RealGM

I noticed a discrepancy just now in two statistics that I had heretofore assumed to be the same thing, a measure of a given players' teams' offensive and defensive efficiency with him on the court.

I'll use Corey Maggette, since he's the player whose stats I was viewing (in a curiosity-driven attempt to determine how "empty" his stellar personal stats were relative to his team's success with him on the court ). Per Basketball-Reference.com, his ORtg and DRtg were 116 and 113, respectively, this season. Compare those to the Warriors' overall ORtg and DRtg of 108.1 and 111.7, and I would conclude that Maggette's presence on the court made their defense slightly worse but their offense considerably better, making him a net positive overall.

Flip over to 82games.com, and look at his Floor Time Statistics, which their definition says "represent how the team performed while the player was on the floor." Off 1.08, Def 1.14, which per100 possesions would be 108 and 114.

Which stat am I misinterpreting (or does one site have its numbers wrong)? Did Corey Maggette's team win by 3 points/100 possessions with him on the court, lose by 6 points/100 possessions with him on the court, or something altogether different*?

*I saw one GSW fan attempting use his Net48 of -6.7 (per 82games) to argue that he made Golden State 6.7 points per 48 minutes (in their case, 100.4 possessions) worse than they already were! I'm pretty sure that's not correct.

That is not the same. 82games.com is using the real result for their On/Off court rating (how many points scored the team, how many points scored the other team during the time the player was either on or off the court), while b-r.com is using the boxscore stats and Dean Oliver's formula to calculate ORtg and DRtg. You can get a good approximation of the On court ratings by using Oliver's ORtg and DRtg under the assumption that said player is responsible for 1/5 and 4/5 is the average team ratings.

http://www.82games.com/0910/09GSW13.HTM#onoff

You are using that Off, Def wrong, btw, because these are NOT pace adjusted numbers, but just the real points scored per minute.

Overall the Warriors played actually better with Maggette off the court and the reason for that was his poor defensive performance. He has the worst Net DRtg, which means the difference between the DRtg while him being on the court and off the court is the lowest of all players- -5.6 means the defense gets 5.6 points per 100 possessions worse.

^Thanks for the info.

That still leaves me wondering more, however. What exactly do ORtg and DRtg tell us then? The definitions on Basketball-Reference's glossary page are incomplete and try to link you to Dean Olliver's page for a full explanation, which in turn tries to sell you his book. Does no one know how they are calculated? Are we supposed to just take them at face value, even in cases like Maggette's, where Olliver's numbers say he's great and 82games' numbers say he stinks? Without knowing the actual formulas, how can we explain the difference?

Edit: Are you sure about the 82games ones not being pace-adjusted? The top stats from the link you sent me say per 100 possessions, and even players on very slow-paced, defensive-minded teams have Off, Def ratings over 100. (Example: Gerald Wallace is 1.06, 1.04.)

I've often thought that Offensive Rating should be called Inoffensive Rating. Don't miss many shots and don't turn the ball over much, and your ORtg will be pretty high.

Highest ORtg seasons:
http://www.basketball-reference.com/pla ... i?id=f722A
Topped by Steve Kerr, Fred Hoiberg, and Tim Legler. These are not the greatest offensive players of all time. They simply didn't do much, and so they didn't "waste" many possessions.

ORtg doesn't account for the fact that a guy who isn't very mobile, doesn't pose a threat from many locations, doesn't penetrate or handle the ball much, ... may indirectly make it easier for a defense to shut down a possession.

Kerr was a specialist who now recounts that he was on his way out of the league when his career was saved by the Bulls. Certain others on that team allowed him to be useful, as a 5th option, from deep.

If you multiply ORtg by Usg%, you'll have a composite description of a player's Offense. For some reason, the guys at b-r.com think this is a bad move.

For interpreting b-r Off and Def ratings and making them somewhat transferable I calculate the standard deviation of a teams range in those ratings and then calculate their Z scores. I'm not sure if using the team off or def rating as a median is preferable to using the player median, but either way it is interesting to see where a player falls and I think it eases the comparison value of these ratings cross teams.

Good information, but still no one knows how it's actually calculated?

It seems to be normalized around the team's overall ORtg and DRtg, then uses some combination of each player's personal stats, not +/-, to assign them a higher or lower rating. Am I on the right track here? This stat seems arbitrary and meaningless, if so. That also wouldn't explain how they calculate DRtg, since it would be even harder to base an entire defensive stat on players' personal boxscore stats, since you'd have only blocks, steals, and defensive rebounds to work with.

Need more takes guys.

jicama wrote:They simply didn't do much, and so they didn't "waste" many possessions.

Uh, you know how to interpret that stats, but somehow you wanted to use that as something else. Why? Best offensive players are those who have the best combination of usage rate and offensive rating.

John Doe [MIN] wrote:It seems to be normalized around the team's overall ORtg and DRtg, then uses some combination of each player's personal stats, not +/-, to assign them a higher or lower rating. Am I on the right track here? This stat seems arbitrary and meaningless, if so.

Nope, you are not on the right track. The complete way to calculate the ORtg and DRtg is described in Oliver's book in appendix 1 and 3. And believe me I'm not bored enough to type all 12 pages here. ;)
I can only suggest to read the book, it is a good read, probably the best book regarding basketball statistics out there. Way better than the Hollinger stuff or the books by Winston or Berri.

mysticbb wrote:Best offensive players are those who have the best combination of usage rate and offensive rating..

So how do you do that?
And what do you call it? "ORtg" is already taken.

mysticbb wrote:

John Doe [MIN] wrote:It seems to be normalized around the team's overall ORtg and DRtg, then uses some combination of each player's personal stats, not +/-, to assign them a higher or lower rating. Am I on the right track here? This stat seems arbitrary and meaningless, if so.

Nope, you are not on the right track. The complete way to calculate the ORtg and DRtg is described in Oliver's book in appendix 1 and 3. And believe me I'm not bored enough to type all 12 pages here.
I can only suggest to read the book, it is a good read, probably the best book regarding basketball statistics out there. Way better than the Hollinger stuff or the books by Winston or Berri.

Oh, come on, you've gotta give me something! If it's too complicated to fully explain, can you at least tell me:
Are players' Ratings normalized to those of their team?
Do on/off stats or +/- factor in at all?
Is DRtg a function of opponents' ORtg, players' defensive stats, or neither?
Is ORtg the efficiency of teams' possessions with the player on the court, or only the rate at which said player converts his personal possessions into points? Or neither?

I haven't read the book, but I do know there's no on/off +/- in the ORtg and DRtg seen at b-r.com .

DRtg is basically the team's DRtg, tweaked up or down depending on defensive rebounds, steals, and blocks. The assumption is that 5 players participate equally in all defensive possessions, except for those extra contributions.

ORtg assumes one player determines a possession, that being the player who takes the shot or turns it over. OReb and made shots add to this rating; missed shots and TO diminish it.

These are an attempt to refine what can be seen from boxscore stats. The new wave of stats, adjusted +/- and all that, use play-by-plays to determine points for and against while a given player is on the court.

PBP's do not exist for very many years (6 or 8, maybe?) and so cannot apply to earlier eras. That's one big reason why folks are still digesting the boxscore numbers in various ways.

jicama wrote:If you multiply ORtg by Usg%, you'll have a composite description of a player's Offense.

I decided to do that very thing recently and got a nice ranking of players from it. (LeBron, Wade, Paul, etc.).

For some reason, the guys at b-r.com think this is a bad move.

Do you know why?