Methodology
We believe in showing our work. Here's a plain-English breakdown of how every MatchEdge AI probability is generated.
Each nation has an overall strength score combining historical performance, squad quality and competition tier.
Ratings update after every international fixture. Wins against higher-rated opposition move ratings more than wins against weaker opponents.
Attacking and defensive xG distributions are estimated per team, then combined into a match-level scoreline simulation.
More recent matches carry more weight. Form is a decay-weighted average of result quality and opponent strength.
Host advantage, travel distance and altitude are modeled per fixture using historical World Cup base rates.
Every prediction is sampled across thousands of simulations to capture variance. Outputs include a confidence score.
Decimal odds are derived directly from model probabilities. No bookmaker margin is applied — these reflect the model's view.
Low / medium / high tiers based on input sample size and prediction variance. Low confidence means treat the output with extra caution.
Model v2 beta
Public predictions now use Model v2 beta. Inputs:
- external Elo rating (primary team-strength input)
- FIFA ranking (secondary reference)
- internal attack, defense and team strength
- expected-goals assumptions
This is a transparent beta model and will continue to improve as more source data becomes available.
Predictions are probabilities, not guarantees. This is not betting advice.
Next steps
Future improvements may include:
- recent international form
- goals scored and conceded
- opponent-adjusted results
- squad and availability updates closer to the tournament
External data will be added only when source quality and licensing are suitable.
Team comparison tool
Compare ratings, verified recent form and group-stage context using public beta data.
Explore team comparison →