Projections combine three ingredients: a weighted average of the last three seasons, regression to the league mean proportioned to games played, and an age curve that lifts young players and fades veterans. Projected games account for injury history and for the 2026-27 calendar, expanded to 84 games under the new CBA; nobody is projected at 84 games by default.
Goalies are handled separately: save percentage, starts, wins and shutouts are regressed, then projected onto a realistic role.
Hits, blocked shots, goals against, saves and overtime losses are projected by the same method, from the same three seasons.
The exact parameters are not published. They are calibrated by backtest: 2025-26 projections, recomputed from earlier seasons only, were compared against actual results, and the chosen parameters are those that minimize the error — particularly for top scorers and players under 23.
The displayed range covers roughly 80% of outcomes (P10 to P90). It measures the normal randomness of a season, not major injuries or role changes.
The probabilities in player cards, the comparison tool and "My team" come from Monte Carlo simulations. Each simulated season first draws talent uncertainty — wider when the three-season sample is thin —, then a number of games consistent with injury history, then goals and assists game by game. Two thousand seasons per player, identical draws as long as the scoring is unchanged. Players are simulated independently: joint probabilities run slightly tight, especially for linemates.
Value is salary-based. For each rank of the projected standings, the expected cap is the median cap of players ranked nearby (±15 ranks for skaters, ±5 for goalies). The column shows the gap between that expected cap and the player's actual cap: +3.0 reads "he produces like his ranking neighbours but costs three million less" — the number that matters in a salary-cap pool. A dash: outside the draftable pool (top 250 projected skaters, top 40 goalies) or no listed contract.
Rebound estimates the pool points to be regained over the 2025-26 season: the gap between it and the lower of two benchmarks — our projection and the best of the last two seasons. Capping at the level already demonstrated is deliberate: a rebound is only deemed likely if the player has produced at that height before. The column also shows the reverse: a player who produced above his projection in 2025-26 shows, in red, the expected drop. Gaps under 5 points are not shown; an asterisk flags a benchmark built on few games. Published rankings (NHL.com Top 250 of August 26, 2026, CBS Sports of August 13, ESPN of July 17) remain in the player card as an external benchmark.
The Teams tab projects all 32 teams. Each team's strength starts from its roster: the sum, at a per-84-game pace, of our projections for its forwards, defensemen and goalies, on current teams — games an injured player misses are played by someone else — so off-season moves count. It is combined with the team's recent trajectory, weighted toward last season, then the real 84-game schedule is simulated 2,000 times, home ice and the overtime point included; current-season results gain weight as they accumulate. Caveat: lineups can still move at training camp. The PP1 tag marks the five skaters with the most power-play ice time on their team: per 2025-26 at first, per actual deployment once the season provides enough games.
Net value brings everything back to pool points: projected points, plus contract savings converted to points — at the exchange rate observed among draftable players, the points-per-million slope recomputed under your scoring —, plus half the rebound. The half is deliberate: the projection already believes in the rebound; the remaining half pays for the draft sheet's lag. Hovering a value shows the breakdown.
Data: NHL stats API (api.nhle.com), 2023-24 through 2025-26 regular seasons; 2026-27 calendar: 84 games per team. Age as of October 1, 2026. Salaries (2026-27 cap hit and remaining contract years): capwages.com, August 29, 2026 snapshot; a dash means no listed contract at that date. "Pts/$M" divides projected pool points by cap hit. External benchmarks: NHL.com "32 in 32" series, August 2026. Season results refresh every morning.