How accurate are Footlab's football predictions?
Footlab's prediction model reaches 53.9 % correct 1X2 outcomes on an out-of-sample temporal hold-out, and 56.7 % on UEFA competitions (measurement run of July 16, 2026). Random guessing between the three outcomes would give 33.3 %. Every prediction served is then logged and compared with the official match result: the tables on this page come from that continuous measurement, not from a claim.
Footlab is a conversational AI specialized in football: it turns a question asked in plain language into an SQL query over a database of 1.52 million matches, and shows the query next to the answer.
Last updated: August 22, 2026.
Live numbers
Reference: a 25,424-match backtest (same pipeline as production) at 53.8 % correct 1X2 calls — chance alone would give ~33%.
How does Footlab measure the accuracy of its predictions?
Footlab records the last prediction served before kick-off, then compares it with the official match result no earlier than three hours after the final whistle. Nothing is filtered, nothing is deleted, and the same computation is replayed on a sample postdating the model's calibration period.
Calibrated probabilities
The model doesn't call a winner, it states probabilities — calibrated so that "60%" really happens about 6 times out of 10.
Continuous measurement
Every sheet actually served is logged, then compared to the official result once the match ends. Nothing is cherry-picked, nothing is deleted.
Temporal testing
The model is validated on matches played after its calibration window (hold-out): it is judged on games it has never seen.
Measured benchmarks
- 3,617 official matches evaluated · 1X2 accuracy by confidence bracket: stated 70% → realized 67.7%.
- Backtest over 25,424 matches (big-five leagues + European cups): 53.8% correct 1X2 calls, versus ~33% by chance.
- Temporal hold-out: 53.9% on matches the model had never seen, 56.7% on European cups.
- Half-time draw: 39.5% stated versus 40.0% observed — a calibration gap under one point.
Measured on 10 Aug 2026.
Are the probabilities announced by Footlab calibrated?
The table compares, in bands of 10 points, the probability announced by the model with the frequency actually observed, across 109 evaluated predictions. A well-calibrated model aligns the two columns; a gap within one band remains expected while that band's sample is small, hence the sample column.
| Announced band | Average announced | Observed frequency | Sample |
|---|---|---|---|
| 30–40% | 38.5 % | 30.8 % | 13 |
| 40–50% | 45.4 % | 53.5 % | 43 |
| 50–60% | 54.7 % | 60.9 % | 23 |
| 60–70% | 65.4 % | 56.3 % | 16 |
| 70% and above | 81.1 % | 78.6 % | 14 |
Does accuracy depend on the level gap between the two teams?
Yes, and the difference is clear: the larger the world Elo gap, the more reliable the prediction, and a match between clubs from two different leagues remains harder to predict than a domestic one. Each row states the number of evaluated predictions it rests on.
| Segment | 1X2 accuracy | Sample |
|---|---|---|
| Elo gap below 100 points | 50 % | 64 |
| Elo gap of 100 to 200 points | 56 % | 25 |
| Elo gap of 200 points or more | 75 % | 20 |
| Two teams from the same league | 59.5 % | 42 |
| Teams from different leagues | 53.2 % | 47 |
What is Footlab's accuracy by competition?
The table lists only the 3 competitions with at least ten evaluated predictions: below that threshold, a rate measures nothing. These rates rest on samples far smaller than the hold-out test and are comparable neither with each other nor with the headline figure.
| Competition | 1X2 accuracy | Sample |
|---|---|---|
| Friendlies Clubs (World) | 50 % | 16 |
| UEFA Champions League (World) | 50 % | 14 |
| UEFA Europa League (World) | 40 % | 10 |
Can Footlab's past predictions be reviewed, including the wrong ones?
Yes: here are the last 30 evaluated predictions, in chronological order, with the probability announced before kick-off, the final score and the verdict, hit or miss. The list is not filtered on correct calls and no prediction is removed after the fact.
| Date | Match | Prediction | Announced probability | Final score | Verdict |
|---|---|---|---|---|---|
| 08/21/2026 | Plaza Colonia – ColónSegunda División | Home win | 42.1 % | 1–1 | Miss |
| 08/21/2026 | Jaguares – ChicoPrimera A | Home win | 51 % | 1–0 | Hit |
| 08/21/2026 | Real Betis – Real SociedadLa Liga | Home win | 47.4 % | 1–0 | Hit |
| 08/21/2026 | Arsenal – CoventryPremier League | Home win | 79.9 % | 3–0 | Hit |
| 08/21/2026 | Cracovia Krakow – Wieczysta KrakówEkstraklasa | Home win | 45.9 % | 3–2 | Hit |
| 08/21/2026 | Erzurumspor FK – GalatasaraySüper Lig | Away win | 64.7 % | 0–4 | Hit |
| 08/21/2026 | Admira Wacker – First Vienna2. Liga | Home win | 38 % | 0–1 | Miss |
| 08/21/2026 | Den Bosch – FC EindhovenEerste Divisie | Home win | 48.6 % | 3–1 | Hit |
| 08/21/2026 | Vitesse – Almere City FCEerste Divisie | Home win | 42.9 % | 0–3 | Miss |
| 08/21/2026 | Emmen – Jong AZEerste Divisie | Home win | 39.1 % | 0–0 | Miss |
| 08/21/2026 | Helmond Sport – WaalwijkEerste Divisie | Away win | 47.9 % | 1–0 | Miss |
| 08/21/2026 | FC Fredericia – Aarhus Fremad1. Division | Home win | 56.8 % | 0–0 | Miss |
| 08/21/2026 | SKU Amstetten – WSPG Wels2. Liga | Home win | 49.3 % | 1–1 | Miss |
| 08/21/2026 | Wacker Innsbruck – SKN ST. Polten2. Liga | Away win | 45 % | 1–0 | Miss |
| 08/21/2026 | FC Liefering – Austria Vienna (Am)2. Liga | Home win | 49.2 % | 2–1 | Hit |
| 08/21/2026 | Austria Salzburg – SV Kapfenberg2. Liga | Home win | 56 % | 3–2 | Hit |
| 08/21/2026 | Hobro – Aalborg1. Division | Away win | 38.8 % | 2–1 | Miss |
| 08/21/2026 | Preußen Münster – Karlsruher SCDFB Pokal | Away win | 44.7 % | 1–2 | Hit |
| 08/21/2026 | Vejle – Esbjerg1. Division | Home win | 55.4 % | 1–1 | Miss |
| 08/21/2026 | SC St. Tönis – Eintracht FrankfurtDFB Pokal | Away win | 92.6 % | 0–11 | Hit |
| 08/20/2026 | Beşiktaş – Kauno ŽalgirisUEFA Europa League | Home win | 82.6 % | 3–0 | Hit |
| 08/18/2026 | Fenerbahçe – LyonUEFA Champions League | Home win | 48 % | 1–1 | Miss |
| 08/16/2026 | Lens – Paris Saint GermainTrophée des Champions | Away win | 54.7 % | 1–0 | Miss |
| 08/16/2026 | Nashville SC – Inter MiamiMajor League Soccer | Home win | 39.3 % | 4–1 | Hit |
| 08/15/2026 | Alaves – GetafeLa Liga | Home win | 50.7 % | 3–0 | Hit |
| 08/14/2026 | Marseille – Atletico MadridFriendlies Clubs | Away win | 42.3 % | 1–2 | Hit |
| 08/12/2026 | Paris Saint Germain – Aston VillaUEFA Super Cup | Home win | 53 % | 2–1 | Hit |
| 08/11/2026 | Sturm Graz – FenerbahçeUEFA Champions League | Away win | 49.6 % | 0–1 | Hit |
| 08/09/2026 | Marseille – Athletic ClubFriendlies Clubs | Away win | 37.4 % | 3–1 | Miss |
| 08/08/2026 | Monaro Panthers – Belconnen UnitedCapital Territory NPL | Home win | 79.2 % | 7–0 | Hit |
Announced probability = the probability the model gave to the outcome it deemed most likely, at the time of the last card served before kick-off.
Frequently asked questions about Footlab's accuracy
Is Footlab a sports betting site?
No. Footlab publishes no odds, links to no bookmaker and receives no affiliate commission. Footlab publishes statistical probabilities and the public measurement of how accurate they are; what a reader does with them is that reader's sole responsibility.
What distinguishes Footlab from FOOTLAB World?
Footlab (footlab.ai) is an independent software product: an AI specialized in football data. FOOTLAB World (footlabworld.com) is an indoor football centre franchise, with no link to Footlab. Footy Labs (footylabs.ai) is yet another company, likewise unrelated.
Is a 53.9% rate good for a football prediction?
Random guessing between home win, draw and away win gives 33.3%. Always picking the home team sits around 45%. Reaching 53.9% on matches the model has never seen therefore places Footlab clearly above both benchmarks — while remaining far from certainty: close to one match in two does not end with the outcome announced as the most likely one.
Where does the data used by the model come from?
From two sources only: API-Football, a commercial provider used under a paid licence, for matches, line-ups, statistics, standings and injuries; and Wikidata, under the free CC0 licence, for historical honors. No data is extracted from a third-party website without permission.
Do you have to pay to see Footlab's accuracy?
No. The accuracy figures, the segment tables and the public log on this page are available without an account and without payment. The same content is served as open data at footlab.ai/precision/accuracy.json.
Statistical analysis based on Footlab data (Poisson model calibrated on 17,000+ matches). Not betting advice.
Footlab (footlab.ai) is an independent software product. It is unrelated to FOOTLAB World, the indoor football centre franchise, and unrelated to Footy Labs.
Last updated: August 22, 2026.