Stanley Cup overtime is the most random format in sports
Twenty minutes of full-strength, full-ice, sudden-death hockey is the most dramatic format in North American professional sports. It is also, by a meaningful margin, the most random. The next goal wins. There is no shootout to reach for. There is no overtime period after this one that lets the better team reassert itself. The first puck across either goal line ends the night, ends the series if it's late enough, and writes the storyline that the broadcast will replay for the next forty years. The process that produces that storyline is, statistically, much closer to a coin flip than the way we talk about it afterward suggests.
That isn't a critique of the format. Stanley Cup overtime is wonderful television precisely because of the variance. The drama is the variance. But the way fans and broadcasters read overtime results — as confirmation of which team was "better" or which goalie "stepped up" or which star "delivered" — almost always overstates how much the result tells you about either team. The math on playoff overtime is unforgiving, and the narratives attached to it routinely ignore it.
What the goal rate looks like in OT
At five-on-five even strength, NHL teams score at a rate of roughly 2.5 to 2.7 goals per sixty minutes combined, which is about one goal every twenty-two to twenty-four minutes of play. Playoff hockey runs slightly tighter than the regular season because both coaches tend to coach more conservatively and both rosters tend to be deeper at the defensive end, but the basic rate is in the same neighborhood. A single twenty-minute overtime period is therefore expected to produce roughly one goal, on average, with substantial variance around the expectation.
The variance is the part that gets understated. A Poisson-style model of goal arrival, calibrated to the observed rates, gives roughly a 40-45% chance the period ends without a goal at all, around a 35-40% chance of exactly one goal, and a long tail of multi-goal periods. Across a single overtime, the team that produces 60% of the expected goals — a meaningful but not overwhelming edge — still loses something like 38-40% of the time based on the goal arrival math alone. The team that's "supposed to" win, in any reasonable model of in-OT play, loses much closer to half the time than the broadcast framing suggests.
Sudden death amplifies the noise
Sudden death is the variance multiplier. In a normal twenty-minute period, a team that gets the first goal can be outplayed for the rest of the period and still win the period; the score reflects the full twenty minutes of evidence. In sudden death, the moment the first goal goes in, the rest of the evidence is unobserved. The period could have gone another fifteen minutes, the other team could have generated four high-quality chances, the play could have shifted entirely. None of it shows up. The single chance that ended the period is the entire record of what happened.
This is the same structural feature that makes penalty shootouts so widely understood as random. The shootout is a small number of attempts, each one resolved independently, with the result determined by a few high- leverage moments rather than a sustained run of play. Stanley Cup overtime is more sustained than a shootout, but it shares the structural amplification of small samples that any sudden-death format imposes. The first goal is the entire game. Everything that would have happened later is censored.
Goalie save percentages in OT are not what they look like
One of the most quoted stats in playoff broadcasts is a goalie's overtime save percentage across a postseason or a career. The numbers can look astonishing — .950 or higher over substantial overtime workloads. The implied story is that elite goalies elevate in overtime, that the moment produces something extra, that the great ones get greater. The underlying data does not actually support that read.
The censoring problem distorts overtime save percentage in a specific direction. A goalie's overtime save percentage only includes shots that came before the goal that ended the period. If the goalie's team scores after ten minutes of OT in which the goalie faced six saves, all six saves count, and the period ends with no goal against. If the goalie's team gives up the goal after one shot, the period ends with one goal against and a save percentage close to .000 for the period. The distribution of overtime save percentages by individual period is bimodal — either close to 1.000 or close to .000 — because the period ends the moment a goal is scored.
Aggregated across many periods, the headline number can look impressive without telling you anything about the goalie's underlying performance. Two goalies with identical shot-quality-adjusted performance can post wildly different overtime save percentages depending on whether their teams happened to score first more often than not. The number is almost entirely a function of which team broke the deadlock, which is the thing the goalie's performance was supposed to be evidence about in the first place. The reasoning loops.
The format's effect on series outcomes
Roughly 20-25% of playoff games go to overtime in a typical Stanley Cup postseason, and a meaningful share of series get decided by one or more overtime results. In a seven-game series where two of the games go to overtime and split, the impact of OT randomness on the series outcome is moderate. In a series where one team wins three overtime games by a single goal each, the OT variance accounts for most of the series result. The team that won three coin flips moves on; the team that lost three coin flips packs up.
Underlying play tells a different story in those series than the final score does. Five-on-five expected-goals differentials across a seven-game series are usually much closer than the games-won column suggests, and the teams that win on the strength of overtime results usually have expected-goals splits that look like coin flips. The series narrative — toughness, clutch performance, killer instinct — gets written backwards from the result. The result, in the close ones, was mostly the OT bounce.
Why the league doesn't change it
The honest reason is that the format works as entertainment. A regular-season tie was the format the league used to have, and it produced no drama. Multiple overtime periods until somebody scores produces a kind of sustained collective stress nothing else in sports replicates. The format is genuinely the best moment in North American sports television. There's no commercial argument for changing it. A more "fair" format would be a worse one.
The cost is that the records of certain teams and certain players are partly artifacts of variance the format imposes rather than skill differences anyone could have observed. Several Cup winners over the past twenty years won series in which they were outplayed at five-on-five and survived on OT bounces. Several Cup losers were the better team across the series and lost the bounce. The history books don't have a column for that. The history books only have the score.
How to read OT without getting fooled
Two practical adjustments. First, weight the regulation five-on-five run of play more heavily than the OT result when evaluating which team actually played better. Expected goals from the regulation portion of the game is usually the more honest summary of who deserved the points. The OT goal is the deciding event, but it is not the most informative event about either team. Second, be skeptical of any narrative about clutch performance that's built on a small number of overtime games. The variance band is wide enough that almost any pattern can survive a single postseason. The pattern usually doesn't survive the next one.
The drama of Stanley Cup overtime is real. The information content of any single overtime is much smaller than the drama would have you believe. The format is the best one in sports. The records it produces should be read with a wider error bar than the celebrations after them suggest.