Methodology

Our predictions come from a Poisson distribution model — a real, well-established statistical method used throughout professional sports analytics, not a mysterious "AI algorithm" making unexplainable guesses.

Here's the actual process, in plain language:

1. We calculate real scoring averages. For each team, we work out their average goals scored and conceded per game — home and away tracked separately, because teams genuinely perform differently at home versus on the road.
2. We compare against league-wide averages. Every team's home/away scoring numbers are measured against the league's overall home and away scoring averages, which produces an attack strength and a defense strength rating for each team.
3. We calculate expected goals for the specific matchup. The home team's attack strength is combined with the away team's defense strength (and vice versa) to produce each side's expected goals for this particular fixture.
4. We run the Poisson formula. Using each team's expected goals, the Poisson formula calculates the probability of every realistic scoreline — 0-0, 1-0, 2-1, and so on. From that full grid of scoreline probabilities, we derive the 1X2 result, Over/Under 2.5 goals, Both Teams To Score, and the single most likely correct score.

All the underlying numbers — goals scored, goals conceded, games played, home/away splits — are real, current team statistics pulled from football-data.org, an established football data provider. We refresh this data regularly as new matches are played, so the averages feeding the model stay current throughout the season.

Early in a new season, there simply isn't enough current-season data yet to calculate reliable home/away scoring averages. When that's the case, we fall back to using that team's home/away form from last season instead, so the model still has real numbers to work with rather than guessing. Any prediction built this way is clearly marked with an asterisk (*) on the site, and a note near the top of the predictions page explains it. Once a team has played enough games this season, we automatically switch back to using real current-season data.