Football ratings

Thursday 10 September 2026 · times UK

The model's strongest calls

Nothing clears the bar today. The thresholds are set where the model has historically been as good as it claims — home win 70%, away win 60% — and a quiet day is a real answer, not a gap.

Every fixture

League One

Stevenage v Luton20:00

	Home	Draw	Away
Model	41	29	31
Market	31	27	42

	Over 2.5	Under 2.5
Model	51	49
Market	48	52

Under 4.5 goals: 88% · happens 86% of the time anyway

Model is 10 points above the market on home win (pinnacle). Worth a look at why.

Primeira Liga

Estrela v SC Braga20:15

	Home	Draw	Away
Model	19	27	54
Market	19	25	56

	Over 2.5	Under 2.5
Model	47	53
Market	49	51

Under 4.5 goals: 89% · happens 86% of the time anyway

UEFA Champions League

New competition — experimental. The model has never been trained or tested on it, these clubs are rated across leagues whose ratings have never been calibrated against each other, and the probabilities use a fallback calibration. Treat them as much weaker than the domestic ones.

Fenerbahçe v AS Roma17:45

	Home	Draw	Away
Model	42	28	30
Market	28	27	45

	Over 2.5	Under 2.5
Model	53	47
Market	55	45

Under 4.5 goals: 85% · happens 86% of the time anyway

Model is 14 points above the market on home win (pinnacle). Worth a look at why.

PSV Eindhoven v Shakhtar Donetsk17:45

Not rated — Shakhtar Donetsk has 0 prior matches in the store, most likely promoted from a division we do not collect. A rating built on that would be a guess wearing a percentage.

Bayern München v Bodo/Glimt20:00

Not rated — Bodo/Glimt has 0 prior matches in the store, most likely promoted from a division we do not collect. A rating built on that would be a guess wearing a percentage.

Como v RB Leipzig20:00

	Home	Draw	Away
Model	55	24	21
Market	53	24	23

	Over 2.5	Under 2.5
Model	60	40
Market	61	39

Under 4.5 goals: 80% · happens 86% of the time anyway

Manchester United v Sabah FA20:00

Not rated — Sabah FA has 0 prior matches in the store, most likely promoted from a division we do not collect. A rating built on that would be a guess wearing a percentage.

Slavia Praha v Lens20:00

Not rated — Slavia Praha has 0 prior matches in the store, most likely promoted from a division we do not collect. A rating built on that would be a guess wearing a percentage.

How the calls did

✓ST Johnstone 0–1 CelticCeltic away win 75%

✓Twente 1–0 TelstarTwente home win 74%

✓Barcelona 5–1 Feyenoord over 2.5 goals 74%

✕ST Johnstone 0–1 Celtic over 2.5 goals 68%

✕Twente 1–0 Telstar over 2.5 goals 68%

✓Moreirense 0–4 BenficaBenfica away win 67%

✓Liverpool 2–1 Atletico Madrid over 2.5 goals 62%

✓Napoli 0–1 Arsenal under 2.5 goals 57%

6 of 8 settled since yesterday · 158 of 233 since the model went live.

Every match, for the record

The model rates the whole card, not just the calls, and learns from all of it. Judged as probabilities rather than picks — a 38% favourite losing is the model working, not failing.

·Resulttop pick right 9/11, we claimed 56% · 226/453 all-time

what actually happenedwe gave it 54% on average

·Over/under 2.5top pick right 6/11, we claimed 59% · 264/453 all-time

what actually happenedwe gave it 51% on average

A well-calibrated model's average claim matches how often it happens. That is the number Rule 9 gates on, and it is tracked for results and goals alike whether or not a fixture ever produced a call.

What these numbers are. The model's own estimate of how likely each result is, from form, Elo, rest, shots and xG as they stood at 06:00 UK on match day. One bivariate Poisson produces all of them, so the result and the goals figures share one underlying view of the match. Each published figure is then calibrated against what actually happened in past matches, one market at a time — so they agree closely but are not forced to agree exactly. Market figures are the bookmaker's prices with their margin removed, so the two are comparable.

What they are not. Across 57,633 out-of-sample fixtures the market separated matches better than the model in every league, every season, every outcome and every confidence band — 24 slices, no exceptions, re-tested in August 2026 on the current model. This is a second opinion, not an edge. Where it earns its keep is honesty about its own confidence: when it says 74%, that happens about 74% of the time.

Why a call appears. Thresholds — home 70% · away 60% · over 60% · under 55% — sit where the model has historically been as good as its claim. Raising them would surface fewer calls and be less accurate about them, because its confidence above those levels is inflated everywhere except home wins. Draws are never flagged: the highest one ever rated was 62%.

4 rated · 4 not rated · model goals_v7_cal · built 2026-09-10 07:05 UK