Sunday, June 17, 2007

In Defence Of Brind'Amour's Selke

Rod Brind'Amour of Carolina won his second straight Selke Trophy at the NHL Awards for best defensive forward in the NHL. I would have voted for him if I had a ballot and I support his victory.

Some in the blogosphere think that Sami Pahlsson of Anaheim should have won the award. While Pahlsson was a better defensive forward when you combine the playoffs and regular season, I do not believe he was in the regular season alone. Pahlsson was one of the best players in the playoffs, while Brind'Amour was on a team that failed to qualify.

One of the most high profile blogosphere supporters of Pahlsson over Brind'Amour is Earl Sleek from the Battle of California, but he is far from the only one here is a statistical analysis by James Mirtle.

In theory, I believe the Selke Trophy should go to the forward who earns the most wins by his defensive play. I believe that, in theory, we could imagine being able to statistically analyze the contribution of all the players on a team and deduce how many wins they produced in a manner where the total wins of the players on a team strongly correlate with the total number of wins the team has (this would be the hockey equivalent of Bill James's baseball win shares method.

Now there are several problems with this idea in practise. First hockey is not nearly as statistically complete a sport as baseball. Much of the game is not recorded well on the statistical record and thus is hard to recover in any sabermetric theory. This is especially true in regards to defensive play, which the Selke Trophy is intended to reward.

Though James Mirtle made as good a try as possible with his statistical analysis, there are problems with it that are brought in because numbers tend to be averages over the whole season and are subject to the manner that a player is used. If a player is a very good defensive forward, but has little offensive value (at least compared to their teammates) then that play will be used in exclusively defensive situations and be up against a very strong strength of opposition (this is the case with Sami Pahlsson and Jay Pandolfo who were both the 10th highest scorers on their teams). This is certainly not the case with Rod Brind'Amour who was the second highest scorer on his team (and only one point back of first). There were situations where Brind'Amour was on the ice in an offensive role and hopefully against as weak as possible an opposition so that he would be able to score. For the most part, Pahlsson and Pandolfo were not used this way since they lack offensive upside. These offensive situations Brind'Amour was used in were in addition to all the important defensive situations. Brind'Amour had an incredible amount of ice time. His ice time per game of 24:06 minutes placed him third in the NHL among forwards (behind Martin St Louis and Brad Richards of Tampa Bay). It's not much of an exaggeration to say Brind'Amour was always out there. He was on the ice whenever there was a key defensive situation for Carolina and he was on the ice whenever there was a key offensive situation as well. This horribly distorts any average statistics (such as his average strength of opposition).

What about when Brind'Amour was on the ice? There was more even strength goals per minute scored then for either of Pahlsson or Pandolfo. That is entirely true, and related to the fact that both New Jersey and Anaheim had better defences then Carolina and allowed less goals all season. It is also due to the fact that New Jersey and Anaheim have other very good defensive forwards who were usually paired with Pandolfo and Pahlsson. In New Jersey, Pandolfo regularly played with John Madden (a former Selke winner) and Sergei Brylin. In Anaheim, Pahlsson regularly played with Rob Niedermayer and Travis Moen. In defensive situations, Brind'Amour had a less regular set of linemates (though he was usually paired with Justin Williams). Also, the defenceman pairings on the ice were not as good in Carolina's case (afterall Mike Commodore lead Carolina in minutes played on defence). Average goals against rates are strongly linked to the quality of teammates (even when we try to correct for this) and Brind'Amour had lesser quality teammates (which explains why he missed the playoffs).

There is one final stat James Mirtle shows which makes a case for Sami Pahlsson, he played more minutes in 4 on 5 situations than anybody in the NHL. That is directly related to the fact that Anaheim took more penalties then anybody in the NHL. There was more time for him to play short handed because his team was more often shorthanded.

Statistically showing quality defensive play is hard (if not impossible) given the stats collected by the NHL. Any numbers need to be filtered through the way a player was used and to see the effects of his teammates on those numbers. For the most part, any defensive numbers have huge error bars and tend to preferentially select players who play in defense only situations (as opposed to those who play in offensive situations - often against lesser opposition - as well). As long as Brind'Amour is the go to defensive forward in any defensive situation in Carolina, he is clearly a top defensive forward and a potential Selke candidate. We don't have decisive enough statistics to determine who the best defensive forward is (and we likely never will). It comes down to watching games and judging among the top few players for any awards. The differences cannot be reliably pulled from Mirtle's statistics. There is too much lost in the numbers and too much open to interpretation. That is probably the nature of any defensive numbers in hockey. Brind'Amour was a very valuable defensive forward to Carolina. I believe he probably created the most win shares for his team with his defensive play (among all defensive forwards in the league). I cannot show this statistically. I do not believe it is possible to show this statistically. I do believe that the statistics that Mirtle cites to support Pahlsson do not show that he produced the most win shares for his team with his defensive play. There is too much clouded by the way he was used and his teammates.

I think I've said my peace on your other post, but to your points:

--Brind'Amour's TOI. The difference between his TOI and the others' is almost entirely because Rod plays 5:00 of power play time per game whereas the others do not. I don't want to say that it's bad that Rod plays PPs, but I don't know if I'd use it as a reason to suggest that he's a better defensive forward.

--I'm not just on Brind'Amour for his quality of minutes, which I do realize are averages. But given that his minutes were on average easier, you would think that the defensive rates would show improvement in Brind'Amour's minutes vs. the minutes he is not playing. That was not evident.

--Win share I don't think is necessarily the key metric here, but even so, we are talking about a team in Carolina that didn't win enough to make the playoffs or didn't really demonstrate strong defensive play throughout the year.

--I'd fully support Pahlsson over Brind'Amour, but as I said last post, I think Pandolfo's regular season numbers are stronger.
You've got a great blog and I enjoy reading it regularly. I've got a hockey site & i've been trying to promote it lately.

Would you consider doing a review, article, or post a link to it for your readers?

The site is and is all for helping hockey players shop for hockey equipment by telling them the best prices. I don't sell equipment so i'm trying to save fellow hockey players money.

Please consider helping me promote my site but keep up the great work with your blog!

Thanks for everything!

- Cody
Good post. In re: statistics -

Hockey will probably never have as good a statistical set as any of the other three American majors, and especially baseball. Not to say that people can't try to create a better systematic look at hockey numbers, but hockey is a freeform game specifically designed to minimize stops in play.

Baseball, on the other hand, is easily broken down into pitches, batters, innings...
Football stops and lines up after every play...
Even basketball is easier to break down and digest because there is so much scoring, and therefore it's easier to quantify a player's contribution in offensive terms, by points, rebounds, and assists per minute and per shot taken, and so on.

In hockey that's obviously a lot harder. A bad player on a good unit will hurt his teammates; and then there's the goaltender. I vividly remember an Isles/Flyers game from the '01-02 season, where then-Islander Dick Tarnstrom suffered through a terrible performance: giveaways, a lousy penalty, another penalty taken by a teammate to cover a mistake. He was a giant sucking black hole. Garth Snow, however, shut out Philadelphia and that was that.

I can also cite a game against the Devils in which Oleg Kvasha was so bad in the first period that I started leading the opposing fans in chants against him. He was benched after that - so do you go with his very low TOI? Do you emphasize that his giveaway led to a goal? What if DiPietro had robbed the shooter - does that make the play less poor? Statistically it didn't hurt the team as badly...

And so on.
Post a Comment

<< Home

This page is powered by Blogger. Isn't yours?