 10 favourable |
| Authenticated PRO Final Call | Not established |
All are separate research categories. None of these retrospective results is automatically a certified selection. Personal bet-slip information and financial details are not published.
What is being tested
The 122 definitions consist of 18 shared, 35 football, 37 basketball and 32 tennis variables. Inclusion in a dictionary does not prove actual model use, correct time alignment, or measurable out-of-sample benefit. We will test baseline Poisson, Dixon–Coles and supported higher-dispersion football models on chronological holdouts, with Brier score, log loss, calibration and high-scoring tail diagnostics.
A prior, separately archived score-only 1,432-match football holdout reported Under 4.5 Brier 0.12488 (Poisson) against 0.12535 (historical-rate baseline) and Over 1.5 Brier 0.15817 against 0.15686 (worse than baseline). These are historical diagnostics, not newly reproduced AIONORA 5.5.5 results.
History and validation status
Authenticated forecast-to-history readback and independent prospective calibration are not yet confirmed. We have not changed model coefficients, converted past research successes to official records, or declared a model improvement. A release requires genuine prematch snapshots, independently verified results, qualified prices and authenticated persistent readback.
Independent statistical sports research. No predictions are guarantees. For adults 21+. Betting carries financial risk.