A team that keeps its 5 players carries last year’s rating forward as a prior anchor, instead of resetting to zero. Roster turnover penalizes the carry-over: each player swap pulls the prior closer toward 0. Mainly relevant at the start of each season.

For every map a team plays in VCT, the model records the final round score against the opposing team. For instance, a 13–2 win carries far more signal than a 13–11 win.

A linear algebra solve finds the rating vector that best explains all observed round differentials simultaneously.

Game weights follow exp(−λ × weeks ago). Half-life = 5 weeks.

Per-map ratings with smaller sample sizes are blended toward the team’s overall rating.

Each team runs through 10,000 simulated BO3 vetoes against league-average opponents using historical ban/pick patterns. Expected round-diff across the surviving maps becomes the headline rating. Thus, a great ban target is worth as much as a great map.

Each team’s rating is the expected round differential they’d post against a league-average opponent across 10,000 simulated vetoes — raw map strength times how well their pool survives ban/pick.