Compare any two skaters

Pick any two NHL skaters and see their 2026-27 projected stat line, per-game rates, game-to-game consistency, and next scheduled game, side by side. It's free for every skater the projection covers.

The same season projection published on each player's own page and on /2026-27/players: games x ice time x rate, projected separately per skater with an 80% range fitted from the model's own out-of-sample error.

What the tier is, measured. Backtested over three complete season transitions (2021-22 -> 2022-23 -> 2023-24 -> 2024-25, 520 player-seasons), last season's STEADY tier does predict a lower next-season cv than BOOM/BUST, a gap of 0.224 cv units at p = 4e-25. But cv is mechanically about 1/sqrt(mean) for a low-count stat, and within every season measured 1/sqrt(points-per-game) explains 77% of cv. Control for it and the gap collapses to 0.003 cv units at p = 0.79; the tier adds R^2 of 0.0007 over prior-season points-per-game alone, which out-predicts the tier outright (0.335 vs 0.218); and a placebo tier built from points-per-game rank-thirds with cv never consulted reproduces the whole effect (0.229) and agrees with this tier on 77% of player-seasons. So: read this as a scoring-rate ordering with a volatility name. It is NOT evidence that one skater is more reliable than another at the same scoring rate. Two further limits: the cut is a rank-third within the season's own field, so the cv that is STEADY one year drifts by up to 0.08 cv units across seasons and relabels up to 30% of a position group -- the tier is not comparable year to year; and a bootstrap of each skater's own games puts the tier in a different third in about 30% of resamples at EVERY sample length, a full 82-game season included.

Yes. The comparison tool works across positions; PuckCast's separate "Players like this" similarity search on each player page is position-matched, for finding a comparable player rather than an arbitrary one.