Tuesday, June 05, 2018
xG in the Stanley Cup
?I've been watching the fantastically presented Corsica chart that @MannyElk has, waiting for the ideal game to talk about this.
One of the popular metrics in hockey and soccer is the concept of "Expected Goals", which is really the shortform of "What is the Expected Value, in Goals, of the Observed Shot Frequency and Distribution". That is to say, the conversion of an empirical shot heat map into a theoretical goal value.(*)
(*) You may recognize this from my post just yesterday on Estimated wOBAcon, which uses the same concept of taking a speed+angle heat map and converting it into a theoretical wOBA value.
Since the Knights were "all over" the Capitals in the first period, but posts and random variation getting in the way, the "empirical shot heat map into a theoretical goal value" score was 1-1, when the ACTUAL score was 3-0 for the Capitals. And near the start of the third period, it was an even wider discrepancy, as the Expected Goals had the Capitals down by 1, while their Actual Goals had the Caps ahead by 4.
But, it's not all random variation to account for the difference. Because once a team takes a shot, there's a goalie there. And what these "Expected Goals" methods do is ALSO say "oh and by the way, let's pretend the goalie doesn't do anything other than randomly do what goalies non-randomly do". Even if it's Tim Howard or Dominik Hasek.
So, this is not a Capitals v Knights situation, but rather Capitals v Knights while ignoring the goalies. Even if the goalies make up a large share of the Stanley Cup MVP (almost one-third, or 16 of 52 Conn Smythe winners are goalies). Therefore, while technically correct, it's important to appreciate that just like FIP is a COMPONENT of pitching, and not ALL of pitching, so is Expected Goals a COMPONENT of team playing. The goalie is also part of the team.
***
Addendum: this is how the implied win% worked out. Naturally, this looks very different from the xG chart, since this includes "goals in the bank". That is, once the Caps score the first goal, their chances of winning goes from 55% to 65-70%, and by the time it's 3-0, they are already "winning" at a 90% clip. This chart shows "Future Win Expectancy given Actual Goals to that point".
?