SSAC Paper: Parker, Ginobili and Duncan perfect trio, Pop is the best coach

[timvp]

The annual MIT Sloan Sports Analytics Conference is currently taking place, which is basically a bunch of people discussing advanced stats and different ways to analyze basketball (and other sports). Anyways, each year there are some interesting research papers. So far this year, I found this one interesting and especially Spurs related:

http://www.sloansportsconference.com/wp-content/uploads/2012/02/Robert-Ayer-Sloan-Sports-Analytics-Conference.pdf

Basically, it's a research paper on team construction. It looks at various types of players and aims to figure out which types of players fit best with each other. In doing so, the research also ended up ranking NBA coaches.

Not to spoil the results but Pop grades out as the best coach. And the research shows the best combination is a dominant bigman surrounded by a score-first point guard and a wing that can score, hit threes and rack up assists. That sounds quite a bit like a certain Big 3 who has won some games around these parts . . .

[DBMethos]

Man, if only the stats-based articles that I read in my own field were this compelling.

[DesignatedT]

Pop grades out as the best and it isn't even close.

[Fabbs]

Pop "grading out" as the best coach is far better then actual playoff wins, we all agree on that. Pops is hands down better then the following:
2008 Phil J
2009 Rick Carlisle
2010 Alvin Gentry
2011 Lionel Hollins

[timvp]

^MultiFabbs fabbing the fabs he's supposed to fab. No fabbrises here.

[jjktkk]

Pop "grading out" as the best coach is far better then actual playoff wins, we all agree on that. Pops is hands down better then the following:
2008 Phil J
2009 Rick Carlisle
2010 Alvin Gentry
2011 Lionel Hollins

Always good to hear from one of the top posters on here. Keep em coming Flabbs.

[therealtruth]

That's why he's being winning so many championships.

[Arcadian]

That is awesome. I'm currently learning regression techniques in a graduate statistics class. I should show this to the professor!

[Fabbs]

That's why he's being winning so many championships.
multiTimvp and schtikk know it. :rollin

[Fabbs]

Not to spoil the results but Pop grades out as the best coach.
What year(s) was CIA Pop "grading out" as "the best"?

[Spurs7794]

Pop "grading out" as the best coach is far better then actual playoff wins, we all agree on that. Pops is hands down better then the following:
2008 Phil J
2009 Rick Carlisle
2010 Alvin Gentry
2011 Lionel Hollins


Wait so with this system Pop is both worse and better than PJax and better and worse than Rick Carlisle? Interesting. Is Doc Rivers also better than Phil Jax? But then Phil Jax is better than Doc Rivers from 2010. And Stan Van Gundy is better than Doc Rivers in 2009. But PJax is better than Stan Van Gundy also in 2009. Then how is Doc Rivers better than PJax in 2008?

Pretty conclusive logic.

[Snow]

Pretty cool stuff. He seems to make it a little confusing by listing cluster-8 as multi-faceted, high scoring wings, with high assists for their position and are great 3 point shooters (Paul Pierce, Danny Ainge, MANU) then in his regression results and conclusion call his cluster-8 small fowards which threw me off.

[cantthinkofanything]

Always good to hear from one of the top posters on here. Keep em coming Flabbs.

That sig's funny enough to not use the Photoshop version.

http://www.anvari.org/db/fun/Misc/News_Anchor_Rapist_Search.jpg

[temujin]

The key element is the so called team talent.

Which is measured as the sum of individual talent.

The individual talent is based on an "efficiency" formula used on NBA.com2: ((Points + Rebounds + Assists + Steals + Blocks) - ((Field Goals Att. - Field Goals Made) + (Free Throws Att. - Free Throws Made) + Turnovers)).

This is highly biased team-wise. A team with a high ratio of assists per turnover, for exmple, will have more points (easy layups, open jumpers), better percentages, even more FTs. Similarly, a team with a focus on defense will have the same as above plus rebounds, steal etch. On average, each player.
So essentially, a personal talent is highly biased by the SYSTEM in which that person is performing.

This makes pretty much a circular argument: efficiency is not a pure parameter, but highly influenced by the team system: the system influences the talent, enhancing it no matter what, on average.
If team talent, PURE talent, irrespective of the SYSTEM, cannot be measured, this whole analysis falls apart.
basically, unless one can measure pure talent (easier said than done) this thing just confirms that:
I. Thoma + Pierce + Olajuwan (Robinson)>> Webber, Gasol and Mark jackson (Avery Johnson).

[temujin]

The fact that Pop comes on top of the coaches, if anything, confirms, in my opinion, that this is a rating of a system (defense plus fewer mistakes) more than individual combinations.

[jesterbobman]

I can't support any paper that uses the NBA Efficiency formula as a basis of player value, it's terrible. On a team level, the team may well be accurate(As it is summing good values and bad values) the rating of individual players is terrible. It's also not a scientifically based approach, and arbitrarily assigns values as being equal(In reality, 1 Block is not equal to 1 steal).

On the Spurs front, not shocking that they do well, the spurs FO is very smart, I'm fairly certain that they have a regression based model in place with picking players, though they seem to avoid it occassionally(Corey Joseph wouldn't show up well at all.)

[temujin]

I can't support any paper that uses the NBA Efficiency formula as a basis of player value, it's terrible. On a team level, the team may well be accurate(As it is summing good values and bad values) the rating of individual players is terrible. It's also not a scientifically based approach, and arbitrarily assigns values as being equal(In reality, 1 Block is not equal to 1 steal).

On the Spurs front, not shocking that they do well, the spurs FO is very smart, I'm fairly certain that they have a regression based model in place with picking players, though they seem to avoid it occassionally(Corey Joseph wouldn't show up well at all.)

I agree.

[spectator]

i would like to see his research results with PER, instead of EFF. however, i think that would not offer conclusive results.

[RuffnReadyOzStyle]

Interesting stuff, thanks timvp. :tu