Foul play: Does PFL even matter?

[kucoach7]

Nathan Sliver
Editor in Chief
sevenfortyseven.com

There has been a whole lot of hubub about personal foul training this offseason and it resulted in our favorite thing here at 747, new data. Commissioner Wiggly Theodore Nosy nonchalantly posted the player propensity to foul ratings and we went crazy here. We had so many questions but most of all we wanted to know, does it even matter?

How does the raw correlation stack up?
First you need to understand our metric. We created a simple standard measure of fouls, fouls per minute by dividing fouls per game by minutes per game. We then started with a simple correlation. The correlation is very significant but a small number that doesn't really mean anything. The main question we wanted to answer was how much variation in fouling does PFL explain? The answer 48.8%.

Okay, that means nothing to me. So what? We did the same exercise with assists per minute and current passing rating, the result? 84.1%. We then repeated the same analysis for offensive rebounds, the result? 84.9% So what does this mean? To put in terms of the debate that has been happening this offseason, it seems that foul training provides a less clear return on investment than other current attribute training.

Relationship status: Complicated
The raw relationship isn't near as strong as other attributes and their corresponding stats so what else is affecting fouling? I used stepwise regression to create the best predictive model I could with the data I had and here are the other things that affect fouling:
Position: Centers, power forwards, and small forwards are more likely than guards to foul.
Weight: The heavier you are, the less likely you are to foul. Not sure about this.
Jump: The better you are at jumping, the more likely you are to foul. Jumping on defense causing fouls makes sense in real life.
Defensive rebounding: The better you are at defensive rebounding, the less likely you are to foul. Again, this makes sense in real life if you think about over the back calls.

Thanks for all that. I don't care. Just tell me how effective foul training is.
Drum roll please. Holding all else equal, improving PFL by five points can result in between .11 and .15 fewer fouls per 30 minutes. This means that for a guy like Boogie Cousins to go from 3.6 fouls per 30 minutes, like he averaged last season, to 3.0 fouls per 30 minutes (and stay on the court longer and get more fouls) an owner would have to invest about 72 points in foul training. However let's remember that other things are working in these guy's favor. As they improve in rebounding and gain some weight, they foul less. DeMarcus is actually a great example of how complicated this is. Overall, he has improved in his career at not fouling. He gained 10 pounds, which helps, but he's a slightly worse rebounder than when he entered the league. Two years ago his rebounding surged and he received foul training but his jumping also increased and his fouls per minute actually went up slightly. Obviously my results are on average, holding all else equal so they aren't going to explain every situation but the results are strong so I hope they can be of use to the league.

[Darth Vegito]

Fantastic stuff as always KU!

Sent from my VS988 using Tapatalk

[IamQuailman]

Just another example how foul training needs to change. Current set up yields a change so minimal that it's a waste (or you need to sink 72 points in a player over 6 seasons, which just isn't feasible).

Sent from my SM-G935P using Tapatalk

[IamQuailman]

Ps - great work todd! Did you actually find correllation between JMP and fouling? Or was that more of your assumption?

Sent from my SM-G935P using Tapatalk

[WigNosy]

Tougher to do would be to control for team pace too.

[kucoach7]

Ps - great work todd! Did you actually find correllation between JMP and fouling? Or was that more of your assumption?

Sent from my SM-G935P using Tapatalk

Yes, I found a correlation there. All findings listed above are statistical results.

[kucoach7]

Tougher to do would be to control for team pace too.

So I was curious about this and tried out the model with team fixed effects to see if I could capture that certain teams foul more (maybe because of defensive scheme) but found no results. That didn't really get at it though because it isn't game level so it can't control for playing a team that gets fouled more or the defense actually employed. I don't have game level info on pace but I do have game level info on defense for my team's games and could do something like that.

[Inner_GI]

You can claim 5 points