APBRmetrics

italia13calcio and I discussed this in the Mark Jackson thread, but I wanted to post another thread so I could solicit feedback on this idea specifically.

JE already has a RAPM coach rating. As far as I understand it, he adds the coach to his RAPM equation and determines how much a coach is worth. Essentially, he takes the leftover value from the discrepancy between a team's APM and each player's RPAM to find the value of the coach.

This idea is a little different.

Each team would get a cumulative VORP rating. This is just the cumulative VORP of each player added up for the entire roster over a season*. This number would then be adjusted for the aging curve of each player to predict what their VORP for the next season is projected to be. The coach rating be the difference in the value of the VORP between what the team actually achieved and what was projected.

A coach's total VORP rating would then be this number divided by the amount of years they have been coaching that team. The last step is necessary as coaches that have already made their players great in previous seasons would not get credit if you looked at only a one year sample.

RAPM and VORP don't measure a players skill, but the impact they have on the game. Nobody thinks Nick Collison is an elite player, but many argue that he has a huge impact on the game. It is the job of the coach to get the most out of their roster - to have their lineups make the biggest impact possible. It is to Scott Brook's credit that Collison could have such a huge impact on the game, and he should be awarded for this.

Thoughts on this?


*would need to be vorp/per game instead of a player's actual vorp to take into account missed games. Instead of using a one year vorp sample, it might be worth it to use a 33-67% blend of a player's past two VORPs so its more consistent.

Would be interesting to see, not necessarily to provide firm answers but to possibly raise more questions in search of answers. It would test for / find possible coaching impact and variation of individual players from the average aging curve. It would be hard to tell how much was each, especially immediately. Maybe a bit more 3-5-7 years later, maybe not.

RAPM measures the impact a player has on the floor, not their skill. A player's impact is hugely influenced by the construction of their team's roster, their role on the team and the system employed by the coach.

When we look at the Spurs, do we say that Pop was lucky to always have such an amazing roster filled with high RAPM players? or do we say those players are lucky to be on the Spurs where Pop puts them in a position to achieve such a high RAPM?

With a similar roster, Jason Kidd took the Bucks from last place with a 111.8 D rating in 2014 to a fourth place d rating in 2015 of 102.2.

Sure there was some player development, but the impact Kidd had on the team's D and consequently, the players RAPM is too large to ignore.

look at the BPM improvements over one year for these players:

Giannis Antetokounmpo 2.3
John Henson 2.4
Ersan Ilyasova 3
Brandon Knight 3.1
O.J. Mayo 2.9
Khris Middleton 2.9
Zaza Pachulia 2.1
Larry Sanders 3.2

Using the VORPs provided (both Sanders and Knight left the team in the winter, so the numbers should be even larger), this gain alone is equal to a value of 22.68 additional wins (over a replacement level player).

ampersand5 wrote:...
RAPM and VORP don't measure a players skill, but the impact they have on the game...

Is this the VORP that is produced from BPM as seen at bk-ref.com?
If so, how does it measure impact on the game? Isn't it just another boxscore metric?

Does it make sense for a coach to be valued relative to a replacement player?
Should you call it VORC? Or is it relative to an average NBA coach?

Mike G wrote:

ampersand5 wrote:...
RAPM and VORP don't measure a players skill, but the impact they have on the game...

Is this the VORP that is produced from BPM as seen at bk-ref.com?
If so, how does it measure impact on the game? Isn't it just another boxscore metric?

Does it make sense for a coach to be valued relative to a replacement player?
Should you call it VORC? Or is it relative to an average NBA coach?

yup - VORP from BKREF made up from BPM.
While BPM is a boxscore metric, its a bit different in the sense that BPM is just used as a proxy for RAPM. RAPM would be better to use, but the data is much easier to handle with BPM.

If this is to be calculated, the numbers would be in comparison to players, but once this is determined, it could then be adjusted for the average coach.

ampersand5 wrote:With a similar roster, Jason Kidd took the Bucks from last place with a 111.8 D rating in 2014 to a fourth place d rating in 2015 of 102.2.

Sure there was some player development, but the impact Kidd had on the team's D and consequently, the players RAPM is too large to ignore.

look at the BPM improvements over one year for these players:

Giannis Antetokounmpo 2.3
John Henson 2.4
Ersan Ilyasova 3
Brandon Knight 3.1
O.J. Mayo 2.9
Khris Middleton 2.9
Zaza Pachulia 2.1
Larry Sanders 3.2

Using the VORPs provided (both Sanders and Knight left the team in the winter, so the numbers should be even larger), this gain alone is equal to a value of 22.68 additional wins (over a replacement level player).

Just wanted to add a few notes.

1. When looking at gains, we should adjust for age. Younger players develope faster, so you'd want to compare to that.

2. Attributing all to Kidd may not be the right way to do it. How much of it is due to Drew the year before, causing them to develop faster, and is just being realized now? Splitting up the effects between player progression between two years may be tricky.

italia13calcio wrote:

ampersand5 wrote:With a similar roster, Jason Kidd took the Bucks from last place with a 111.8 D rating in 2014 to a fourth place d rating in 2015 of 102.2.

Sure there was some player development, but the impact Kidd had on the team's D and consequently, the players RAPM is too large to ignore.

look at the BPM improvements over one year for these players:

Giannis Antetokounmpo 2.3
John Henson 2.4
Ersan Ilyasova 3
Brandon Knight 3.1
O.J. Mayo 2.9
Khris Middleton 2.9
Zaza Pachulia 2.1
Larry Sanders 3.2

Using the VORPs provided (both Sanders and Knight left the team in the winter, so the numbers should be even larger), this gain alone is equal to a value of 22.68 additional wins (over a replacement level player).

Just wanted to add a few notes.

1. When looking at gains, we should adjust for age. Younger players develope faster, so you'd want to compare to that.

2. Attributing all to Kidd may not be the right way to do it. How much of it is due to Drew the year before, causing them to develop faster, and is just being realized now? Splitting up the effects between player progression between two years may be tricky.

1. 100%

2. I think the more important notion is how bad was Drew, that this is just regression to the mean?

ampersand5 wrote: 2. I think the more important notion is how bad was Drew, that this is just regression to the mean?

Yup, I wasn't even thinking of that, but that's 100% an angle. Seperating the effects of two coaches is tricky. What I was getting at though, is that stuff coaches do can often affect a player's developement, which may not materialize in the season that they are actually coaching but rather in the next season or down the road. It may not be applicable here as Drew probably didn't speed up their development, but it is something to keep in mind. When you have potentially two causes for a player increasing his BPM (better player development from the season before, plus better usage/coaching in the current season) the two effects may be impossible to seperate.

I'm interested to hear what you guys think. Is this a valid concern? Would a player's development not materialize for another season or so?

Sometimes after a coaching change, players may benefit by the better qualities of both coaches.

Of course, the acid test is whether a given coach continues to elicit improvement over several seasons.
Last year, it was Hornacek who seemed to get improved play from almost everyone on his roster. This year, has he done something differently?

Last year, Kidd was a dudd. Maybe he's just better with younger guys?

The Bucks are why you should use a multi-year metric value instead of single season. It's a better estimate of talent that a coach has.

Was Kidd really a dud last year though? Brooklyn was very good after New Years, and after his first few months he seemed to be doing very well.

AcrossTheCourt wrote:The Bucks are why you should use a multi-year metric value instead of single season. It's a better estimate of talent that a coach has.

100%

bondom343 wrote:Was Kidd really a dud last year though? Brooklyn was very good after New Years, and after his first few months he seemed to be doing very well.

I think history shows that Kidd was a good coach in Brooklyn. I imagine prejudicial views influenced people's opinion of him early on (especially the dropping the cup thing).

Also, its important to keep in mind that coaches can
a) improve with experience
b) make a bigger impact with certain rosters/organizations

That's pretty much my sentiments as well. He was put in a bad spot early, but I think he is actually a pretty good coach.