Tuesday, May 04, 2010

Putting Corsi In Its Place

The playoff version of the 2010 Canadiens are an affront to well-trained hockey stats gurus everywhere. The team is testing the limits of shots and chances allowed vs shots and chances generated and coming out way behind. Yet, they somehow manage to find a positive balance in goals and wins up to now.

Those who've staked the biggest claim in these stats ventures are digging in their heels now. The Canadiens, they say will fall if they continue with the strategy they used to beat the best team in the NHL.

The problem I have with this argument is that it often relies too heavily on one single parameter (shots), and often isn't accompanied by observation.


Once, I was browsing a popular stats site, and there were questions being asked about how the Canadiens could possibly be winning the games they were.

In a frank and simple response, I suggested the stats man should watch the games.

Naturally, in his own backyard, he took offense. I stand by my advice, though. I don't think it's wrong. Although one can indeed prove that shots directed at the net correlates to more shots and shots in turn correlate to more goals and they in turn to more wins. One also shouldn't forget the exceptions at every step. There will always be teams one needs to watch to understand.

Whenever one takes this argument to the number crunchers, one gets accused of cock-eyed bias, of blind fandom. It's certainly true that fans of a team can ignore the bits they don't like and focus on the positive. Interesting as well, however, is the point to which these some-time name-callers invest the same amount of energy and passion as you and I do into a sports team we like to watch into the very numbers they claim to be objective defenders of. I think at some point the hardest fans of shots directed at goal and Corsi numbers have to realize the cracks in the thing they choose to support with blinders.



The problem with Corsi


I like the Corsi number (or the total shots directed on net plus minus, if you prefer). I think it does give something to think about. But it only gives something extra, not something new and complete for us to ignore everything else, such as actual goals, actual play.

One reason I think it is at times and imperfect stat is that it doesn't often account for what the smart thing to do is. For example, a high Corsi number is easily achieved by bombarding the opposition net with shots (even off net or into defenders) when in the opposition zone. I think we and some of Alexander Ovechkin's Washington teammates from the last round might agree that sometimes taking wild shots from everywhere is actually a bad idea. I think even the staunchest Gomez fans might advise him to stop padding his Corsi and concentrate on what he's great at (passing) a little more often.

The same goes at the other end. A team can allow fewer shots by leaving men in prime position open so that passes are attempted. Even though it's probably better to force players outside and clean that first rebound, Corsi numbers sometimes say that a team would be better off on average by tempting a pass into the net.

If you watch the game, this is less of a problem. One watching understands that a breakdown that results in a 2 on 0 with a post (no shot) is more dangerous than a shot from beside the net (even from 4 feet away). It is a problem for those who don't watch, however, as that missed shot (post) doesn't outstrip the close range exercise in futility.


The quality of shot

Today I read that there is no way that Montreal can continue with their strategy of allowing shots from outside or that Cammalleri can continue to shoot at 26%. While I agree, I also take issue with the complete dismissal of the idea.

What shot counts, blocked shots and missed shots don't record is the quality of the shot. Attempts have been made to approximate the shot quality, but they rely thus far entirely on very shaky ground (the distance of shot release, estimated by an NHL recorder). They do not interpret context (i.e., a shot from 50' when the goalie has fallen and the net uncovered is actually a high percentage chance) and without concerted effort to record these efforts live (like Olivier is doing for the Habs) it cannot really be done very completely at all.

Mike Cammalleri could shoot 26% from now if he only took high success rate shots, for example. A team can let up 35 shots 5% success shots from the perimeter if they score enough on counter attack. It's not a strategy I enjoy watching, particularly, but neither is it completely invalid.

If one doesn't know the quality of shot (i.e., watch the game) then one shouldn't really be making catch-all pronouncements on what happened in that game, or series, or season. Ron MacLean take note.


Fatigue

The real chances in hockey come from breakdowns. Breakdowns can be manufactured in the opposition through skill, or they can start happening on their own when players can't keep up.

I've already said that the Stanley Cup playoffs are probably the most grueling test in sports when it comes to a single trophy. So many games over so little time. Recovery time is short and energy systems are tested.

When looking at the prospects of winning this tournament, fatigue has to come in to consideration. Here the question of game strategy and our friend Mr. Corsi come in yet again. As Jacques Martin has designed his tactics for these two series, he seems to me to have had the fitness of his players at least somewhat in mind. I think that's where the strategy of collapsing and allowing certain shots comes in. A player like Gill running around trying to prevent the puck carrier from passing or shooting, or actually trying to strip the puck at every possession will tire him, much more so that waiting for the low percentage shot and the player to turn it over of his free will.

The same theory has been applied in the offensive zone. The reason we are all fretting that the Canadiens aren't sustaining possession is because they aren't really putting any effort into that game at all.

I am a massive proponent of the possession game. I think that holding the puck is better than chasing it, both from a game and a fatigue standpoint. I am, however, also a realist. If Martin thinks his team is incapable of keeping up with our skill match, then I don't find many arguments for that. I think Cammalleri probably would get burned out if he spent all his energy on the boards fighting it out, I think Gill' energies would be mis-spent pursuing with his skating.

While Corsi says a game plan must seek massive shots for, with few against as the ideal, it doesn't really account for contexts where Tom Pyatt is playing Sidney Crosby or Hal Gill is defending Malkin.

I think that while shot management is important, it is not the whole picture. As stats interpreters must consider lots of numbers, so coaches can't get so streamlined in thinking.

Last game, the Canadiens seemed to conserve energy early on with a pay-off of a tie in the first. Going in, maybe not a good strategy, but coming into the second, it's hard to say that it was a loser. From there, they actually visibly wore on the Penguins. A PP goal turned the Pens into more desperate and less thoughtful shooters and the 3rd goal was available because star players were worn out by the tenacity of the Montreal zone defence.

You can say they lost the shot game, the Corsi game. But they won the game in my opinion because they won in other areas of play undetected by stats. One was energy management.

No comments: