Quote:
Originally Posted by DeluxeMoustache
This leads to a question about what the whole value of the statistical analysis is, though
|
Quote:
Originally Posted by Bingo
It's counting. Counting isn't a model. Summarizing a bunch of counting stats isn't a model, its a summary.
I'm just summarizing that ... it's not a model.
|
When what's the point of all the counting? In the first post you state that you want to be above the red line in each case. FanIn80 points out that Calgary was above the red line in games the Flames lost (one of them a 9-1 defeat) and below the line in games they won handily. There were explanations on why the data didn't match the final score.
So, I will go back to my original question. What's the point of all the counting? I don't mean the question in a snarky way. I actually want to know the point. If all of the counting leads to all of the graphs, and the counting and graphs don't match up with winning or losing in the NHL, then what's the point? It seems the point of counting is just to do some counting. If you could do the counting for the metrics, and then it would be able to reasonably predict something, then there is value in the counting. There should be a cause and effect or correlation from the counting to results.
In your chart there are 17 games The green line is above the red line 14 times. During those 14 games the Flames record was 8-5-1. Of the 3 times the green line was below the red line the Flames won all 3 games. So here's my problem with this stat/metric/count: It doesn't correlate to anything. When the green line is below the red line the Flames have won 100% of the games. The point of hockey is to win games. If I'm coaching a hockey team I want my team to do things that will help us win games. If I see that we win 100% of the games when the green line is below the red line, I want our team doing this more often, because it will lead to more wins. If this is true, then the metric has value. In this case we have been told that the green line being below the red line is bad. We aren't supposed to be down there. We are told we want to be above the red line. However, when the green line is above the red line, Calgary only wins 57% of the games. If you tell me that I can play one way that has shown I have a 57% chance of winning or another way that has a 100% chance of winning, I'm going to go with the way that has a 100% chance of winning. However, what the metrics crowd has stated is that we want to follow the other path. We want to be on the other side of the red line, even though that doesn't lead to more wins. If this is the case, why are we counting? There is no value in counting something if it doesn't help predict something or tell you the story of what happened. It's simply an exercise in counting.
If you require an explanation on why everything is backwards from the output, then methodology of collecting the data in this specific way is flawed. The method of counting individual markers can be correct, but the output of displaying them can still be flawed. Let's suppose you were buying stocks and you believe the best stocks to buy are ones where the company has a P/E ratio of greater than 50% of the highest P/E ratio over the last 5 years. You go through all the companies and you find the companies that fit this metric. You buy these stocks and then notice the companies you've invested in have share prices that continue to decline. You've lost a lot of money and time. Would you still continue using this metric? Of course not. It does not predicatively lead to you picking good stocks to buy even though the raw data and counting is done perfectly. In this case you'd say the output has no bearing on future events, it doesn't explain what happened in the past, and it was simply an exercise in counting.
Why summarize these data points if it doesn't point to anything? Why not find data points that lead to more victories in the NHL? This is why advanced metrics work better in baseball than they do in hockey. In baseball you can find data on how many times a left handed hitter has hit the ball to the left side of the infield against any pitcher. You can use that data to shift, or not shift, the infield's defence to the right side. Of course the batter may still put the ball where you don't want it to go on occasion, but more often than not, the stats can be used as a predictive tool. If they weren't useful in predicting something, they wouldn't be used. Looking at the green line, it's not predicting anything, nor is it explaining anything. It's just a line that doesn't correlate or correspond to anything.