Postseason Probabilities
Loading the season 0 seasons simulated — pp at 95% Loading model
In Outno games picked
Checking the live feed
Loading the season 0 seasons simulated — pp at 95% Loading model
In OutPreliminary forecast. Each number comes from the selected number of simulated seasons. Progress reflects completed simulations. Dashes mean a forecast is unavailable.
The six outcomes are Playoffs, Division, Round 2, Conference Finals, Cup Final, and Cup. Expand a club for division places 1–3, wild cards 1–2, Out, and any outcomes hidden on a narrow screen. W–L–OTL records include accepted finals and your picks immediately. A loss marked * includes an Any finish pick: it may become an overtime or shootout loss. Selecting a finish makes the record exact. The line beneath shows projected points.
Small arrows compare all your picks with a no-pick baseline using the same data, seed, and sample count. Hypothetical results constrain outcomes; they never train the ratings. Accepted actual finals alone update team strength.
Sampled 0 or 100 is not proof of elimination or qualification. The sampling margin describes simulation noise, not predictive accuracy. The line after the eighth row marks the size of each conference’s playoff field and follows your selected sort.
Elo is the fast default. It samples winners and hockey finishes, ranking simulated ties by points, regulation wins, regulation-plus-overtime wins, total wins, then a seeded lottery. Enhanced retains the original detailed simulation, adding score draws and head-to-head and goal-based tiebreaks. Both use the same preliminary team ratings and exact rules for completed actual standings and playoff results. Enhanced takes longer; greater predictive accuracy has not been established.
Both models omit goalie, player, injury, and travel adjustments. The chronological evaluation is descriptive because historical inputs were inspected beforehand. No predictive-superiority claim is made. Enhanced also uses a seeded lottery for hypothetical ties surviving all published criteria.
Picking games. Choose a winner, then optionally choose Regulation, Overtime, or Shootout on the metadata line. Any finish samples the finish conditional on your winner. Changing winners preserves the finish choice; selecting the same winner again clears both.
The official 2026–27 slate contains 1,344 regular-season games: 84 per club, split evenly between home and away. Every matchup and date was reconciled against the NHL’s published schedule. Preseason and exhibitions are excluded.
Filter one or several clubs, then close the selector with Done. Today preserves filters and moves to the nearest matching date. Earlier and later games load in bounded chunks. Start times display in Eastern time; Today follows your local calendar.
Live scores are display-only. Accepted finals update records and ratings and retire affected picks. Shared links retain game identities, opponents, winners, optional finishes, model, samples, and seed. Matching data and model revisions are required for identical forecasts. Postseason results are replayed within the playoff bracket; postseason games are read-only.
NHL Forecast · 2026–27
Preview · Preliminary model
Schedule and historical game inputs come from NHL-operated feeds. The schedule was checked against the published 2026–27 slate. Current team artwork comes from NHL assets; conference artwork is from Wikimedia. This independent project is not affiliated with the NHL or its clubs; names and marks identify their respective owners.
Fast Elo and the detailed Enhanced simulation cover the full regular season and four playoff rounds using the same preliminary team ratings. Historical finish probabilities are conditional on the selected winner’s strength. Enhanced also samples scores for detailed tiebreaks. Future draws keep current ratings fixed. Old scenario links retain the original detailed engine as Enhanced. Market is unavailable. No paid or player-level inputs are used.
This preview is unindexed. Publication and live updater activation require separate authorization. Local fixture checks do not establish production live-to-final behavior or real-device verification.