So in part 1, I looked at some data and made some interesting findings. One, that teammates have a far greater impact on performance than opponents. Secondly, that defensive performance is impacted greater by quality of teammates/opponents than offensive performance. And finally, that forwards seem to have greater control over their performance than defensemen.
In the comments, I got some great suggestions for future work. One commentator suggested looking through the same lens at defensemen, comparing offensive and defensive defensemen.
So I thought the best way to do this was looking at zone starts. While this is not a perfect measure, in this situation I have to trust that in general coaches put their offensive defenseman out for more offensive draws and put their defensive defensemen out for more defensive draws.
This isn't a perfect system, and as a result could affect the data and results.
So what I did was calculate each players adjusted zone start % (player zone start% - team zone start%) at even strength 5v5 play in the offensive, defensive, and neutral zones.
I then separated the defensemen into two categories. One group was players with a greater than or equal to 1.5% adjusted offensive zone start % and the second group was players with a greater than or equal to 1.5% adjusted defensive zone start %. This gave me about 300 players in each group (note: there were a few players who resided in both categories, but not enough to really affect the data meaningfully)
So hopefully, I have broke up players into Offensive defensemen and defensive defensemen based on their zone start%.
Before we look at the results, I am going to paste the results for defensemen from the previous post for reference.
So let's look at the results shall we.
So first let's look at offensive performance for these defensemen.
I am going to hold back on the graphs in this post, only because it is time consuming to take the steps to upload the graphs to a website and than link the graphs to this fan post.
So this is pretty close to what we see in the first chart. Nothing much to see here.
On to defensive performance for these guys
Again, this lines up pretty similarly to what we originally saw for all defensemen. The R^2 values are definitely close enough that any difference could easily be accounted for just because we descreased our sample size from over 1,000 defensemen to about 300 defensemen.
Let's take a look at offensive defensemen
First, let's look to offense again.
The effect of teammates seems to be pretty similar to defensive defensemen. Where we see a see a little difference is the affect opponents have on offensive performance. We are seeing about a 7% difference in explained variance by quality of opponents between offensive and defensive defensemen. If this difference in explained variance is true, we are seeing that perhaps offensive defensemen have a bit more control over their offensive performance, which makes sense.
Now for defensive performance.
Again we are seeing some slight differences, both with the effect of teammates and opponents. Opponents are explaining about 10% more variance on the defensive performance here then what we saw with defensive defensemen. This points towards offensive defensemen having perhaps slightly less control over their defensive performance then defensive guys.
So what can we pull from this data. Well to be honest, I'm not exactly sure whether or not those differences in R^2 values are large enough to tell us anything. Any statistical savvy commentators can let me know that below. But here is all the data that many of you were asking for after my last post.