Hello all! You may remember me as that guy with the HAL9000 avatar who constantly makes silly picture comics. Besides shopping photos, I do have a passion for numbers and the ultimate understanding of the games of hockey and baseball, the latter of which you are most likely not interested in because baseball is too SLOW!
Anyway, now that I'm done rambling, on to the show! As you may recall, Steve created a measure called Shutdown Index which basically looks at player performance taking into consideration the situation that the player played in. I liked this metric, but I felt it could be expanded upon to look into overall performance since the defensive side represents a portion of a player's game. As such, I used multivariate regressions to evaluate the top players (among forwards and defencemen) since the 07-08 season at even strength. I created a metric similar to SDI, but for offensive performance, called dCF/20 (differential between expected and actual CF/20). I, then, added dCA/20 (expected CA/20 - actual CA/20) and dCF/20 (actual CF/20 - expected CF/20) to give me delta Corsi per 20 minutes (dCorsi/20).
Like in Steve's study, I factored in quality of teammates (TMCF/A20), quality of competition (OppCF/A20), and defensive zone faceoff% (DZFO%). However, while running the regressions, what I found interesting was how significant NZFO% was for CF/20 for both forwards and defencemen (both in the negative direction in relation to CF/20; maybe this is explained by the higher those two are, the less the players play in the offensive zone), but OZFO% was significant for only forwards and neither OZ nor NZFO% was significant for defencemen.
So, now on to the numbers. I listed them in order of top dCorsi (adjusted for TOI) since 2007-08 season (minimum 50 minutes):
A few interesting surprises, others not as as surprising. It's not perfect (couldn't tease out all the team effects), but it's the best I can do for now. Now my computer is lagging, so I will end here. Hope you found this interesting.
Update: I took Steve's advice and ran the regression analyses where the minimum TOI was 200 minutes, since the players who played less would throw the data off due to SSS. It resulted in much better R^2 values (as high as 0.52) and I updated my spreadsheets (see links above) as a result. As well, I removed the tables because it was causing some lag on my computer, and others seemed to have trouble with the table.