Originally Posted by mabus

Did you know that Lidstrom is the only D to finish in the top 5 in D scoring more than twice since the lockout (7 seasons ago)?

Originally Posted by Hey Robbie

Thanks for the translation work, J Mac; in trying to squeeze out my post before dinner burned I did a pretty poor job communicating.



Above is a plot of the HPG Value of the top 20 forwards for 4 seasons ago, 3 seasons ago, and 2 seasons ago. There are fewer than 60 data points because a few players were in the top 20 one year and then were not rated the following year. The vertical axis is the HPG Value of that forward the next year. If the value of a forward in a given season were an excellent predictor of that player's value the next year (eg if having an HPG value of 20 last year meant that a guy is pretty moch a lock for between 15-25 the next year), then the points would follow a line running from lower left to upper right. The more tightly the points clustered around the line, the higher the correlation (Rsquared) is said to be, approaching a perfect Rsquared of 1.0.

Note how there is a rough trend showing higher values one year mean higher values the next year, but that there is plenty of scatter. That scatter is why the correlation comes out to 0.2, closer to 0 than to 1. Those three points in the upper right are all Ovechkin, The lowest one is Malkin losing his value this year due to injury.

Ovechkin's points are examples of the ones I referred to in my original post in saying that a few points exerted a great influence over the correlation. Because Ovie was so much more valuable than even the rest of the top 20 forwards each year, much of the pattern measured by Rsquared was established by his performances. In fact, if we remove Ovechkin from the analysis, the correlation drops to 0.03!

Originally Posted by Hey Robbie

For comparison, this is the plot for the defensemen. Here, there is even more scatter away from the trend, meaning that having a high HPG value one year did not necessarily mean a high value year was coming the next year. The lower correlation of 0.10 represents this shakier predictive value.

If we just take these two Rsquared values at face, er, value, we would say that top end forwards do a better job repeating their top end numbers the next year than top end defensemen do. But there are problems with this that I haven't worked through yet, and am hoping some more noodling on my part and some criticisms on the community's part will help me address. First, as I pointed out with Ovechkin, ae the super-elite the consistent ones, and would we be better off comparing the two classes with these super-elite players removed from the analysis? Without Ovechkin, aren't defensemen actually MORE consistent than forwards? Second, what about injuries? Sixty of each class is a small enough data set that a few freak injuries easily could skew the data. It might be that defensemen are more likely to suffer such injuries, and that is part of the reason for their lack of consistency, but I don't think 60 seasons is enough to say that comfortably.

Originally Posted by mabus

Did you know that Lidstrom is the only D to finish in the top 5 in D scoring more than twice since the lockout (7 seasons ago)? Drafting D early is very risky. Just ask Zubov, McCabe, Phaneuf, Green, Gonchar, Campbell, Streit or Markov owners about what it's like to spend big money early on a top D man coming off a huge season. It's time to wake up and stop drafting D men early - let someone else take the risk.

Posted this on the main page today, and was just thinking this might be a reasonable forum topic. What do you think, am I out to lunch, or is it much riskier to take a D man early?

Mabus