Do The Thrashers Have Large Talons?

Sunday, September 14, 2008

What Sort of Career Will These Thrashers Have?

One of my dreams is to complete a player forecasting program along the lines of Nate Silver's PECOTA for baseball players. In the PECOTA system you compare a player's performance at a particular age, let's say Tobias Enstrom at age 23. Next you find the defensemen who put up similar numbers at the same age. Then you can see what these "comparable" players did in the next year of their career.

A similar idea is used by the National Weather Service to forecast where hurricanes will travel. They look at past storm tracks, water temperature and wind direction and they are able to produce a prediction. Those predictions are not always right but they reduce the uncertainty about where the storm is likely to travel.

So our question here is this: what NHL players are most comparable to Enstrom, Armstrong, Christensen and Little based on their last 2007-08 seasons? I came up with a list of the following statistics for comparison 1) Even Strength Scoring Rate per hour; 2) Power Play Scoring Rate per hour; 3) Total Scoring Rate per hour; 4) Even Strength Time on Ice; 5) Power Play Time on Ice; 6) Short Handed Time on Ice. Finally I averaged all six categories to produce an overall "most comparable player" for each. Let's get to the results.

Enstrom, Age 23 Season
Most Comparable Player: A. Markov and J-M Liles
ES Scoring Rate: Mattias Ohlund
PP Scoring Rate: Dimitri Kalinin
Total Scoring Rate: Martin Skoula
ES TOI: Sergei Gonchar
PP TOI: Ryan Whitney
SH TOI: Redden, Kalinin

Comments: Enstrom received huge amounts of ice time so if we look at his usage stats his Age-23 comparable players are all very solid defensemen. Enstrom probably received more ice time than he would have on a stronger team, but you might be able to say the same thing about a couple of these guys during their age 23 season. When we look at his scoring rate comparable players we mostly see mostly 2nd tier offensive defensemen. Enstrom's numbers are a bit inflated due to all his ice time so I don't expect him to end up being quite as offensive as say Gonchar, but if his career ends up looking like those of Ohlund, Kalinin and Skoula that's not bad at all for an 8th round pick.

Colby Armstrong, Age 25 Season
Most Comparable Player: Malhotra
ES Scoring Rate: Jason Williams
PP Scoring Rate: Rene Bourque
Total Scoring Rate: Andrew Brunette, Kris Beech
ES TOI: Brendan Morrison
PP TOI: Erik Rasmussen
SH TOI: Brett McLean

Comments: Armstrong's comparable players are a real mixed bag of offensive and defensive players. His upside includes top six players like Brendan Morrison, Jason Williams and Andrew Brunette and his downside comparables include checking line guys like Rasmussen, Brett McLean and Malhotra. Virtually all of comparables have had solid NHL careers--the big question is whether Armstrong will be a 2nd or 3rd line sort of guy going forward from this point.

Erik Christensen, Age 24 Season
Most Comparable Player: Raffi Torres
ES Scoring Rate: Stacey Roest
PP Scoring Rate: Andreas Karlsson
Total Scoring Rate: Tomas Surovy
ES TOI: Oleg Kvasha
PP TOI: Pavel Datsyuk
SH TOI: Kyle Calder

Comments: There's no way to put a pretty face on this one. So far in his career Christensen has been considered a bit of a disappointment and his chief comparable--Raffi Torres--is also a disappointment in his NHL career. The thing that concerns me the most is that Christensen has received a decent amount of ES and PP ice time but put up a rather poor scoring rate in that ice time--so far he has shown the skill you typically see from checking line players. This is probably make or break season for Christensen in terms of securing a career as a scoring line player.

Bryan Little, Age 20 Season
Most Comparable Player: Marcus Nilson
ES Scoring Rate: Marcus Nilson
PP Scoring Rate: Alexei Morozov
Total Scoring Rate: Rotislav Olesz
ES TOI: Daniel Cleary
PP TOI: Marian Hossa
SH TOI: Josef Vasicek

Comments: The fact that Little made the NHL at age 20 and held his own suggests he will have a decent NHL career. What type of career it will be is still up in the air. His scoring rate comparables are not All-Star caliber, at the moment the forecast puts him on the border between a 2nd and 3rd liner. Little played the fewest NHL games of the four I considered and so he might suffer from small sample size issues. The defensive game is already there for Little the question is to what degree we see his offensive side develop in the NHL.

Conclusion: I hope you enjoyed this. This is basically a crude attempt at a more systematic comparison I hope to do in the future. For this exercise I only used a single season's worth of comparables, in the future I hope to set it up to use three seasons worth. Hopefully the larger sample size will narrow down the range of possible career outcomes.


  • Falconer, you are my hero. I love stats. :D

    Just some observations that went through my head while reading all of this...

    What kind of linemates did some of these comparable players have? A guy like Kovy can really boost the numbers for someone like Christensen - Kovy can finish well and he can also set up well - if Christensen can finish better than Rucchin or White, his stats will see a boost above his projection levels.

    Another thing I thought about was experience. Age can be deceiving in that some guys get pushed into NHL action faster than others. For example, looking at Enstrom - were the comparable guys also rookies or did they have a few years of service in already?

    I also have to say I like the choices of categories you looked at - those were good stats to represent individual achievements as opposed to other stats (like plus minus) that can be a reflection of playing on a really poor or really excellent team.

    Thanks, as always, for your excellent data and analysis.

    By Anonymous Sara, at 7:14 PM  

  • Falconer:

    Nice work. You and I need to collaborate on some of this stuff. I think I mentioned to you elsewhere that a couple of years ago I compiled a database of all active players at that time, with annual stats, including age and number of seasons in the league. The intent was similar to what you are working on here.

    I think one key thing you need to consider in this type of comparitive forecasting is not only the age, but how long the player has been in the league. The stats of a 23-year-old rookie will look different than the stats of a 23-year-old who has been in the league for 3 years...

    By Anonymous Jarndyce, at 11:41 AM  

  • re: age factors.

    I will concede that there is a big difference between the way many European players arrive in the NHL and many North American players arrive. Also linemates can make a difference, but perhaps less than you think at ES--those numbers show less season to season variation than say PP rate numbers.

    That said, my hope is to create something that looks like PECOTA in which it used the prior three years worth of data including both major and minor league patterns to come up with a comparability match. Using the top matches it produces a forecast for the coming seasons.

    By Blogger The Falconer, at 12:56 PM  

Post a Comment

Links to this post:

Create a Link

<< Home

Who links to my website?
View My Stats