The AI Goal Lab algorithm
Quantitative analysis of football matches, built on data and published with full transparency on results.
More than 20 factors per match
Each match is assessed across six categories of data. The algorithm combines them into a single probability estimate per market.
Team form and strength
Recent performance at home and away, goals scored and conceded, and the longer-term level of each team.
Chance quality (xG)
Expected goals measure the quality of chances created and conceded, a more stable indicator than the final score alone.
Team news
Injuries, suspensions, probable line-ups and rotation risk.
Match conditions
Competitive stakes, rest and fixture congestion, travel, weather, derbies and managerial changes.
Market data
Betting market prices and their movement before kick-off carry valuable information and are an input to the model.
Match details
Corner and card tendencies, referee profile and head-to-head record, for markets beyond the result.
From data to pick
- 01
Data collection
Statistics, team news, line-ups and market prices are gathered daily from multiple sources.
- 02
Modelling
The model estimates the expected performance of each team and converts it into probabilities for result, goals, corners and cards.
- 03
Validation and calibration
Estimates are compared with the market and calibrated. Weak signals are excluded, and when no option meets the criteria, no pick is published.
- 04
Publication
One pick per match with supporting reasoning, recorded before kick-off and never altered.
Transparency and verification
- check_circleThe algorithm was evaluated on thousands of historical matches across nine leagues, using a forward-looking simulation that never uses future information.
- check_circleEvery pick is recorded before kick-off. All results are published, favourable and unfavourable alike.
- check_circleNo pick is published when the analysis does not support one.
- check_circleProbabilities are estimates, not guarantees. For persons aged 18 and over only.
Algorithm training period
The algorithm is currently in a trial period of training and optimisation. For its duration, all tips are provided free of charge. Results are public, so you can assess performance for yourself.