AI Football Prediction Accuracy and Results
Every tip our AI publishes is checked against the final score. Our most accurate market over the last 30 days is over 1.5 goals: 82% correct (7 014 tips). Positive-ROI system: AI tips on a home win · odds 1.50–2.50, average 1.76 · AI confidence ≥ 60% → +18.4% ROI (43 tips, last 30 days).
Period: 30 days (02.10.2026–31.10.2026)
Positive-ROI system
+18.4% ROI in 24 days
AI tips on a home win · odds 1.50–2.50, average 1.76 · AI confidence ≥ 60%
Period: last 30 days (02.10.2026–31.10.2026), including tips made before the model update · 02.10.2026–25.10.2026
How we calculate (historical simulation): flat stake of 1 unit per tip, no tax, average odds from several bookmakers recorded before kick-off. Profit is shown in units; ROI = profit ÷ units staked; max drawdown = the largest drop from a profit peak. The rule is re-selected every day from 840 combinations (market × odds range × AI confidence) with at least 40 tips. Past results do not guarantee future results. Odds below 1.20 are left out; rules that were in profit in both halves of the period are preferred.
Where AI tips were profitable: odds and confidence bands
We split every settled AI tip with pre-match odds by tip type, odds band and AI confidence. Only bands in profit on at least 40 tips are listed, best ROI first. Odds for Over/Under and both-teams-to-score have been recorded since 12.09.2026, so these markets have a shorter history.
Since 15.09.2026 – updated model
| AI tip | Band | Tips | Hit rate (break-even) | Avg. odds | Profit (units) | ROI | Signal |
|---|---|---|---|---|---|---|---|
| Under 3.5 goals | odds 2.50–3.20 | 65 | 46.2% (36.9%) | 2.72 | +15.84 | +24.4% | promising |
| Home win (1) | AI confidence 60–70% | 123 | 74.0% (69.4%) | 1.47 | +9.03 | +7.3% | promising |
| Away win (2) | AI confidence 50–60% | 279 | 60.2% (57.9%) | 1.79 | +12.92 | +4.6% | could be luck |
| Home win (1) | odds 1.20–1.40 | 274 | 78.5% (76.7%) | 1.31 | +6.59 | +2.4% | could be luck |
| Home win (1) | odds 1.40–1.60 | 381 | 67.5% (66.8%) | 1.50 | +2.62 | +0.7% | could be luck |
Last 30 days (02.10.2026–31.10.2026)
| AI tip | Band | Tips | Hit rate (break-even) | Avg. odds | Profit (units) | ROI | Signal |
|---|---|---|---|---|---|---|---|
| Both teams to score – no | odds 2.00–2.50 | 111 | 53.2% (46.1%) | 2.18 | +18.29 | +16.5% | could be luck |
| Both teams to score – no | AI confidence 40–50% | 102 | 55.9% (51.6%) | 1.99 | +9.61 | +9.4% | could be luck |
| Home win (1) | odds 2.50–3.20 | 61 | 42.6% (39.1%) | 2.56 | +5.46 | +9.0% | could be luck |
| Home win (1) | AI confidence 60–70% | 91 | 71.4% (67.6%) | 1.53 | +8.04 | +8.8% | could be luck |
| Away win (2) | AI confidence 50–60% | 105 | 64.8% (60.5%) | 1.69 | +7.85 | +7.5% | could be luck |
| Away win (2) | odds 1.40–1.60 | 82 | 70.7% (66.6%) | 1.50 | +4.85 | +5.9% | could be luck |
| Home win (1) | odds 1.20–1.40 | 173 | 79.8% (76.6%) | 1.31 | +7.46 | +4.3% | could be luck |
| Home win (1) | odds 1.80–2.00 | 298 | 53.7% (52.9%) | 1.89 | +4.50 | +1.5% | could be luck |
| Under 2.5 goals | odds 1.80–2.00 | 93 | 53.8% (53.2%) | 1.88 | +0.72 | +0.8% | could be luck |
| Over 2.5 goals | AI confidence 40–50% | 226 | 56.2% (55.4%) | 1.86 | +1.30 | +0.6% | could be luck |
Historical simulation: flat stake of 1 unit per tip, no tax, average pre-match odds. Break-even = hit rate needed to finish at zero at these odds. Signal: “solid” = well above break-even and positive in both halves of the period; “promising” = above break-even in both halves; “could be luck” = the sample is still too small to tell. Past results do not guarantee future results.
Card predictions: how accurate they are
2,823 settled matches, 02.10.2026–31.10.2026. Next to every accuracy we show the baseline — what you get without any prediction, by always calling the side that happens more often — so it is clear where the AI prediction adds something.
| Market | Accuracy | Settled | Baseline | Edge |
|---|---|---|---|---|
| Over 2.5 cards | 76.5% | 2,823 | 76.2% always over | +0.3 pp · no edge |
| Over 3.5 cards | 63.3% | 2,823 | 59.4% always over | +3.9 pp |
| Over 4.5 cards | 61.6% | 2,823 | 57.5% always under | +4.1 pp |
| Over 5.5 cards | 72.5% | 2,823 | 72.1% always under | +0.4 pp · no edge |
| Both teams booked | 76.6% | 2,823 | 76.4% always yes | +0.2 pp · no edge |
| Red card | 81.1% | 2,823 | 81.1% always no | +0.0 pp · no edge |
| Penalty | 100.0% | 2,823 | 100.0% always no | +0.0 pp · no edge |
The baseline is a prediction-free guess: always calling the side that turned out more often in this sample. A market says something about the model only when the accuracy is above it — a high number on a line where one side happens in most matches says more about the market than about the model.
Card predictions have been stored since 18.09.2026, so shorter periods can show the same numbers. Every prediction is saved before kick-off and settled against the match record; a yellow and a red card count as one each. The call is over/yes when the AI prediction gives at least 50%.
Accuracy by market and period
| Market | 7 days | 30 days | 90 days | All time | Since 15.09 (new model) |
|---|---|---|---|---|---|
| 1X2 | 47% (1 509) | 49% (8 165) | 48% (21 174) | 48% (24 340) | 50% (13 554) |
| Double chance | 73% (1 504) | 72% (8 141) | 72% (20 949) | 72% (23 996) | 72% (13 518) |
| BTTS | 56% (1 509) | 57% (8 165) | 55% (21 174) | 55% (24 340) | 56% (13 554) |
| Over 2.5 | 58% (1 509) | 60% (8 165) | 58% (21 174) | 58% (24 340) | 59% (13 554) |
| Over 1.5 | 81% (1 158) | 82% (7 014) | 83% (13 830) | 82% (16 231) | 83% (10 194) |
| Correct score | 9% (1 509) | 9% (8 165) | 9% (21 174) | 9% (24 339) | 9% (13 554) |
Percentage of correct tips, number of settled tips in brackets. Void and postponed matches are not counted. Over 1.5: matches where the AI tipped over 2.5 goals — how often at least two goals were scored.
Daily 1X2 accuracy
Days with at least 20 settled tips.
Leagues where the model performs best
| League | Tips | Accuracy |
|---|---|---|
 Süper Lig Turkey | 23 | 78% |
 Cup Bulgaria | 29 | 76% |
 Damallsvenskan Sweden | 20 | 70% |
 Oberliga - Schleswig-Holstein Germany | 23 | 70% |
 NB III - Northwest Hungary | 22 | 68% |
 Pro League Saudi Arabia | 25 | 68% |
| U19 Bundesliga Germany | 73 | 67% |
 Serie A Italy | 27 | 67% |
| Oberliga - Rheinland-Pfalz / Saar Germany | 21 | 67% |
 Tweede Divisie Netherlands | 21 | 67% |
Latest settled tips
Value bets record
Historic value bets were selected with an earlier, less precise method. From September 2026 value bets are selected only from the full 1X2 probability model.
How we count
Only tips published by the model before kick-off are counted. A tip is frozen when it is first settled, so later changes cannot improve the record. Results come from official match data after the final whistle; for extra time and penalties the 90-minute score is used, the standard for settling a result.
Transparent results, updated automatically
Our most accurate market over the last 30 days is over 1.5 goals: 82% correct on 7 014 settled tips. Every prediction we publish is settled automatically after the final whistle, based on official results. Over the last 30 days the AI predicted the correct 1X2 outcome in 49% of 8 165 matches. The tables below break results down by market (1X2, double chance, correct score, BTTS, over/under 2.5 goals) and by league.
What the numbers mean
A hit rate on its own does not say whether a pick was worth backing: a model that always picks heavy favourites wins often but pays little, so read accuracy together with the odds. Postponed, cancelled and abandoned matches are excluded. Past performance does not guarantee future results.
Why we publish every result
Many tipster sites only show their winners. We publish every settled prediction, losses included, so you can judge the model by league and by market before you trust it. Only predictions published before kick-off count towards these statistics.