APBRmetrics

Crow wrote:Chicago next to last? That seems worth noting and investigating by other means. Perhaps it was affected in part by Rose's absence due to injury, uncontrolled or discussed in the analysis?

No, Rose' injury has nothing to do with that. The issue is that the Bulls had a lot of frontcourt players with high WP48 (Noah, Gibson, Asik, Boozer), but only 96 minutes at PF/C. While they had a bunch of guards/forwards with low WP48 numbers, while having 144 minutes per game for them. Obviously, the Arturo guy didn't think the stuff through which then ends up with the conclusion that Tom Thibodeau is obviously a moronic coach who has no clue how to use the Bulls players. They would have had 100% correlation, if the Bulls had given more minutes to Asik, Noah, Gibson and Boozer instead of giving them to Deng or Rose.

And in that way you can get the result, that the coach of the team, which had the best record in the league twice in the row despite having big trouble with injuries, is actually one of the worst coaches in the league.

For the teams where correlation of minutes was very low with WP like Chicago and Houston, I wonder what the correlations were to other things like winshares per 48, RAPM, age, salary, etc. I might get to looking at that later. If anyone gets to it first, great.

Crow wrote:For the teams where correlation of minutes was very low with WP like Chicago and Houston, I wonder what the correlations were to other things like winshares per 48, RAPM, age, salary, etc. I might get to looking at that later. If anyone gets to it first, great.

Correlation to minutes is misleading here, because if the cast has not an even distributed talent among the positions, the correlation will be per se bad. That is described by my example of the Bulls. And also, even if you just look at the distribution of the minutes among a certain position, you will make the BIG assumption, that the players could have kept their performance level when playing more minutes and more different matchups, while assuming that possible team chemistry issue in terms of possible overlapping skills do not exist.
This method just has too much flaws in order to evaluate coaching, because a coach has to use the available talent and maximize that also with the premise to accommodate things like developing talent, increasing team chemistry or improving the confidence of needed players.

I agree that the simple correlation tests are flawed / misleading. I would want to add a lineup layer to the analysis ideally. The other simple correlations should not be taken at face level far, but it is might still be more useful to check them than not. Then do something more advanced and helpful, as one can identify those methods.

Are some teams really bad on the correlation with WP largely because of their level of priority given to defensive rebounds or their particular shot distribution pattern or are the other reasons dominant? There may be some insights yielded even working with an insufficient method. But whether it is worth it is personal choice and there is room for differences of opinion.

mystic wrote:The issue is that the Bulls had a lot of frontcourt players with high WP48 (Noah, Gibson, Asik, Boozer), but only 96 minutes at PF/C. While they had a bunch of guards/forwards with low WP48 numbers, while having 144 minutes per game for them.

In other words, you just can't do such analysis without taking into account quality of options.
But I wonder, what is be the most optimal way to handle depth? Let's assume player A is slightly better than B, they play at the same position and both are not easily fatigued. How should you distribute minutes among them? 36/12? 30/18? 24/24? Should matchups be a deciding factor?