Fantastic Post on Hoya Prospectus

[SFHoya99]

www.hoyaprospectus.blogspot.com/

Great post, one (pretty major) quibble on the analysis.

"Moreover, and here's where I may be stretching the statistics a bit, the scatter plot can tell us a bit more: if a player is above the line, he makes the team more efficient than expected based upon Off. Rating (i.e. the player is underrated by Oliver/Pomeroy/etc.) while if he is below the line, he is overrated. Keep in mind that there are considerable uncertainties for the data on both axes that are not shown or even calculated, because that would make my life a lot harder. But it looks like Jessie Sapp and Roy Hibbert were underrated offensively last season, while Patrick Ewing Jr. and Vernon Macklin were overrated."

I'd argue this is faulty on offense and therefore on defense.

The difference between +/- and straight offensive efficiency is that +/- is biased by two things:

1. Who you are on the court with
2. Who you replace when you come on the court

It's really obvious when you have something like "Macklin was overrated by his offensive efficiency and Hibbert was underrated."

I doubt that was true. Macklin's +/- is probably a bit worse than it would normally be because he never got to play with Hibbert, who was obviously the best player on the team last year.

Furthermore, net +/- for Macklin is basically the exact opposite of net +/- for Hibbert -- in short, we aren't measuring Hibbert's or Macklin's effectiveness versus the rest of the Hoyas, we're measuring Hibbert and Macklin's effectiveness solely versus each other.

How it relates to defense is this: to really use +/- effectively, rotation construction probably needs to be taken into account on both offense and defense. Is Wallace's good rating really an effect of his D? Or did he play with Hibbert a lot and his replacements -- Rivers and Wright -- play more with Macklin or Ewing at center?

[SirSaxa]

There is another factor... who are you guarding.
If Rivers came in to shut down the other team's best player (which he often was) that is not taken into account by this statistical analysis.

nonetheless, it is pretty impressive. I could only skim it as I have not taken a statistics course in decades and at this time of night... I come on the board to relax and enjoy some good Hoya chat. I felt like I was taking an SAT or something!

But as I said... well done! I was impressed.
There are a million ways to look at stats, so there will always be critics.
Still, once again, good work!

[CO_Hoya]

Oct 22, 2008 18:02:52 GMT -5 SFHoya99 said:

www.hoyaprospectus.blogspot.com/

Great post, one (pretty major) quibble on the analysis.

"Moreover, and here's where I may be stretching the statistics a bit, the scatter plot can tell us a bit more: if a player is above the line, he makes the team more efficient than expected based upon Off. Rating (i.e. the player is underrated by Oliver/Pomeroy/etc.) while if he is below the line, he is overrated. Keep in mind that there are considerable uncertainties for the data on both axes that are not shown or even calculated, because that would make my life a lot harder. But it looks like Jessie Sapp and Roy Hibbert were underrated offensively last season, while Patrick Ewing Jr. and Vernon Macklin were overrated."

I'd argue this is faulty on offense and therefore on defense.

The difference between +/- and straight offensive efficiency is that +/- is biased by two things:

1. Who you are on the court with
2. Who you replace when you come on the court

It's really obvious when you have something like "Macklin was overrated by his offensive efficiency and Hibbert was underrated."

I doubt that was true. Macklin's +/- is probably a bit worse than it would normally be because he never got to play with Hibbert, who was obviously the best player on the team last year.

Furthermore, net +/- for Macklin is basically the exact opposite of net +/- for Hibbert -- in short, we aren't measuring Hibbert's or Macklin's effectiveness versus the rest of the Hoyas, we're measuring Hibbert and Macklin's effectiveness solely versus each other.

How it relates to defense is this: to really use +/- effectively, rotation construction probably needs to be taken into account on both offense and defense. Is Wallace's good rating really an effect of his D? Or did he play with Hibbert a lot and his replacements -- Rivers and Wright -- play more with Macklin or Ewing at center

Thanks for the promotion.

I'll apologize in advance to everyone else but SFHoya99, since the rest of this discussion would probably be more appropriate over at the blog rather than here.

Your idea that we're measuring Macklin solely versus Hibbert did come to mind while I was working out the numbers. But if that were truly the case, what you'd end up with is that Macklin would show up as underrated and Hibbert overrated in my analysis. This is due to the slope of the line comparing Net Eff. (+/-) and Off. Rating, and a bit of algebra that would be far too involved to cover here. I'd be happy to work it out for you over there if you're really interested.

With regards to your second point about rotation bias, that was a something that I meant to discuss, but it ended up getting forgotten. Briefly here, Off. Rating is assumed to be a teammate-independent stat. Since Off. Net. Eff. is highly correlated to Off. Rating, it appears that the rotation bias is small. In fact, I'd think that the residuals of the fit are as likely to be this rotation effect as my claim of under- or overrating. That's why I said "I may be stretching the statistics a bit."

Since Off. Net. Eff. shows only a relatively small teammate effect, I assume that Def. Net. Eff. would have roughly the same bias. I make this assumption since the underlying math is identical, just using defensive possessions. I have no other evidence to defend this assertion.


My goal was merely to show that, with HD box scores, you can make a quick set of calculations of how well each player performed on offense and defense, one that is defensibly comparable to much more complicated metrics. These stats are also better than traditional +/- stats, since they a) are tempo-free and b) isolate offense and defense, rather than just on-court vs. off-court.

[jgalt]

my brain hurts

[SFHoya99]

But if that were truly the case, what you'd end up with is that Macklin would show up as underrated and Hibbert overrated in my analysis. This is due to the slope of the line comparing Net Eff. (+/-) and Off. Rating, and a bit of algebra that would be far too involved to cover here. I'd be happy to work it out for you over there if you're really interested.

Thinking about it, I think I get your point. Okay, so the idea is that the difference between the +/- system on offense and basic offensive efficiency is (mostly) the effect of teammates because you're basically assuming the basic offensive efficiency is right. If they correlate highly, then then this effect is small.

So that I think I get.

But if Hibbert's +/- efficiency is calculated as a number relative to the offensive efficiency of the team when Macklin is in and vice versa, wouldn't Hibbert's +/- offensive efficiency have been different if Macklin had been a more competent sub?

Or is that offset by the fact that overall team efficiency would have increased?

[CO_Hoya]

Oct 23, 2008 1:42:52 GMT -5 SFHoya99 said:

Or is that offset by the fact that overall team efficiency would have increased?

You got it.

[hoyaboy1]

Great stuff, CO_Hoya. This is very good analysis that you aren't going to find anywhere else.

It also makes me feel good about myself since I always thought Macklin was a terrible defender and that Rivers wasn't nearly as good as advertised on defense.

[SFHoya99]

Oct 23, 2008 7:36:20 GMT -5 CO_Hoya said:

Oct 23, 2008 1:42:52 GMT -5 SFHoya99 said:

Or is that offset by the fact that overall team efficiency would have increased?

You got it.

Just thinking out -- but it seems that if Vern had been more effective -- say, making his FTs -- he would have raised his personal Off Efficiency much more significantly than he would have raised the team's.

So if Vernon had made 75% of his FTs, wouldn't his individual numbers gone up more than the team's numbers would have? Making Hibbert relatively less valuable even though he did nothing different?

I think part of the cause is there is no weighting for PT/usage.

[hoyaboy1]

Yea, my understanding of +/- via 82games.com was always that rotations and sub quality had a pretty major impact.

[joey0403p]

edited Macklin - he is gone

Attempts to circumvent language rules here by abbreviations will not be tolerated.

[prhoya]

;D

Good job, CO. Pretty interesting...

[FewFAC]

I find it difficult to believe that someone on the staff doesn't keep the types of defensive statistics that would provide for more compelling individual ratings, though the system seems rather geared (at least in my mind) to eliminate differences between individual and team offensive ratings, and it seems your data appear to verify such.

Also, I like the average worth plot, because for whatever reason it seems more intuitive to me that average worth is zero-sum, yet the data clearly demonstrate that the players are effectively making each other equally efficient.

Also, I am not entirely sold on the rotation effect, though I acknowledge its inherent faults. Some rotations which, by the numbers, should be the most efficient, appear to only be effective for situational use. Also, the opponent is completely unaccounted for. It is something akin to a batting average versus a batting average against a certain pitcher. But that type of statkeeping remains only a dream for basketball fans.

[kellycpcm]

Hopefully, JTIII considers one of the 3 Arizona recruits who are now looking elsewhere;

sports.espn.go.com/ncb/news/story?id=3667137

Gaddy would be a great pick up for the Hoyas