PlusMinus - direct player comparison

[EvanZ]

So that's what I call 190 possessions, which is really small.

[mystic]

Thanks J.E for the answer.

So that's what I call 190 possessions, which is really small.

Yeah, that is basically what my question should imply. Taking the average pace into account we are talking about roughly 100 minutes of basketball.

Well, it is a nice tool, but I doubt that those numbers are holding much water when it comes to predicting future possessions. We can see that certain players are doing better than others, maybe it can show some chemistry issues for some lineups/players, but I'm hard pressed to see anything more in those numbers. APM is trying to recreate those results, if we want to believe that those 4SPM data is worth a lot more than I implied, we can also go a step further and can use APM.

[Statman]

On Kevin Love -

True, but his impact was much greater than what rAPM predicts (+9.55 better, weighted per Poss)

Not to mention that rAPM should favor Love some in that he usually would be playing more against other team's starters, and not be playing as many "garbage" minutes... the regression would adjust some for that.

[mystic]

Not to mention that rAPM should favor Love some in that he usually would be playing more against other team's starters, and not be playing as many "garbage" minutes... the regression would adjust some for that.

The Net numbers, if I understand J.E. correctly, are also adjusted for the strength of the opponents.

[J.E.]

We can see that certain players are doing better than others, maybe it can show some chemistry issues for some lineups/players, but I'm hard pressed to see anything more in those numbers

It's certainly not meant for predictions. As already stated, I think it's a nice tool when explaining why player A is most likely better (in terms of influence on team points) than player B to someone who does not know regression or chose to ignore RAPM results. It's pretty straight forward and more intuitive. It's certainly useful when trying to make a point like Nowitzki being better than Marion(or Collison over Perkins) because it's a more straight forward way to compare players, without a complicated algorithm behind it

[EvanZ]

Jerry, my only suggestion is to divide the #possessions by 2. I think that is more consistent with the convention people typically use. Even better (but more work) would be to convert to pace-corrected minutes or something.

[J.E.]

Jerry, my only suggestion is to divide the #possessions by 2. I think that is more consistent with the convention people typically use. Even better (but more work) would be to convert to pace-corrected minutes or something.

What are "pace-corrected minutes" or "something"? Why is one attack by team A and a following attack by team B one possession?

I suppose I could test all of the results for significance with a paired T-Test?!

[mystic]

What are "pace-corrected minutes" or "something"?

I guess, Evan wants to see pace-adjusted minutes. Meaning: If player A plays 10 minutes on a team with pace 90, he played 11.1 minutes per 100 possessions or so.

Why is one attack by team A and a following attack by team B one possession?

Well, usually only the offensive possessions are counted, because possession means "possessing the ball". (Zu Deutsch: Possession = Ballbesitz)

I suppose I could test all of the results for significance with a paired T-Test?!

Yes, that would be helpful. And you should really think about adding the overall Net (ORtg-DRtg) for the respective lineups, not just the difference between lineups with player X vs. lineups with player Y. That should lower the amount of confused people.

[EvanZ]

What are "pace-corrected minutes" or "something"?

Something I just made up. Just multiply by 48/(average # possessions in a game). Also, just to be clear once again, it's not that I want it for myself. I'm perfectly fine with what you have. I'm just thinking in terms of what more casual statheads might understand. In fact, I should probably start doing this in my own work.

Why is one attack by team A and a following attack by team B one possession?

Well, they occur simultaneously in time. One team is playing offense, and the other is playing defense. Also, and others can feel free to give their own opinion, but my assumption is that most statheads are thinking the same way I do about it. (I think in American football, this would also be the convention. Is soccer different?)

I don't disagree with your logic, it's perfectly reasonable. Just different. Not wrong or right as far as I'm concerned. It just seems that it's better to stick with the convention being used by others.

[J.E.]

Well, usually only the offensive possessions are counted, because possession means "possessing the ball".

If we only count offensive possessions, everybody in the league was on the floor for exactly 0 defensive possessions last year, because we're not counting those?

Yes, that would be helpful. And you should really think about adding the overall Net (ORtg-DRtg) for the respective lineups, not just the difference between lineups with player X vs. lineups with player Y. That should lower the amount of confused people.

haha

[EvanZ]

Well, usually only the offensive possessions are counted, because possession means "possessing the ball".

If we only count offensive possessions, everybody in the league was on the floor for exactly 0 defensive possessions last year, because we're not counting those?

Well, they were on the floor while the opposing offense possessed the ball. I agree with Mystic that by definition of the word, "defensive possession" is an oxymoron, since the offense is the one with possession.

[Mike G]

Is it any more an oxymoron than "defensive turnover" (turnover caused), defensive FGA (allowed), etc?
At the end of the game, players are often subbed in and out for the offensive possession and the defensive possession.
The defender is not playing a possession for the opponent; he's playing the defensive (for his team) half of a possession.

[Jeff Fogle]

how about hoops analytics settling on the terms:

offense/defense
possession/defended possession
turnover/takeaway
field goals made/field goals allowed

or alternatives others think of.

Seems like we can take care of the oxymoron's...

[mystic]

http://en.wiktionary.org/wiki/possession

No idea, but that is related to the ball. The team with the ball has possession (of the ball). The other team plays defense in order to get possession of the ball.

Nobody says "defensive turnover" and "field goal allowed" is not an oxymoron at all.

When a player is subbed in while his team is on the defensive end, his team has not possession of the ball. If we count both things, we end up counted twice as many possessions as it really are. That's why the pace is roughly 90, not 180. That means a team had 90 possessions (of the ball) in that game, while we can easily assume that the other team had also 90. It is just the case that a "defensive possession" is already counted as the "offensive possession" of the opponents.

[Mike G]

Nobody says "defensive turnover"...

That's good if so, but not long ago we were dealing with someone's use of "defensive offensive rebound".