Next game day

Win odds come from the site’s best-tested game model: a logistic blend of TOI-weighted player Elo (K=12), each dressed skater’s mean net goals over their last 60 games, the goalies’ GSAx over their last 60 appearances, and a back-to-back rest edge. It is fitted on every complete season, 2010-11 to 2025-26. Tested walk-forward (each season predicted from earlier seasons only, 2012-13 to 2025-26, 16,312 games) its log-loss is 0.6648, against 0.6682 for player Elo alone and 0.6885 for “the home team wins 55% of the time”; it picks the winner 59.5% of the time. Hockey is close to a coin flip, and no model gets far from these numbers.

What moves the odds shows each factor’s effect: the win probability minus the probability with that factor set to even. Home ice is the intercept. The factors overlap, so they need not add up exactly.

Lineups are projected: in season from each team’s last ten lineups, before the season from depth charts. Starting goalies are not confirmed when this page is built (early morning), so each team’s goalie term is weighted by the expected share of starts shown in brackets.

Playoff stakes come from the same 10,000-season simulation as Projected Standings. Each team’s playoff odds are simulated twice with tonight’s game forced to a win and to a regulation loss, on otherwise identical seasons. The ± figure is half that swing, in percentage points. The game with the largest combined swing gets the “Biggest stakes” badge.

Players to watch are the top three skaters on each projected lineup by net goals per 82 over their last 60 games, the Players table’s default view (fewer games count the missing ones as league average), plus both goalies by GSAx per 82.