Saturday, November 6, 2010

The Best Players in the NHL?

One thing I've been wanting to do for some time is to compare Corsi QoC/QoT stats versus Corsi Rels. I figure the players who are highest above league average are some of the best, and the guys the most below league average are some of the worst. Courtesy of, I took the 07-08, 08-09, and 09-10 Corsi stats pages and subtracted each player's QoT from QoC and plotted it against the Corsi Rel. I did this for both raw Corsi-based QoC and QoT and also for the Corsi-Rel based measures.

Someone who has a high differential of QoC-QoT faces tough competition with weak teammates, and someone with a high (in the negative) differential  faces easy competition with strong teammates.

In a few days I think I'll end up adjusting for zone starts, but for now, here's what I got. You can find the program I used here.

First, using Corsi-based QoC and QoT:

There isn't much here. I guess there's too much underlying talent disparity in the NHL for any strong, significant trends here. What about Corsi Rel-based QoC-QoT?

Now here's something. Maybe it's a bit circular (I haven't worked out all the logic in my head), so I should have expected a correlation in the first place, but this looks promising. An r squared of 0.43 is pretty decent (r would be 0.656).

Now let's plug in the best-fit equation and get vertical residuals. (Crosses fingers)

The list looks okay. It's not great, but good Corsi players like Mikko Koivu and Sergei Fedorov get near the top of the list and noted poor Corsi players like Jack Johnson are at the bottom. I guess factoring in zone starts would help guys like Chris Drury and Jay McClement get higher. I also guess playing "Corsi hockey" isn't always how those two, and others like Anton Volchenkov, play defense. It'll be interesting to see scoring chance tallies for Drury.

For defensemen, though, I think it doesn't work as well. Douglas Murray shows up several times towards the bottom and Wade Redden shows up near the top. Nicklas Lidstrom shows up a few hundred spots down as well. I guess next time I'll exclusively go with forwards.

On a Caps-related note, John Erskine, Milan Jurcina, Donald Brashear, and Tomas Fleischmann all make appearances near the bottom. Out of 1013 players, Erskine's highest rating is at 874. Yikes. But then again, Jeff Schultz shows up near the bottom as well, as does Tom Poti, so I guess this methodology could use a lot of some work.

I'm having some issues when I try and include team and year, so I'll try and get that fixed (it shouldn't be too hard to figure out that the guy topping the list, though, was in 08-09). I guess in the next installment I'll recalculate Corsi using zonestarts (each offensive zone is +0.8 extra, each defensive zone is -0.8, so I subtract 0.8 for each O zone start and add 0.8 for each D zone start) and get three year averages. I guess when I do that guys like Ovechkin and Crosby will jump to the very top of the lists flukes like Fehr near the top will be eliminated. That'll probably involve me recalculating Corsi ON and Corsi OFF for each team and player. Good thing I have a computer.

Just for the heck of it, here are the numbers using the raw Corsi-based QoC-QoT:

Players known as pure defensive forwards like Drury, John Madden, and Kris Draper are hurt a lot by using raw instead of relative Corsis. Defensemen are probably undervalued. Guys like Ovechkin, the Sedins, Koivu, Parise, and others jump to the top, though, so there is something there, although using a trendline with an r of about 0.2 isn't good technique. Maybe there is something more in Corsi On numbers. But that's another post for another day.

No comments:

Post a Comment