November 29, 2006 Hockey stats leave alot to the imagination. Hopefully, not for long. Fans of the game are embarking on a statistical revolution: expanding the limited scope of current player data, and establishing entirely new standards for measuring true on-ice productivity. Greg Wyshynski talks to the man behind, my new favorite resource. hockey fans are being mislead by the current numbers. He calls plus/minus "worthless"

posted by garfield to hockey at 06:16 PM - 18 comments

Please link up to other examples of new-stats. I know there are a ton around, but I have been out of the loop for a few weeks.

posted by garfield at 06:19 PM on November 29, 2006

Great link garf. I'll be back later with some input.

posted by MrFrisby at 06:34 PM on November 29, 2006

Pretty interesting stuff, Garf. I nominate Dr. John to jazz up this fellow's site. As for the new stats, I think he's pretty dead-on. I'd love to see some goalie information based on the area from which goals are scored. Hell, even if you could get a better feel for how much a poor defense impacts a decent goalie. I can also hear Grum oiling up his slide rule -- go Grum go!

posted by wfrazerjr at 07:02 PM on November 29, 2006

Oh, and if you haven't visited this site, get there -- it kicks ass!

posted by wfrazerjr at 07:07 PM on November 29, 2006

Yeah, I just added it to my favorites. Other things that need to be counted in addition to the where the goals are scored scored from: Screened shots, deflections, off of a defenseman, offense to defense ratio (2 on 1, 4 on 4, 3 on 1, etc.).

posted by MrFrisby at 07:52 PM on November 29, 2006

"But Cohen may be hockey's first statistical revolutionary": The Holy Grail of sports stats is to find one number that encompasses everything a player does. In baseball, the most stat driven sport around, there have been a few attempts such as Bill James Win Shares, and "Pete Palmers"í "Linear Weights". The claim in this article for Cohen to be hockey's 1st statistical revolutionary is spurious when the likes of Ray Flowers (THT and Z-Transforms) and Michael Brownstein (Brownstein Index Theorem) and others have been using applied statistical analyst with hockey stats for many years. These so called "new standards" aren't so new after all or perhaps only new to the author. Maybe what is new is Joe Public Hockey Fan's interest in the subject. The categorization of the +/- stat as "useless" is meaningless hyperbole, as it is just one of many variables that can and should be be used in applying statistical analyst to hockey successfully. I do think Cohen is making a valuable contribution to the field but he's hardly the 1st or alone in the quest to add more meaning to the numbers.

posted by skydivedad at 08:40 PM on November 29, 2006

I can also hear Grum oiling up his slide rule -- go Grum go! Surprisingly, I'm not so gung-ho for trying to "play" with hockey stats. The intermingling of player results with each other makes it difficult to cull individual and discrete numbers. If you can't break out individual players from their teammates, it become difficult to determine where the value lies. As well, there are aspects of the game that are almost impossible to categorize (at least, until the advent of some second-by-second play-by-play collation system). For example: last year, Joe Thornton won the MVP. But would he have been deemed most valuable if he played with someone other than Cheechoo, who might not have converted all those passes into goals? Since we can't really "rate" passes that don't get converted into goals (as assists), it seems we'd be missing out on a valuable player. In baseball, if you get a hit but your teammates don't drive you in, you are still going to get credit for that hit. In hockey, that tape-to-tape pass that doesn't get converted into a goal? It's lost in the ether.

posted by grum@work at 09:58 PM on November 29, 2006

Actually, I don't think it's particularly difficult to tell who are the better players in the league outside of the stats - which have seemed to explode in recent years (hits, takeaways, +/-). Watching the games is really all that is needed. In fact, if you look at simply Points per Game - the list you come up with is pretty close to his "most productive" list. Afiniganov, Crosby, Jagr, Vanek, etc. However, I do agree that +/- is a pretty worthless stat to throw at an individual. It's like giving credit to goalies for the amount of goals the team scores in front of them. At best it indicates some situational play - and only indirectly at that. And if he could come up with some convincing way of better measuring goalies statistically, it would be a big help. Save percentage, GAA and wins don't really suggest who the best goalies are. (i.e. Would you take Vesa Toskala over, say Martin Brodeur? Or Ray Emery over Ryan Miller?)

posted by WeedyMcSmokey at 08:12 AM on November 30, 2006

Where are the links, dagnabit?! Glory stats aside, there are players whose value to their team doesn't make it to the score sheet, and some team stats are deceptive, like power play chances : power play goals. One glaring omission from the official record is penalties drawn. ESPN boxscores track it, but are at times inaccurate when compared to your own viewing experience. JABS, along with other team specific bloggers, is making a study of it this season, compiliing a penalty +/-, which in the new whistle happy league is becoming a more significant indicator to team success. There are other angles to add this line of thought. Period of penalty taken/drawn. Zone of penalty taken/drawn. Score when penalty taken/drawn. (I'm just brainstorming on that last one, as that information is available, but not accessible at a glance.) There is the shot chart from's 'gamecenter', which is somehow not a standard statistic. However even this lacks location of the shot on the net. This additional information would help determine a goalie's weaknesses. hockeynumbers is developping a complex shift analysis. Most of the meat & potatos on this site are above my head, but I like the tables. Check out his wishlist. There is IrreverentOilFans and a trippy graph thing and division strength (NE represent!). mc79 has a great ES Strength of Schedule study, among many others. TheMadHockeyBlogger is ambitiously trying to develop a Hockey IQ stat. And you might be surprised by some of the correlations between hits, takeaways, giveaways, etc and team success. And for those that think face off wins is a meaningless stat. I know there is more out there, so share it up, dudes.

posted by garfield at 09:17 AM on November 30, 2006

The best way to know a players worth IS to watch them. That's how I knew to pick up Parise in both of my fantasy leagues before he became NJ's scoring leader. But I can't watch every single game every night, especially the late games on the west coast. So for players on teams I don't watch on a regular basis, all I have to go by is stats. I don't think the +/- stat is totally worthless, but on a scale of 1-10, I would weight it at 3. The problem with it is you just have to be on the ice when a goal is scored, it doesn't matter if you were involved in the play at all. Garf, I had no idea there were other places to get all those stats. I've only been using and for the common stats. The only other site I can bring to the table is

posted by MrFrisby at 09:35 AM on November 30, 2006

Fris, checkout have some nice sorting options.

posted by garfield at 09:45 AM on November 30, 2006

While not exactly a 'stat' site exactly, has some great searching features. I love clicking on a player and seeing his history, including college and euro leagues, etc... And each year for that player is clickable so you can see the teams roster for each year.

posted by myshtigo at 10:17 AM on November 30, 2006

Last week, The Maven posted an interesting stat involving Scott Gomez: Interestingly, with Gomez in the lineup, Claude Julienís team is a mere 5-5-1. In the eight games without Scott, the record is 7-1. Normally that's a stat they like to show during a pregame show when a key player comes back from injury. Do you think a stat like that gives any indication to a players worth?

posted by MrFrisby at 10:29 AM on November 30, 2006

I'd never seen hockey-recap before. This totally explains my serious poolie suckitude. I am totally bookmarking this thread.

posted by chicobangs at 10:33 AM on November 30, 2006

I can't forget to include hockeyrodent, and his "real" stats; categories on the left, tables on the right.

posted by garfield at 11:49 AM on November 30, 2006

After looking at some of these stats, I get the feeling that a lot of coaches would benefit by taking a peak. Some of these guys aren't getting the ice that their productivity would suggest. And some of the guys who are getting the ice have stats that are slightly inflated. But all in all - do you guys see any surprises? I guess maybe the only one I saw was that the death of Glen Murray may have been greatly exaggerated. Other that that - no, it seems the players that we consider to be good are also statistically seen to be good - with a modicum of reshuffling. There just isn't a tremendous disconnect here, so I'm not sure ultimately of the value of more stats. I am however supportive of getting rid of +/-. It's a pretty dumb stat And for the love of sugar-coated Jesus, this "hits" stat has got to be standardized or ignored. I know there is more out there, so share it up, dudes. I'm not so sure you didn't mention them all there, dude. Nicely done.

posted by WeedyMcSmokey at 02:31 PM on November 30, 2006

My candidate for marginally useful statistics are the power play scoring and penalty killing numbers. Expressing these as a percentage of the power play opportunities in which a team scores or successfully defends gives only half of the truth. Frequently, because of consecutive penalties, a power play may last only a few seconds, but if killed successfully, it counts the same as a full 2 minutes. What I'd like to see is a number that represents how long, on the average, it takes a team to score on the power play. Conversely, the same statistic can be applied to show how long a team can avoid being scored upon. It could be done by adding up the minutes and seconds with a manpower advantage (or short-handed) and dividing by the number of goals scored (or allowed). The numbers can be further refined to 5-on-4, 5-on-3, 4-on-3, etc. Granted, this is a team statistic, and not an individual one, but I think it tells a lot more about special teams than the ordinary percentage numbers.

posted by Howard_T at 03:24 PM on November 30, 2006

Stan Fischler had a series of books in the late 80s where he tried to do some of this sort of analysis. As an example, he broke power plays and penalty killing down exactly as Howard_T suggests, into PP time required to score. If you haven't seen them, they're well worth checking out. About the only thing I remember now is the nicknames, though. The Brantford Biped? I still use that. Oh, and plus-minus is a useful stat if you compare it to the rest of the team and adjust based on team goal differential. My Stratomatic card generator did this and its rankings were fairly good.

posted by alex_reno at 05:22 PM on November 30, 2006

You're not logged in. Please log in or register.