My new player performance formula
So I was up the other night thinking there has to be a good new age way to analyze hockey players, ala Sabermetrics in baseball. All I kept thinking is that Plus/Minus could be a very useful statistic, but it doesn't truly measure the goals a player contributes to his team and directly measure the ones in the other team's favor. So I came up with a formula using a player's goals, primary and secondary assists, powerplays drawn and the effectiveness of those powerplays, giveaways and the results of those giveaways, and penalties taken and the result of those penalty kills. The resulting formula is as follows.
((Goals * .5 (assuming that the player scoring the goal is 50 percent responsible for it's actual being scored)) + (Primary Assists * .35 (assuming the primary assister is 35 percent responsible)) + (Secondary Assists * .15) + (Powerplays Drawn * (League number of powerplays drawn / league number of powerplay goals)) + (Shots on drawn powerplays * (league number of shots on drawn powerplays / league number of powerplay goals) + (goals on drawn powerplays)) - ((giveaways * (league number of giveaways / league number of giveaway goals)) + (shots off giveaways * (league number of giveaway shots / league number of giveaway goals)) + (giveaway goals) + (penalties against * (league penalties against / league penalties against goals)) + (penalties against shots * (league penalties against shots / league penalties against goals)) + (penalties against goals))
In my estimation this formula should measure the true number of goals a player contributes to his team or against is, weighted appropriately against league averages. Unfortunately the statistics I require are not available or recorded or I just could find them, so I went through play by plays by hand of the Stanley Cup Finals and recorded stats from those seven games. The results of my formula are as follows.
<!-- BODY,DIV,TABLE,THEAD,TBODY,TFOOT,TR,TH,TD,P { font-family:"Arial"; font-size:x-small } -->
| Justin Abdelkater 0.9605 |
| Ville Leino 0.6605 |
| Brett Lebda -0.7473 |
| Ruslan Fedotenko 0.5569 |
| Valtteri Filppula -0.1044 |
| Marc Andre Fleury -1.3834 |
| Brad Stuart -0.4334 |
| Brian Rafalski 0.6717 |
| Chris Osgood 0.1322 |
| Evgeni Malkin 1.6803 |
| Niklas Lidstrom -0.7536 |
| Niklas Kronwall -3.6576 |
| Mikael Samuelsson 0.2632 |
| Phillipe Boucher -0.0395 |
| Henrik Zetterberg 1.8202 |
| Hal Gill -2.0148 |
| Jonathan Ericsson -0.9140 |
| Maxime Talbot 2.2195 |
| Bill Guerin -1.7104 |
| Jordan Staal 0.5421 |
| Segei Gonchar -0.9666 |
| Kris Letang 1.1525 |
| Matt Cooke 1.1789 |
| Craig Adams -0.0341 |
| Marian Hossa 0.3079 |
| Johan Franzen 3.1554 |
| Sidney Crosby 0.6894 |
| Brooks Orpik -1.6174 |
| Mark Eaton 0.3765 |
| Jiri Hudler 0.6482 |
| Tyler Kennedy 1.1920 |
| Tomas Holmstrom 1.3554 |
| Dan Cleary 2.3177 |
| Darren Helm 1.4912 |
| Chris Kunitz -2.4622 |
| Miroslav Satan -0.3753 |
| Rob Scuderi -0.9018 |
| Kirk Maltby -0.1580 |
| Mathieu Garon 0.3807 |
| Pascal Dupuis -0.1534 |
| Kris Draper 0.8807 |
The average is somewhere around 0.2. So by my estimation the most effective player was Johan Franzen, the least effective was Niklas Kronwall, the most effective Penguin was Maxime Talbot, the least effective Penguin was Chris Kunitz.
This is only a rough formula come up with at about four in the morning last night. It seems to skew towards forwards and be innacurate towards defensemen, but seems to work well enough for me. The results are also probably pretty skewed due to the extremely small sample size, but I was too lazy to look at any more play by plays.
Let me know what you guys think.
Connor
The content expressed in fanposts does not necessarily reflect the opinions of the staff here at Pensburgh.com. FanPosts are opinions expressed by fans of various teams throughout the league but may be more Pittsburgh-centric for obvious reasons.
0 recs |
11 comments
|
Comments
Wow
That’s a lot of work you put into that. It’s much appreciated. Aside from quibbles about the weights assigned, it’s pretty good.
But I wonder about the volatility over a 7 game series. For instance, I can’t buy Fleury as a -1.3. No doubt, game 5 hurt his score on your scale. Maybe it’s just a goaltender flaw, maybe it’s a problem with that one huge outlier, or maybe there’s some way to discount outlier peformances/cap each game’s effect on the total outcome.
Otherwise, I congratulate you on your fine work….but you may have a bad case of insomnia if you stayed up at night doing this.
by CarlWeathersMustache on Jun 24, 2009 11:46 AM EDT reply actions
Obviously, the outliers would work themselves out over the course of a season. This might just be a problem with using it to analyze such a small sample set.
by CarlWeathersMustache on Jun 24, 2009 11:48 AM EDT up reply actions
Fleury
I do have a reason for Fleury actually, as this formula measures offensive and defensive production without taking goaltending into account. So fleury’s score was cause by the fact that he had some giveaways, at least one of which led directly to a goal, and didn’t contribute offensively. If I was to analyze goaltenders another formula would have to be come up with.
And as for insomnia, school’s out, I’m back on the 3 AM to 1 PM sleep schedule.
by thecheeseisblue on Jun 24, 2009 2:51 PM EDT up reply actions
That’s are pretty genius idea. Well done. Perhaps this can take on the same cult following as Corsi numbers.
Follow the Penguins on SBN @ Pensburgh.com and twitter.
Hockey Sabermetrics
I would encourage you to check out Puck Prospectus and the articles on GVT (Goals Vs Threshold, very similar to VORP) and Fixing Plus Minus.
it's a good idea and great effort
but as you said formula is only an offensive statistic and doesn’t reward any defensive play (just punishes bad defense & offense). so this formula is only really a good measure of offensive production. not defensive. Thus it gives a skewed result against defensive defensemen and goalies (look at scuderi, orpik, and fleruy). So you’d have to come up with a totally different formula for defensive production (although that would be pretty cool to look at how great players were at just scoring and how great they are with keeping the other team from scoring).
2) i think your powerplay weighting should be (pp drawn * (pp goals/pp opportunities) that way you are weighting their pp by the odds that, that pp will turn into a goal. maybe that is what you intended.
3) what i just said, but with everything else when you’re looking at the odds that play will result in a goal. (shots on goal, etc.)
4) I don’t think you should take into account the end result of the what the player did (giveaway, pp drawn/taken) into his number, just the odds that it will lead to a goal. If 2 players on the same team draw a penalty, one shouldn’t be rewarded less than his teammate because the team didn’t score on his pp but scored on his teammates pp. They both gave their team the same opportunity (unless one pp was drawn against someone on the pk and the other wasn’t). For a defensive formula, it may be a good idea to weight penalties taken more harshly for those who are on the pk unit. Or giveaways in their own zone or resulting in odd man breaks….but then again I think this is what you’re trying to take into account by looking at the end result of what the player did (giveaway, pp drawn/taken). But if that’s the case you shouldn’t be adding them together. You should be multiplying that opportunity by the degree of opportunity it is and the chance that the opportunity will lead to a goal.
5) you take into account giveaways (which leads to a loss of puck control, less time on the offensive, and less scoring opportunities). But you don’t take into account takeaways, which lead to more scoring opportunities.
And it might be a good idea to weight takeaways, because a takeaway in the other teams zone probably results in more goals for your team than a takeaway in your teams zone (although it should be weighted the opposite for a defensive breakdown). Also takeaways that result in an odd man rush should be weighted more.
5) it’d be good to list shoots on goal*(yearly # of shots on goals/goals scored)—-aka the shots on goal the player gets weighed by the chance the puck will go in. And maybe if pp shots on goal have a higher percentage of going in, also listing those separately would be a good idea (including ones taken on pp other ppl drew). Or it may be a good idea to weight shots on goals (a dumpin from out side the blue line be less meaningful than a breakaway shot).
6) do you have unassisted goals count as a full goal…for example, you aren’t dividing it by half? Or how do you account for goals with only one assist?
Sorry if I come off as harsh cuz I‘m not trying to be. I’m just trying to improve your idea cuz i’m sure you put in a ton of effort.
i just realized you can’t list shots & goals. cuz if you listed shots on goals weighted based upon the odds that those shots went in AND listed goals, you’d be double weighting things so to say. you can really only look at offensive production in terms of opportunities created & the quality of those opportunites…or offensive production in terms of actual production.
take for instance your weighting of a pp drawn vs goals actually scored by the player:
player A scores an unassisted breakaway goal
Player B draws a penalty against a player on the other teams pk unit, but isnt himself on the PP unit for our team. a goal along with 3 shots on goal occur during the pp he created.
player A gets a 1 for his single handed effort in getting a goal
player B would get:
[the odds that his team’s pp would turn into a goal]3*[the odds that a shot on goal results in a goal]1(the goal that was scored)
so player B would get a lot more credit than player A, even though player A actually scored a goal.
and if it’s supposed to be shots and/or goals on the pp drawn only for that player, then why not include even strength shots on goals. but again, including both shots & goals would be double weighting things.
I read something silmilar to this over at Second City Hockey
http://www.secondcityhockey.com/2009/5/13/874544/sch-community-stat-project-proposal
There’s a link in the fanpost to where the author got his idea from.
Very interesting stuff. If I was a math guy, I’d be able to contribute more to the conversation
Turning women into lesbians since 1991...
Very interesting stuff. If I was a math guy, I’d be able to contribute more to the conversation
Kinda how I feel
Follow the Penguins on SBN @ Pensburgh.com and twitter.

by 


















