Statistical studies of National Hockey League shootouts have suggested that, by and large, outcomes are random and no skill is involved. This post on ESPN explored shootout outcomes from 2005-2010, citing St. Lawrence professor Michael Shuckers.
Statistician Michael Schuckers analyzed all 5,711 NHL shootout shots from the tiebreaker's debut in 2005-06 through last season. Of 571 different shooters, not one converted at a significantly higher rate than the NHL average. Meanwhile, only one of 112 goalies outperformed the league save average: the not-exactly-immortal Marc Denis, who is now retired. Just 10 shooters were significantly worse than average, roughly what you'd expect if the results of shootouts were driven purely by chance. "Based on this analysis," Schuckers told me, "I conclude that the NHL shootout is a crapshoot."
If you want more evidence, this article agrees with Shuckers’ conclusions.
In 3-2 shootout victory over Russia, however, American coach Dan Bylsma exhibited a behavior suggesting he does not agree with the idea that shootouts are random. Specifically, Bylsma used Blues forward T.J. Oshie on a remarkable 6 of 8 opportunities (Oshie finished 4 for 6). Here, it’s also important to note that only in Olympic play can players be used repeatedly.
While I’m not sure which side of the “shootouts are random” is right, what is easy to do is explore shootout outcomes for all players and goalies which Bylsma and counterpart Zinetula Bilyaletdinov had at their disposal.
First, I’ll start with offensive players. I grabbed the shootout data for all USA and Russia players (assuming a minimum of 10 attempts in the NHL), and graphed each player’s percentage, along with a 95% confidence interval. Also included are how each player fared in today’s shootout (for the six players with an attempt)
A few things stand out. First, Oshie (54.3%) had the highest career shootout percentage of any player in the game! Of course, it’s difficult to know for sure if Oshie was the best player Bylsma’s could have chosen – but the graph makes it pretty clear that only a few of his teammates boast a similar history of success.
Next, it certainly appears the two coaches used this type of data when choosing players. Specifically, the 6 players used by each team all have career shootout percentages higher than 38.7% (the NHL average, since 2005, is 32.9%). If the two coaches viewed shootouts as random outcomes, we likely would have seen a wider distribution in the shootout percentages of players chosen.
While forwards provided some interesting numbers, I thought the goaltender shootout history was most fascinating. Here, I looked at USA’s three goalies and two more for Russia (I couldn’t find information for Russia’s Alexander Yerymenko). Also, note that in the following graph, we are looking at save percentage, and by looking at goalies (who naturally see more attempts in the NHL), we have much narrower confidence intervals.
For the USA, it’s pretty clear that Bylsma had his choice of three equally talented goalies for the shootout, as Miller, Howard, and Quick each boast career shootout save percentages of around 70 or 71.
With Russia, however, I thought it was interesting that Bilyaletdinov was willing to ride his top offensive players for their great shootout histories (Datsyuk, etc), but didn’t do the same thing with goalies. Specifically, Bobrovsky (65%) was the only goalie in the game with a worse-than-average history, while his backup, Varlamov, has been one of the NHL’s best at this skill (77% for his career). The difference between the two might simply be due to chance, and Varlamov might not be as effective coming off the bench, but it’s at least a decision worth debating.