🤖 Model performance

AI Model Statistics

Our machine learning model analyses every football match it can find, not just the 3 tips we publish daily. Below you find the full performance data of the AI across thousands of matches. This is how the model learns and improves over time.
5,135Matches Analysed
4,751Settled
43.7%Model Accuracy
2,076Correct Predictions

This is the model's raw hit-rate on the exact 1X2 result across every match it analyses, not the win-rate of our published tips. Our tips combine multiple markets (double chance, over/under, Asian handicap) and win far more often. These figures are cumulative since March 2026 and include earlier model versions; our current model (v4) performs more strongly, see the Ensemble section below. See our tip results. What do these numbers mean?

Outcome Distribution

44%
26%
30%
Home Win 44% (2071)Draw 26% (1248)Away Win 30% (1432)

Over/Under 1.5 Distribution

79%
21%
Over 1.5 79% (3368)Under 1.5 21% (900)

O/U 1.5 Model Accuracy: 77.7% (3317/4268 correct)

Over/Under 2.5 Distribution

56%
44%
Over 2.5 56% (2669)Under 2.5 44% (2082)

O/U Model Accuracy: 55% (2613/4751 correct)

Over/Under 3.5 Distribution

34%
66%
Over 3.5 34% (1454)Under 3.5 66% (2814)

O/U 3.5 Model Accuracy: 63.9% (2728/4268 correct)

Both Teams to Score (BTTS)

57%
43%
BTTS Yes 57% (2449)BTTS No 43% (1819)

BTTS Model Accuracy: 54.8% (2340/4268 correct)

Asian Handicap Results

49%
49%
2%
Home Covered 49% (1797)Away Covered 49% (1800)Push 2% (74)

AH Model Accuracy: 44.3% (1594/3597 correct, 74 push)

1X2 Model Accuracy by Competition

This is the raw 1X2 (home/draw/away) hit-rate of the model. Our published tips also use double chance, over/under and Asian-handicap markets, so their win-rate differs; see the archive for tip results.

Strong (45%+)Moderate (38-45%)Weak (below 38%)Dashed line = 33.3% (random pick)
CompetitionMatchesCorrectAccuracy
Serie A13881
58.7%+25.4
Eredivisie13070
53.8%+20.5
Super League10254
52.9%+19.6
Jupiler Pro League15179
52.3%+19
Ekstraklasa15780
51%+17.7
1. Division18492
50%+16.7
Primeira Liga13768
49.6%+16.3
Ligue 112059
49.2%+15.9
3. Liga12458
46.8%+13.5
Allsvenskan17983
46.4%+13.1
Bundesliga10950
45.9%+12.6
Serie B13159
45%+11.7
La Liga16775
44.9%+11.6
UEFA Europa Conference League271120
44.3%+11
Segunda División19985
42.7%+9.4
Brazilian Série A21390
42.3%+9
Premier League415173
41.7%+8.4
2. Bundesliga12852
40.6%+7.3
Superliga10643
40.6%+7.3
UEFA Champions League11847
39.8%+6.5

Recent AI Predictions

DateMatch1X2Model O/U 1.5O/U 2.5O/U 3.5BTTSAH LineActualResult
11 Oct 2026Primeira LigaArouca vs EstrelaHome Win51%Over 1.5 ❌Over 2.5 ❌Under 3.5 ✅BTTS Yes ❌-0.5 (Home) ✅Home Win (1-0)✅
11 Oct 2026La LigaRacing Santander vs ValenciaHome Win38%—Over 2.5 ✅——+0.0 (Away)Draw (2-2)❌
11 Oct 2026Segunda DivisiónCadiz vs Sporting GijonAway Win46%—Over 2.5 ❌——+0.3 (Away) ✅Away Win (0-1)✅
11 Oct 2026Ligue 1Estac Troyes vs MarseilleAway Win42%—Under 2.5 ✅——+0.3 (Away) ✅Away Win (0-2)✅
11 Oct 2026Serie ACagliari vs JuventusAway Win54%—Over 2.5 ❌——+0.8 (Away) ❌Draw (1-1)❌
11 Oct 2026Jupiler Pro LeagueKV Mechelen vs St. TruidenAway Win40%—Under 2.5 ❌——+0.0 (Away)Home Win (4-3)❌
11 Oct 2026Primeira LigaBenfica vs Vitória SCHome Win76%—Over 2.5 ❌——-1.5 (Home) ✅Home Win (2-0)✅
11 Oct 2026La LigaReal Betis vs OsasunaHome Win62%—Under 2.5 ❌——-1.0 (Home) ✅Home Win (5-2)✅
11 Oct 2026Segunda DivisiónMallorca vs Las PalmasHome Win70%—Under 2.5 ❌——-1.5 (Home) ✅Home Win (3-1)✅
11 Oct 2026Jupiler Pro LeagueKVC Westerlo vs AntwerpHome Win40%—Over 2.5 ✅——+0.0 (Away)Home Win (5-1)✅
11 Oct 2026Serie ASassuolo vs AC MilanAway Win52%—Over 2.5 ❌——+0.5 (Away) ❌Draw (1-1)❌
11 Oct 2026BundesligaSC Freiburg vs FC Schalke 04Home Win49%—Over 2.5 ✅——-0.5 (Home) ✅Home Win (2-1)✅
11 Oct 2026Premier LeagueLiverpool vs Manchester CityAway Win45%—Over 2.5 ✅——+0.3 (Away) ✅Away Win (0-3)✅
11 Oct 2026Ligue 1Rennes vs AuxerreHome Win53%—Over 2.5 ✅——-0.8 (Home) ✅Home Win (4-1)✅
11 Oct 2026EredivisieExcelsior vs GroningenHome Win39%—Over 2.5 ✅——+0.0 (Away)Away Win (1-5)❌

Ensemble Model

43.7%Ensemble Accuracy
4,718Settled Matches

Our v4 GradientBoosting model — the AI behind every tip we publish — currently scores 47.4% accuracy on 2,008 settled matches since deployment. This is in line with the inherent difficulty of 3-way football outcomes (random = 33.3%). Accuracy will stabilise as more data accumulates.

35.9%When All Models Agree
1,570All Models Agreed

Ensemble Score Distribution & Accuracy

Score RangeMatchesAccuracy
0-9364
30.8%
10-191256
37.8%
20-29789
40.6%
30-39714
43.7%
40-49692
44.1%
50-59468
56.4%
60-69285
58.2%
70-79116
71.6%
80-8929
72.4%
90-994
25%
100-1091
100%

Model Calibration

Does a higher confidence score actually mean a more accurate prediction? This chart answers that. We group every settled prediction by the internal confidence score the model assigned, then show the real win-rate for each group. A rising staircase is what you want to see, it means the model knows when it is more likely to be right.

Each bar groups model predictions by the model’s internal ranking score (0–100) and shows how often those predictions were correct on settled matches. A rising pattern means the ranking is informative. It does not mean the model beats the bookmaker market: in our own testing it did not. Hover for sample sizes.

Note, 17 September 2026: this chart shows the model’s internal ranking score. Whether its probabilities beat the bookmaker market is a different question, and in our own testing they did not. See our methodology update for the full measurement.

Model accuracy is calculated on settled matches only. The current model uses form, goals average and H2H data as input features. Accuracy improves as more live data is collected.

More transparency: Tip results & track record · How accurate are AI predictions?