View Single Post
02-07-2012, 06:56 PM
Registered User
OrrNumber4's Avatar
Join Date: Jul 2002
Country: Switzerland
Posts: 8,030
vCash: 500
Originally Posted by Vaasa View Post
I've been trying to stay out of these discussions, but I wanted to address this point. The problem is that you do not understand how statistics work. When talking about stats there is cause, correlation, and noise. For example, when you say things like:

You are confusing a fact, that the fourth line was outshot 29-13 without Murray; to a cause that that reason that the fourth line was outshot was because Murray was not on the ice. The fact is that you simply have no basis for that causal statement.

In order to have any statistical validity in saying that the absence of Murray was directly the cause of the fourth line being outshot the situation in ALL of the games this season, both with and without Murray, would have to be the same. That means the exact same teams playing, with the exact same rosters, with all the players in the exact same condition of health, at exactly the same time, on exactly the same ice surface, with exactly the same penalties called at the same time to same players, etc. To prove CAUSE, you must have the some conditions in a testable manner.

But the problem is that you can't even really infer a solid correlation relationship with those stats. Are those shot counts adjusted depending on opposition, line mates, time on ice, health of the players, and other factors that might skew the numbers one way or another? No, they aren't. And even if could infer a correlation relationship, you might be drawing the wrong one. You seem to saying that Murray is so defensively sound that his presence on the ice reduces the oppositions ability to get shots on net. I could just as easily say that the correlation instead is that Murray is so bad defensively that the defensive pairs are more careful when he is on the ice, effectively reducing the shots against. And that when he is off the ice, they trust the remaining forwards defensive abilities more allowing more shots on goal knowing they have additional defensive support to help clear the rebounds.

So basically, you are quoting all these stats that are basically just noise. An individual stat like shooting %, shots against, or any other stat may have some correlational relationship to player impact when looked at over a broad sample size to a generic player. So I might be able to say that in general, over hundreds of NHL players measured over thousands of NHL games, that when a player is on the ice and the opposition gets more shots than the players team gets shots for, that that represents a correlation to the players defensive skills. The broad sample size helps moderate out factors like player improvement or declines in skills; injuries; differences in teams skills, etc. But the statistics simply do not support drawing those same conclusions on a causal basis when looked for an individual player over a particular (especially a small) period of time.

At best, the statistics can help identify trends that MAY be applicable to certain players at certain times, and provide coaches, scouts, fans, and others things to look for to see if they may be true at a particular time. But in the end, individual observation (as subjective as it is) and memory (as faulty as that is) are more likely to support relevant analysis of an individual players impact that trying to apply such specific stats as you are.

While LW's counter-argument that one can never have a controlled game has merit, the core of this post is great.

OrrNumber4 is offline