Soon after we learned Ales Hemsky was done for the year I did a "WOWY" analysis ("With Or Without You") that tried to discern how much of an impact Hemsky's injury would have on the team going forward. At the time I concluded that Hemsky's injury would have a significant impact on the Oilers going forward. Both Bruce and "He Who Shall Not Be Named" (Ender) raised some good objections to the analysis. Neither of them suggested that the team was better off without Hemsky but both of them suggested that the method itself had some serious flaws. One of the objections was that there are so many variables flying around that, even if Hemsky is isolated as the "constant" we can't be confident (let alone sure) that he's the one driving any difference in results. This is a good point.
What I'm doing today is trying to modify the WOWY method (first introduced to me by Tom Tango and Gabe Desjardins) so that it will do a better job of taking this into consideration. I have still isolated Ales Hemsky as the variable. The "without" sample is unchanged in that I've looked at every game that Hemsky hasn't appeared in with the Oilers post-lockout (there are a total of 59). However, for the "with" sample I've only looked at the 59 games that are in closest proximity to the games Hemsky missed. Previously I had looked at all of the games. For example, in 2008-09 Hemsky missed ten consecutive games in the middle of the season. I used those ten games for my "without" sample. For my "with" sample I used the five games before and the five games after Hemsky was out of the lineup. Whenever there was an odd number of games I prioritized the "after his return" portion when possible.
Last time, I looked only at goal differential (no shootout goals included but empty-net goals are). This time I've looked at both goal differential (no shootout goals included, empty-netters are) and shot differential (all situations) to hopefully give us a slightly better look. The results after the jump.
The first graph represents goal differential for the Oilers with Hemsky in the lineup and without him in the lineup. The second graph is the same for shot differential.
In the first piece I measured Hemsky's impact by the difference in overall goal differential. At that time the difference in goal differential was 0.25 per game. In this analysis the number is... similar. Projected over an 82 game season this would suggest Hemsky is worth 20.5 goals above his actual replacements. If we assume that 6 goal differential is equal to one win Hemsky's contribution is 3.4 wins above his replacements which is a big number, one that suggests he's more than worth his contract.
But here's the thing. The really interesting graph is the second one. The Oilers are actually worse at outshooting their opponents with Hemsky in than they are with Hemsky out. Now, in each of the three previous seasons the team was better at outshooting with Hemsky in the lineup. This season, the team is much better at outshooting with Hemsky out of the lineup. The superior goal differential this season is coming on the back of some pretty terrible save percentage by the Oiler rookie goaltending tandem that was only in place once Hemsky went down. On the other hand, when Hemsky was in the lineup this season several other players were ill or injured.
So what of the method? To be totally honest, it looks like Bruce and He Who Shall Not Be Named may have been right. There's just too much going on to really isolate Hemsky as a variable even when we try to isolate him, especially because of just how much variance there's going to be over a 59-game sample. And that's too bad in this case since those goal differential numbers would imply a very good player.