Hey, this is a regression analysis I just did for the last eight years prior to this current year. I'm just starting to do regressions so it would be great if you could give any constructive criticism.
https://dismalcreature.wordpress.com/20 ... n-defense/
Effect of offensive rebounding on defense.
Moderator: Doctor MJ
Effect of offensive rebounding on defense.
-
Jazzfan12
- Assistant Coach
- Posts: 4,294
- And1: 213
- Joined: Feb 07, 2010
Re: Effect of offensive rebounding on defense.
-
Chicago76
- Rookie
- Posts: 1,134
- And1: 228
- Joined: Jan 08, 2006
Re: Effect of offensive rebounding on defense.
There are a lot of murky issues related to ORB and defense that will be very difficult to tease out with a regression, especially when using drtg as y.
Problem 1: we are always working with proxies of defensive performance because so little gets tracked on that end of the floor. DRB%, Age, STL_BLK rate appear to be decent proxies. Of course, these will never capture all aspects of D and they probably aren't linear in nature though, eg, going for a lot of steals to boost this stat comes at a cost, age/experience is good...up to a point, and so on.
Problem 2: Intuitive testing can illustrate that the impact isn't particularly great one way or the other using an overly simplified approach. If the Lakers rebounded on the offensive glass as well as Detroit last year, they would have averaged about 5.5 more ORB per 100 poss. Using actual pt per possession curves based upon first shot taken, we can generically assume that teams will always be able to drag opponents 10-18 seconds into the clock if they don't go for an ORB (and they won't always be able to do this). Also assume that a missed ORB always leads to an early shot. We can also generically assume that maybe aggressive ORB teams will gamble if they feel they have a 40% or better chance at securing the ORB. For the Lakers to get those 5.5 extra ORBs, they'd open themselves up to 9.2 more early shot clock possessions per 100 poss. The defensive cost of each of those attempts is roughly 0.2 pts compared to forcing a team into a middle of the clock first attempt. That's less than 2 pts per 100 poss (or drtg).
I'm not saying the actual impact is that small, but if teams can optimize at multiple ORB% points (an issue for later), then it is that small.
Problem 3: ORB% may in fact be a decent proxy for defensive ability if we could disentangle the transition team strategy component from the "pure ORB ability component". A lot of our best individual ORB rate guys at a given position have been the league's best defenders over time, because ORB gathering requires a combination of athleticism, range, intelligence, and downright nastiness that translates well to defense. So what you think you are measuring (strategy) might be polluted by ORB ability or it may have multicollinearity issues with other defensive proxies.
Unfortunately, I think the only way to begin to test for this is through film/synergy sports kind of data on transition opportunities because it isolates success/failure from so many other possessions buried in drtg. I didn't review the MIT conference stuff, but even this has its limitations:
1-Does it accurately account for not just direct transition opps, but the defensive entropy issues that transition can create on secondary chances? Example: defenders are still scrambling a bit and the ball is brought back out briefly for a shot 9 seconds into the clock?
2-Personnel. It might make sense for one team to shoot a wing in for an ORB chance, but not another. How/when/with whom you attack the glass and vs. what opponent with what players is important. More important than, "they sent 2, 3, or 4 players".
3-Fatigue. ORBs require energy and players have a limited supply. Impossible to quantify this.
I think the important takeaway is that this is a very team specific strategy and one that maybe teams have begun to reassess slightly, but not in a way that would suggest that the optimum strategy should be a 30% ORB rate league-wide the way things were in the late 90s. I find it very hard to believe that with all of the data and video teams have gathered since 1999, that the league average ORB rate dropped 5 pts due to analytical errors. The CV of ORB rates among teams has increased from about .072 in 97-98 to .104 in 13-14, which suggests to me that teams are taking a more nuanced approach based upon personnel at hand. They might have overcorrected a pt or two on average over the last 15 years, but IMO, on average they are more likely to be "more right" about the current ORB philosophy than late 90s teams.
Problem 1: we are always working with proxies of defensive performance because so little gets tracked on that end of the floor. DRB%, Age, STL_BLK rate appear to be decent proxies. Of course, these will never capture all aspects of D and they probably aren't linear in nature though, eg, going for a lot of steals to boost this stat comes at a cost, age/experience is good...up to a point, and so on.
Problem 2: Intuitive testing can illustrate that the impact isn't particularly great one way or the other using an overly simplified approach. If the Lakers rebounded on the offensive glass as well as Detroit last year, they would have averaged about 5.5 more ORB per 100 poss. Using actual pt per possession curves based upon first shot taken, we can generically assume that teams will always be able to drag opponents 10-18 seconds into the clock if they don't go for an ORB (and they won't always be able to do this). Also assume that a missed ORB always leads to an early shot. We can also generically assume that maybe aggressive ORB teams will gamble if they feel they have a 40% or better chance at securing the ORB. For the Lakers to get those 5.5 extra ORBs, they'd open themselves up to 9.2 more early shot clock possessions per 100 poss. The defensive cost of each of those attempts is roughly 0.2 pts compared to forcing a team into a middle of the clock first attempt. That's less than 2 pts per 100 poss (or drtg).
I'm not saying the actual impact is that small, but if teams can optimize at multiple ORB% points (an issue for later), then it is that small.
Problem 3: ORB% may in fact be a decent proxy for defensive ability if we could disentangle the transition team strategy component from the "pure ORB ability component". A lot of our best individual ORB rate guys at a given position have been the league's best defenders over time, because ORB gathering requires a combination of athleticism, range, intelligence, and downright nastiness that translates well to defense. So what you think you are measuring (strategy) might be polluted by ORB ability or it may have multicollinearity issues with other defensive proxies.
Unfortunately, I think the only way to begin to test for this is through film/synergy sports kind of data on transition opportunities because it isolates success/failure from so many other possessions buried in drtg. I didn't review the MIT conference stuff, but even this has its limitations:
1-Does it accurately account for not just direct transition opps, but the defensive entropy issues that transition can create on secondary chances? Example: defenders are still scrambling a bit and the ball is brought back out briefly for a shot 9 seconds into the clock?
2-Personnel. It might make sense for one team to shoot a wing in for an ORB chance, but not another. How/when/with whom you attack the glass and vs. what opponent with what players is important. More important than, "they sent 2, 3, or 4 players".
3-Fatigue. ORBs require energy and players have a limited supply. Impossible to quantify this.
I think the important takeaway is that this is a very team specific strategy and one that maybe teams have begun to reassess slightly, but not in a way that would suggest that the optimum strategy should be a 30% ORB rate league-wide the way things were in the late 90s. I find it very hard to believe that with all of the data and video teams have gathered since 1999, that the league average ORB rate dropped 5 pts due to analytical errors. The CV of ORB rates among teams has increased from about .072 in 97-98 to .104 in 13-14, which suggests to me that teams are taking a more nuanced approach based upon personnel at hand. They might have overcorrected a pt or two on average over the last 15 years, but IMO, on average they are more likely to be "more right" about the current ORB philosophy than late 90s teams.
Return to Statistical Analysis