How our football model did: 1,922 fixtures graded
Public Beathem model record, v2.3.0: 1,922 fixtures (12 Oct–10 Nov), hit rates, calibration, last weekend, and this international-break value board.
Tipster accounts usually lead with winners. We would rather start with a public card. Between 12 October and 10 November 2026, Beathem model v2.3.0 has been graded on 1,922 finished fixtures. The prediction track record shows a 41.9% 1X2 top-pick hit rate and 55.0% on the Over/Under 2.5 side.
Those figures are not a promise that next month will look the same. This post explains what they mean, how last weekend (7–8 November) compared with that wider window, and where the model currently disagrees with the market on this international-break board. Every figure was taken from the live site on Wednesday 11 November 2026. Odds move, so recheck the boards before kick-off.
The headline record
The public window is 12 October to 10 November 2026, model version v2.3.0 — the highest-volume version in that range. Across those 1,922 graded fixtures:
- 1X2 top-pick hit rate: 41.9%. After full time, did the result (home, draw or away) match the side the model had given the highest probability?
- Over/Under 2.5 side hit rate: 55.0%. The model takes Over if its Over 2.5 probability is at least 50%, otherwise Under, and we check whether that side matched the goals total.
- Mean Brier scores: 0.645 (1X2) and 0.244 (Over/Under 2.5). Brier punishes a confident miss more than a cautious one. Lower is better.
These are research probabilities, not tipster guarantees, and past performance does not guarantee future results.
What a hit rate actually means
A hit rate is simply "how often were we right?", counted after the match is finished. It is not profit, and it is not ROI.
A 41.9% 1X2 top-pick mark can look soft if you are used to two-way markets, but 1X2 is a three-way problem. In a typical match the model's favourite is often only in the 40s or low 50s, so the top pick will be wrong more often than it is right. That is expected. It is also not the same as "41.9% of bets won": the record grades every finished fixture the model priced, at every kind of price, not a filtered slip.
Hit rate alone is not enough. A model can look sharp and still be a poor basis for staking if the prices are no better than the market — which is why we also publish calibration and the value board. If you place a bet elsewhere, the only record that matters for you is your own, settled in full in the bet journal.
The Over/Under 2.5 figure is easier to read: it is a two-way market with a 50% decision rule. 55.0% means the side the model leaned toward matched the goals total in just over half of those 1,922 fixtures.
Calibration: when the model says 50–60%, how often does it land?
Hit rate tells you how often the top pick was right. Calibration asks a fairer question: when the model quoted a probability in a given band, did results land about that often? The current record grades the top 1X2 probability in three bands:
- 50–60%: 128 graded fixtures, 83 hits, a 64.8% hit rate.
- 60–70%: 9 graded fixtures, 8 hits, an 88.9% hit rate.
- 70%+: no fixtures in this window, so there is nothing to report.
A well-calibrated 50–60% call should land around five or six times in ten. 64.8% on 128 matches sits a little above that band, which can happen in a few hundred samples — it is not a licence to treat a 55% line as a 65% one. The 60–70% band looks stronger at 88.9%, but nine fixtures is a tiny sample. We would rather show the thin band than hide it.
The methodology explains how the probabilities are built; the record is how they are kept honest afterwards.
Last weekend: 7–8 November
The record page lists days, not weekends, so these figures are the Saturday and Sunday rows added together. On 7–8 November the model was graded on 318 fixtures. The 1X2 top pick matched the result in 133 of 318 (41.8%). The Over/Under 2.5 side was right in 182 of 318 (57.2%).
- Saturday 7 November: 75 / 171 1X2 hits (43.9%), Over/Under 2.5 at 59.6%.
- Sunday 8 November: 58 / 147 1X2 hits (39.5%), Over/Under 2.5 at 54.4%.
The 1X2 rate landed almost exactly on the 41.9% window average. Over/Under 2.5 ran a little hotter than the 55.0% average, especially on Saturday. That is variance, not a new normal.
What a value edge (pp) is — and is not
A value edge is the gap, in percentage points (pp), between the model's probability and the market's implied probability after the bookmaker margin has been removed from the prices. If the model has a draw at 43% and the de-vigged market implies about 26%, the edge is roughly +17pp. The match page shows that gap as a Value badge when it clears a set threshold, with a confidence score.
It is a disagreement signal, not a tip. Markets can be right, models can be wrong, and an edge can evaporate when the odds move. We do not publish bookmaker names or betting links.
A worked example from Saturday's board, pulled this afternoon: Sport Recife vs Goias, Brazil Serie B. The model has the draw at 43%, the market draw price is 3.59, and the listed edge is +17pp. Once you take the 1X2 overround out of the three prices (3.49 / 3.59 / 1.93), the de-vigged draw implied probability is about 26%, and 43 minus 26 is that +17pp gap. If the price shortens overnight, the edge changes with it.
This international-break weekend
With the top European leagues quiet for internationals, this weekend's boards mix UEFA Under-21 qualifiers, friendlies, Africa Cup of Nations qualifiers, and the domestic leagues still playing — Spain's Segunda División, Brazil, Uruguay and the Dutch Eerste Divisie. Friday 13 November currently lists 35 fixtures; Saturday 14 November lists 56.
These were the larger model-versus-market gaps on 11 November.
Friday 13 November
- Finland U21 vs Republic of Ireland U21 (UEFA U21 Championship Qualifying) — Over 2.5 at 65%, edge +19pp.
- Atletico Torque vs Racing Montevideo (Uruguay, Primera División) — Draw at 39% (odds 3.41), edge +12pp.
- Turkey U21 vs Czech Republic U21 (UEFA U21 Championship Qualifying) — Draw at 39% (odds 3.13), edge +10pp.
Saturday 14 November
- Sport Recife vs Goias (Brazil, Serie B) — Draw at 43% (odds 3.59), edge +17pp.
- America Mineiro vs Avai (Brazil, Serie B) — America Mineiro at 41% (odds 3.84), edge +17pp.
- Mallorca vs Cordoba (Spain, Segunda División) — Draw at 37% (odds 3.95), edge +14pp.
The full, ranked lists live on the Friday board, the Saturday board, and the rolling value picks today page. Today's Wednesday board had no value edges that cleared the threshold.
If you follow a match this weekend, log the bet in the journal before kick-off. Eligible 1X2, Over/Under 2.5 and both-teams-to-score bets settle automatically when full-time results are in, so you can review your own ROI by market, league and odds range. Beathem is free to start, and the Pro waitlist is open.
Research only, not betting advice. 18+. Please gamble responsibly.