Methodology · Public and versioned
How we build the numbers.
We treat data like journalism. That means every score, ranking, and recommendation must be defensible with evidence — and the methodology behind it must be public.
Inputs
What goes in.
Every score is built from four input streams: match-level event data (goals, xG, xT, progressive carries, press resistance, pass networks), availability data (injuries, minutes, rest days), context data (competition strength, opponent quality, home/away), and biographical data (age, contract, market value).
Weights
How they combine.
Each of the seven AI subscores — Suitability, Form, Fitness, Experience, Readiness, Consistency, Potential — uses a documented, published weighting. The weights are position-specific: a striker's Form leans on xG conversion, a centre-back's on press resistance and duels won.
Calibration
How we stay honest.
Every quarter we back-test rankings against actual match outcomes and re-weight where needed. When we change a weight, we publish the change, the reason, and the effect on top-25 rankings.
Confidence
When we don't know.
Every number ships with a confidence indicator. Fewer than 500 minutes played? Low confidence. Sample size below threshold? We tell you. If we cannot defend a number, we do not publish it.