Projections
Every headline stat per skater over the full 2026-27 regular season (Sep 29, 2026 – Apr 10, 2027), ranked by penalty minutes. Each total assumes the skater plays every one of his team’s games.
How the shading works — a stronger shade means more, not better, and a skater on a thin sample is left plain
Each figure sits on a tile shaded by its share of that stat’s leading projection in this view, in five steps from zero to the leader. Hits and penalty minutes shade the way points do. The leader is the league’s, among skaters whose rate rests on at least 41 real games, so a shade means the same on every page and under every filter. A skater with fewer is left unshaded, his sample marked beside his name.
322 defencemen · sorted by penalty minutes · assumes no games missed
The × tag compares a skater’s game-to-game swings in penalties taken, not minutes, with chance: under 1.00 steadier, over 1.00 streakier.
How consistency is measured — repeat-measure reliability 0.10, measured 2026-10-02
The tag is a skater’s game-to-game variance in penalties taken divided by the variance a random (Poisson) process at his own rate would show, measured within each season so a change of role between seasons does not count as inconsistency. Each penalty counts once, whatever its length. Measured in minutes, which come in 2-, 5- and 10-minute units, every skater would read streakier than chance because of the units alone.
Its repeat-measure reliability, 0.10, is how closely the figure from a skater’s odd-numbered games matches the one from his even-numbered games, across skaters. A stat is shown only once that agreement is clearly above what chance produces.
How the projections are built — a rate over every season since 2021-22, recent games weighted more, times the club’s schedule
A skater’s rate is his per-game output over every season since 2021-22, each season weighted half as much as the one after it, and pulled a little toward the league rate so a handful of games cannot top the board. The weighting was chosen by testing which one best predicted the rest of a season from the games before it. A skater needs ten games to appear here.
Each total is that rate times his club’s scheduled games. A full-season pace counts every scheduled game, played or not.
Availability is not projected: Team games is the club’s schedule, not a forecast of how many games the player dresses for, and injuries, scratches and trades are not modelled.
How well does projection ordering hold? View the ordering study
Scoring per dollar of cap hit has a board of its own: Cap efficiency