Who's the unluckiest
NHL player ever?
Brad Park. I scored the 305 best finished NHL careers on nine kinds of bad luck, then ran the model 10,000 different ways. Park finished first in 79.5% of them and in the top 5 in 98.5%.
Big number = share of the 10,000 versions of the model where he finished in the top 5. “#1 in” = how often he finished first.
Watch the model get built, version by version.
The case for Brad Park
A Hall of Fame defenceman for the Rangers and the Bruins, from 1968 to 1985. He was in the room every spring and watched the door close.
In 17 seasons, and his team made the playoffs in every single one of them.
At his teams' real odds, he wins at least one Cup in 8,634 of 10,000 simulated careers. That's an 86% chance.
His season was ended by the team that went on to win the Cup 9 times. The pool average is about 3.
1972 vs Bruins, 1977 vs Canadiens, 1978 vs Canadiens.
Four times behind Bobby Orr and twice behind Denis Potvin. He never won it.
Including the 1974 Game 7 against the Flyers and the 1979 too-many-men Game 7 in Montreal. His teams also lost 40 playoff goals to cold shooting and cold goaltending.
His best chance in any one season was only 20% (1971), about 1 in 5. But he got 17 tries.
Highlighted = the champion was a dynasty (the 1970s Canadiens, the early-1980s Islanders).
How Brad Park rose
Park's rank at each version of the model, out of 305. He took over at version 4 and never gave it back.
Change every choice, same answer
I chose the weights, the prime window and the pool. So I ran 10,000 versions with random weights, different pools and different team ratings, and sometimes deleted a whole factor.
Park's #1 share when that factor is removed entirely. He wins 82.5% with all nine.
I also tried a version where bad luck counts more the greater the player was. Park still came first in 54.5% of universes, and the only player who got close was Ray Bourque at 21.9%. In the main model, Bourque is #2 on the full board and finished top 5 in 36.1% of universes.
All 305 players
The full model (version 6, equal weights). Score = the nine factors added up, each measured against the pool, so 0 is average. Tap a column to sort.
Bounce goals: playoff goals lost to cold shooting and cold goaltending against the team's own regular season (shot data starts in 1959-60; earlier seasons count as 0). The usual names rank lower: Roberto Luongo 15th · Joe Thornton 16th · Marcel Dionne 29th · Cam Neely 31st · Eric Lindros 34th · Bobby Orr 40th · Mario Lemieux 69th · Mike Bossy 295th.
How the model works
Version 1 · Great, no Cup
The simplest definition of bad luck: you were great, and you never won the Cup. Career point shares count only if the Cup count is zero. Leader: Roberto Luongo.
Version 2 · Cups expected vs won
Every season, his team gets its fair share of that year's Cup based on how good it was in the regular season. Expected Cups minus Cups won. Leader: Roberto Luongo.
Version 3 · Bodies
How much of his prime, age 22 to 31, he actually got to play. The average player in the pool lost about 9%. Leader: Cam Neely.
Version 4 · The dynasty tax
Every season that ended against the eventual champion, with extra weight for dynasties and for losing in the Final. Leader: Brad Park.
Version 5 · Bounces
Series lost in the last possible game (double if by one goal), plus playoff goals lost because his team's shooters and goalies went cold. Leader: Brad Park.
Version 6 · Circumstances
Weak teammates, 2nd-place award finishes, and bad timing: a team that just got rid of him wins the Cup. Leader: Brad Park.
10,000 universes
Every version of the model with random weights, pools, prime windows and team ratings. Park finished first in 79.5% and in the top 5 in 98.5%.
Can it tell who should have won?
A backtest means I hid one season at a time and asked the model to call it from the final standings, like an old exam with the answers covered up. I did that for all 99 seasons.
When it gave a team about a 12% chance, those teams won 11.7% of the time (12 of 103). It expected 97 of the 305 players to finish without a Cup, and the real number is 96.
The pool: Hockey-Reference's top 250 careers in point shares plus every Hall of Famer with 300+ NHL games, with at least half the career after 1926. Active players are left out, because an unfinished career can't be scored. Goalies are included.
Jean Ratelle kept showing up near the top too (10th on the full board), and he was traded to Boston in the same 1975 deal as Park. They lost the same 3 Finals and 8 of the same 9 exits to the champion. In that 1979 Game 7, Ratelle had 3 assists.
Keep digging
The luckiest NHL player ever →
The opposite question: the career that got the most Cups it didn't earn.
Most overratedThe experts vs the data
Which players the experts rank far higher than what they did on the ice, on every team.
Most underratedImpact vs the experts
The same model pointed the other way: who the data loves and the experts missed.