Schuckers and I have done some work on NHL shootouts, presented at the New England Symposium on Statistics in Sports.
The slides for our presentation are linked here.
A github page with R code is found here. The R code links to the data, which is publicly hosted for anyone to analyze.
Finally, we made a pair of interactive plots using Plotly. The shooter interactive plot is linked here, while the goalie interactive plot is linked here.
tl;dr version:
On their own, shootouts aren’t a crapshoot. First, there’s a decent amount of bias with respect to both when shooters are allocated to take attempts and which rounds those attempts are in. Second, shooters, and to a lesser extent goalies, vary more than we would expect them to if every player was equivalent.
All else equal, and since 2005, the best shootout shooter would have been worth about $700k to his team on shootout performance alone, and the best goalie worth about $1,000,000, on a per-year basis. Given the reduction in the frequency of games ending in a shootout going forward, however, these values are likely smaller going forward.
Finally, conditional on what we know about team behavior, shootouts remain much closer to a crapshoot than a sure thing. And they still aren’t a great way to end a hockey game.
StatsbyLopez 