Who's the luckiest
NHL player ever?
Henri Richard. He won 11 Stanley Cups, the most ever, and my model says his own play earned about 1.6 of them. His teammates alone would have been expected to win about 6.2. I scored 2,163 finished NHL careers and ran the model 10,000 different ways: he finished first in 35.8% of them, about twice as often as anyone else.
Big number = share of the 10,000 versions of the model where he finished first. It's a close race, not a runaway: every name here was #1 in some versions.
Watch the model get built, version by version.
The case for Henri Richard
A Hall of Fame centre who played all 20 of his seasons for the Canadiens. He won the Cup in each of his first five, and no player in history had stronger teammates for longer.
The most by any player. 20 seasons, all with Montreal, 1955 to 1975.
His team's Cup odds with him, minus its odds with his own point shares taken off the standings.
The most of anyone in the 2,163 careers. His Canadiens were good enough to win without him.
At his teams' real odds, also the most in the pool. He still won 3.2 more than that.
In 7 of his 11 Cup years, the pre-playoff favourite was beaten by somebody else before Montreal had to face it.
Goals his teams gained from hot shooting and hot goaltending in the playoffs. He also went 3-1 in series that went the distance.
The 2,163 luckiest careers
The full model (version 5, equal weights). Score = the luck factors added up, each measured against the pool, so 0 is average. Search any player or tap a column to sort.
Earned = the Cups his own play added to his teams' odds. Not earned = Cups won minus earned. Teammates = the Cups his teams would have been expected to win without him. Favourites out = Cup years when the pre-playoff favourite lost to somebody else. Bounce goals start in 1959-60, when shot data begins. #1 in = share of the 10,000 versions where he finished first.
Henri Richard, season by season
Montreal's Cup odds each season, with him (red) and with his own point shares taken off the standings (grey). ★ = won the Cup. The gap between the bars is what he earned. The grey bars are his teammates.
His least likely Cup was 1971: the model gave Montreal a 9% chance, and he scored the tying and winning goals in Game 7 of the Final in Chicago. 1970 = missed the playoffs.
Version by version
Richard's rank at each version of the model, out of 2,163. He took over at version 2 and never gave it back.
Cups divided by career point shares. It rewards being bad: the leader won 4 Cups with fewer than 0 point shares.
His team's Cup odds with him, minus its odds with his own point shares taken off its standings points, is the part he earned. The rest is luck.
The pre-playoff favourite knocked out by somebody else in a year he won the Cup.
Series that went the distance (double if the last game was decided by one goal), plus playoff goals gained because his team's shooters and goalies ran hot.
Cups in a player's first or second season with a team, and Hart, Norris or Vezina wins in seasons someone else at his position had more point shares.
10,000 universes: change every choice
I chose the weights, the pool, the team ratings and even the definition of luck. So I ran 10,000 versions that each pick those at random, then counted who finished first.
Share of the 10,000 versions where Richard finished ahead of each challenger.
Each version also picked one of seven definitions. Richard wins three of the seven, and he's in the top three of every other one except pure luck, the one where teammates don't count.
| Definition | Most often #1 | Share | Richard |
|---|---|---|---|
| Cups he didn't earn, nothing else | Henri Richard | 60.3% | 60.3% |
| Everything except soft awards | Henri Richard | 57.1% | 57.1% |
| Everything except arrival and soft awards | Henri Richard | 42.4% | 42.4% |
| The full model | Bryan Trottier | 41.7% | 34.6% |
| The full model plus health | Bryan Trottier | 48.9% | 24.6% |
| Pure luck: teammates don't count | Yvan Cournoyer | 40.3% | – |
| The NBA model's recipe | Yvan Cournoyer | 25.2% | 17.2% |
Share = how often the leader finished first in the versions that used that definition (about 1,400 each). Richard = his own share; – = he wasn't among the top three.
How the model works
The pool
Every finished NHL career with at least 5 full seasons of games (goalies 4), 1918 to 2026. 2,163 careers; 807 won at least one Cup. Active players are left out, because an unfinished career can't be scored.
The team Cup model
Each playoff team gets a fair share of that year's Cup from its regular-season points percentage (a conditional logit, the same model as my McDavid and Unluckiest videos).
Version 1 · Cups per point share
Cups divided by career point shares. It rewards being bad: the leader won 4 Cups with fewer than 0 point shares. Leader: Kevin McClelland.
Version 2 · Cups he didn't earn
His team's Cup odds with him, minus its odds with his own point shares taken off its standings points, is the part he earned. The rest is luck. Leader: Henri Richard.
Version 3 · Dragons
The pre-playoff favourite knocked out by somebody else in a year he won the Cup. Leader: Henri Richard.
Version 4 · Coin flips and bounces
Series that went the distance (double if the last game was decided by one goal), plus playoff goals gained because his team's shooters and goalies ran hot. Leader: Henri Richard.
Version 5 · Arrival and soft awards
Cups in a player's first or second season with a team, and Hart, Norris or Vezina wins in seasons someone else at his position had more point shares. Leader: Henri Richard.
10,000 universes
Each one randomly picks the weights, the player pool, the team ratings, the scaling, whether the 6-team era is discounted, whether teammates count as luck, and whether a Cup counts when he never played in that spring's playoffs. Richard finished first in 35.8%, ahead of Bryan Trottier (18.7%), Yvan Cournoyer (17.9%) and Jacques Lemaire (11.6%).
A backtest means I hid one season at a time, trained the model on the other 98 and asked it to call the hidden season, like an old exam with the answers covered up. I did that 99 times.
Teams it gave about a 12% chance won 11.7% of the time (12 of 103). Across all 2,163 careers it expected 1,622 Cups, and they won 1,639.
Split every career into odd and even seasons and score each half on its own. Points per game match between the halves (r = 0.95). Cups vs expected (0.19), bounces (0.09) and coin flips (0.07) barely do. That's what luck looks like. Teammates repeat (0.84), because they're circumstance, not chance, which is why some versions leave them out.
- Two judgement calls decide the order at the top: whether teammates count as luck, and whether a Cup counts when he was hurt and missed that whole playoff run (Cournoyer in 1977 and 1979). Both are split 50/50 across the universes.
- Career totals favour long careers on dynasties. Per season, the answer is Ken Dryden.
- The team model only sees the regular season, so it can't see deadline trades, and it slightly underrates strong favourites.
- Bounce goals need shot data, which starts in 1959-60. That slightly hurts Richard, whose first Cups came earlier.
Sources: point shares from Hockey-Reference; games, playoff series, Cup rosters and awards from the NHL stats and records APIs.
Keep digging
The unluckiest NHL player ever →
The same idea pointed the other way: the career where the bounces, the dynasties and the timing went wrong the most.
Most underratedImpact vs the experts
Who the data loves and the experts missed, on every team.
Most overratedThe experts vs the data
Which players the experts rank far higher than what they did on the ice.