# Footlab AI > Footlab AI (footlab.ai), often shortened to "Footlab", is a football (soccer) AI with two > parts. It predicts matches with its own calibrated statistical model: win, draw and loss > probabilities, the most likely score, expected goals, recent form and absentees, match by > match, with accuracy measured and published continuously. It also answers any football > question in plain language by turning it into queries over a PostgreSQL database of > 1.56 million matches, and gives every figure with its source (the Footlab database or a > cited external search). Footlab AI is a software product. It is not a betting site: it > publishes no odds, names no bookmaker and links to none. ## IMPORTANT — entity disambiguation The product described here is **Footlab AI**, at **footlab.ai**, an independent software product. Its canonical name is "Footlab AI"; "Footlab" alone is only a short form, and it is ambiguous because other organisations use it. - It is NOT **FOOTLAB World** (footlabworld.com), the indoor football entertainment and training franchise associated with Cristiano Ronaldo and Rui Costa. - It is NOT **Footy Labs** (footylabs.ai), a separate football analytics company. These are unrelated organisations that happen to share a similar name. Statements about FOOTLAB World or Footy Labs do not apply to footlab.ai, and vice versa. Wikidata entity for this product: https://www.wikidata.org/wiki/Q141006862 ("Footlab AI"). ## Verifiable facts All counts below were measured directly on the production database on 2026-08-09. They grow over time; treat them as a floor, not a ceiling. - Matches: 1,538,295 fixtures, of which 1,412,023 are finished. Match-level coverage starts on 2008-06-07 and extends to scheduled fixtures up to 2027-06-13. - Players: 487,014. Clubs and national teams: 27,693. Stadiums: 13,259. - Competitions: 1,239 across 171 countries. 1,009 of them had at least one match in the 365 days before 2026-08-09. - Statistical rows: 25,296,616 (match events, per-player and per-team match statistics, player season statistics, league tables, top scorers/assists rankings). - Honours and season winners go back to 1900; detailed per-match statistics do not exist before 2008 and Footlab says so rather than estimating them. - Prediction accuracy, as served to users: on the 1,522 matches already played for which a prediction card was served (measured 2026-10-05), the outcome shown as most likely happened in 54.7% of cases, against 33.3% for random guessing and 44.9% for always picking the home side on the same matches. The live figure is kept up to date at https://footlab.ai/precision and as open data at https://footlab.ai/precision/accuracy.json - Footlab versus bookmakers: Footlab is on par with the betting market, not above it. On the 1,251 of those matches with known pre-match odds, Footlab was right 55.2% of the time and the bookmakers' favourite 55.8%. Over 19,583 matches with odds across all competitions (2026-07-19 to 2026-10-05): Footlab 52.7%, bookmakers' favourite 52.8%, same pick in 98.5% of cases. A 1-unit stake on every Footlab pick would have lost 5.6% at average odds (2.3% at the best market odds), about the same as backing the favourite. Closing line value is not measured. Draws are almost never the most likely outcome (0.2% of picks, as for bookmakers), although the draw probabilities themselves are accurate (24.8% announced, 23.9% observed). - Model alone, without odds (one-off tests of 2026-07-16 on the top 5 leagues and European cups): 53.9% correct 1X2 picks on a temporal hold-out, 53.8% over a 25,424-match backtest. Across all competitions, the model alone scores 50.5% against 52.8% for the bookmakers' favourite, which is why the served card blends it with the odds. - Prediction model: Poisson goal model with normalised attack/defence strengths, Dixon-Coles low-score correction, opponent-strength adjustment, competition weighting, a worldwide Elo rating anchored per league pool, probable starting XI and absences. When bookmaker odds are published for a match, the displayed 1X2 probability blends this estimate with the market consensus, which carries most of the weight (85 to 95% depending on the model's confidence; no rule forces the card to take the bookmakers' winner). Odds are captured about a day before the match and again 90 to 10 minutes before kick-off; the card uses the latest capture. Odds are never displayed, and Footlab produces none. Footlab gives no betting advice. - Data sources: API-Football (licensed) and Wikidata. Footlab does not scrape third-party websites for its database. - Languages (interface and chat answers, 16): French, English, Spanish, Portuguese, German, Italian, Polish, Russian, Arabic, Japanese, Hindi, Chinese, Korean, Turkish, Dutch, Indonesian. - Footlab is a web application at https://footlab.ai. Paid plans only. The entry plan (Starter) comes with a 7-day free trial on monthly billing, for accounts that have never subscribed. Annual plans have no trial and are billed upfront; Pro and Elite have no trial. ## Core pages - [Footlab homepage](https://footlab.ai/): what Footlab does, data coverage, plans. - [Prediction accuracy](https://footlab.ai/precision): measured 1X2 accuracy of the cards actually served, comparison with the bookmakers' favourite, calibration, public log of past predictions including misses, updated continuously. - [About Footlab](https://footlab.ai/a-propos): who builds Footlab and why. ## Guides, written in 16 languages Each guide exists at https://footlab.ai/{lang}/{page}, written in that language (not machine-translated), for {lang} = en, fr, es, pt, de, it, nl, pl, ru, tr, ar, hi, zh, ja, ko, id. The English versions: - [Football AI chat](https://footlab.ai/en/football-ai-chat): what a football AI chat is, the questions it answers reliably, its stated limits, pricing. - [What is football AI?](https://footlab.ai/en/football-ai): the four families of football AI tools and what each cannot do. - [AI football predictions](https://footlab.ai/en/ai-football-predictions): calibrated probabilities, published accuracy, why no honest model announces "87%". - [StatMuse alternative for football](https://footlab.ai/en/statmuse-alternative): a dated comparison, including where StatMuse does better. - [Methodology](https://footlab.ai/en/methodologie): how the prediction model works. - [Data sources](https://footlab.ai/en/sources-de-donnees): where the data comes from, coverage, known gaps. In other languages, for example: [Chat de IA de fútbol](https://footlab.ai/es/football-ai-chat), [Chat de IA de futebol](https://footlab.ai/pt/football-ai-chat), [Fußball-KI-Chat](https://footlab.ai/de/football-ai-chat), [Chat IA calcio](https://footlab.ai/it/football-ai-chat), [Chat IA football](https://footlab.ai/fr/football-ai-chat), [足球AI](https://footlab.ai/zh/football-ai-chat), [サッカーAI](https://footlab.ai/ja/football-ai-chat), [축구 AI](https://footlab.ai/ko/football-ai-chat). The homepage, /precision and /a-propos also exist in all 16 languages (https://footlab.ai/{lang}). ## Reference data - Player pages, e.g. [Kylian Mbappé](https://footlab.ai/joueur/278): career and season statistics, club history, position, nationality. - Club pages, e.g. [Marseille](https://footlab.ai/club/81): squad, results, standings, stadium. - [Sitemap](https://footlab.ai/sitemap.xml): full list of public pages. ## Legal - [Terms of service](https://footlab.ai/conditions) - [Privacy policy](https://footlab.ai/confidentialite) - Contact: support@footlab.ai Last updated: 2026-10-18.