A methodology disclosure for the assessment grade system applied to 18,000+ players across 20+ leagues. What the grades measure, how confidence is determined, and the evidence that the system identifies quality.
Each grade places a player on a six-point scale relative to position peers across 20+ leagues. The percentile ranges reflect where a player sits within their position group after league adjustment.
Within each tier, grades carry a modifier: + (top third), no modifier (middle third), or − (bottom third) — producing 18 effective grade levels. An SW-2+ sits at the boundary of Elite; an SW-2− is closer to High Quality.
Every grade is a composite of four computed dimensions. The dimensions are weighted by position — the blend for a centre-forward is not the same as for a centre-back. The composite produces a continuous score (0–100) that maps to the discrete grade tiers above.
The statistical dimension alone does not constitute the grade. A player with a high statistical percentile but low pattern convergence and limited framework evidence receives a lower composite than the statistics alone would suggest. The system is designed to avoid false confidence from any single signal.
A grade describes a specific, position-relative assessment — not a universal ranking. The system is designed to make like-for-like comparisons meaningful.
Trajectory describes the direction of a player's performance relative to what would be expected for their age and position. A 33-year-old centre-forward holding steady is a different signal than a 22-year-old doing the same.
Trajectory compares observed season-over-season change to the expected change from age-curve models, adjusted for league difficulty changes. It requires a minimum of two consecutive seasons of data.
The grade system was retrodiction-tested by applying the current model to historical data across nine seasons. This tests whether the grades, as currently computed, would have distinguished quality when applied to the past.
A STATSWING grade answers one question: how good is this player relative to position peers, adjusted for league context, with what confidence? It is a present-tense assessment of current quality in context. It does not answer several adjacent questions that it may appear to answer.
The assessment model is periodically recalibrated as additional empirical data becomes available. Dimension weights are derived from regression against validated outcomes, not editorial assignment. When the model is recalibrated, STATSWING publishes a research note documenting what changed and why. Published research examining the measurement conventions and frameworks behind these assessments is available at statswing.com/research.