10,000
seasons.
All 1,344 games, every overtime, every shootout and every playoff series of 2026-27, simulated ten thousand times. Here's how often each team lifts the Cup.
Look up any team
See what's driving each team's odds: last season's chances, playoff experience, how the odds moved as the model was built, and who they'd most likely meet in the Final.
Full table
Click a column to sort. Betting markets = preseason Cup odds, vig removed.
Would it have worked?
Leave-one-season-out: each season since 2009 was predicted by a model that never saw it, using only what was known in October.
How a season is simulated
Rate every team on chances
Expected goals from the last two seasons (score- and venue-adjusted), blended with goal differential. The standings alone carry too much luck.
Add playoff experience
Playoff series won over the last three years. It was the biggest single improvement in the backtest.
Play the real schedule
All 1,344 games of the 84-game season, with home ice, overtime and shootouts, plus random October uncertainty in every simulated season.
Playoffs, 10,000 times
Real playoff format, best-of-seven series. Preseason only: not one 2026-27 game is in the model.
Ideas that failed
- Summer roster moves (v3)
- Goaltending, goals saved above expected (v4)
- The Presidents' Trophy curse
- Cup Final hangover
- Special teams
- "Defence wins championships"
- The hot goalie from last spring
- Blending in the betting markets
- Shot attempts (Corsi): cost one champion
- The first few weeks of the season
Each made the backtest worse or picked no extra champions, so none of them made the final model.