Wednesday, October 12, 2011

The Importance of Context

Hockey blogging and stats dredging were born around the same time. Since that time, there has been an expanding following for "advanced" statistics. From the Oilogosphere, the adoption of Corsi, Fenwick, scoring chances and other new ways of twisting the NHL stats record have popped up.

I like to think I was a relatively early adopter of these things. I was excited about the possibilities of Corsi before Toronto missed the playoffs for the 2nd season in a row. I was liking where the future of player analysis was going.

What I've noticed lately, however, is that I am increasingly jaded with the whole lot of them. I still track the numbers and keep a record. I still think there is value to the data. I suppose what I am finding is that I am finding many people to stretch interpretation beyond its bounds. Sometimes simple numbers like shots (on net or off net) are given far too much importance and stretched too thin.

What prompted me to write this piece was a simple segue link from Habs Eyes on the Prize:

In Winnipeg, it was a bit different. The Canadiens scored on their second shot of the game off a steal in the neutral zone. It was a killer play, but in the world of trying to find enough stats in order to create a meaningful sample, it was only a single shot on net, a single positive point towards Corsi. In the context of the game, however, it caused Winnipeg to change gameplan -- something which opened them up to further breakthrough passes.


Remember limitations
You can see the importance of context here. We all watch the Canadiens a fair bit and know that when they get a lead they are not a team that risks much to press for a bigger lead. Their preference is to guard 1-0 or 2-1 than to find breathing room at the risk of leaving that tying goalscoring chance to the opponent.

We may or may not agree with the strategy (my nerves don't agree), but the time for that discussion is later. The point here is that we don't need to leave that observation out of our interpretations of things.

Did Montreal play better against Toronto than they did against Winnipeg? Well, no. If the goal is to win by scoring, the Toronto game for all the defensive poise and nice breakouts did not produce a single seam cutting pass like the Moen PK effort. There was not a rush with a wide open goalscorer getting a clear shot like Pacioretty's goal. There were only shots in volume hoping to go in.

If every win and loss this season follow the pattern of the first two games (unlikely, but bear with me), then let's please not extol the virtues of the positive Fenwick/Corsi man too much. Rather we might need to consider a better way to evaluate a team that buttons up when it's winning and only pours on shots when it's not. Remembering of course that winning is better than not winning.


I still encourage everyone to read up on these stats. I encourage everyone to get their fill of Olivier and Chris Boucher's efforts (and they're big efforts). But if you can be left feeling that the stats do not gibe with what you have just seen, question the statistics as well as observer bias.

The answer to everything has not yet been found. And it's only questioning that will get us there.
"Chris Boucher's analysis shows the Habs played better against the Toronto Maple Leafs than the Winnipeg Jets."

To be completely fair to Boucher, his piece is not one of the pieces I am talking about. He approaches his data like a scientist in a discussion section of a published paper. It's all questions and few presumptions. Sorry Chris that it was your piece that set me off then.

Apology aside now. Let's look simply at the Canadiens first two games.


Stats without context

In the first game the Habs lost to the Toronto Maple Leafs yet outshot them and pretty widely outstretched the Leafs in Fenwick and Corsi. If you only looked at that one element of the game then the Habs were the better team. If the Habs played like this all season, we'd hear how good they were and how they were incredibly unlucky not to score.

In their second game, the Habs were widely outshot, their Fenwick was overall worse than the Jets and so was their Corsi. it was the reverse in terms of these statistics of the Toronto game. The Habs should have lost and if they played like this all season we'd be hearing (like we did in the playoffs of 2010) how lucky the team was and how they didn't deserve anything they were getting.


A dash of context

OK, now how about some context.

In the Toronto game, I would take that above analysis for the first period. The Habs did outplay Toronto and were unlucky (somewhat) not to be winning. They'd have been quite unlucky to be losing. They were a bit unlucky to be losing 33 seconds into the second period, in fact, when they allowed a short-handed goal. But after that, the game changed to the observer. The Leafs seemed to find a new gear and the Canadiens mustered almost nothing of threat for 30 minutes or more. For more than half the game, the Canadiens were in fact the far worse team despite the average Corsi. The NHL record says they had 4 shots to the Leafs 8 in the second period. Olivier recorded chances though as 0 to 12.

No comments: