Regularized Adjusted team ratings

[J.E.]

Lambda ~10. Homecourt and pace adjusted. Does a better job at predicting out of sample point differential than SRS. Playoffs are included in the ranking below

Code: Select all

Miami Heat 	6.26
Chicago Bulls 	5.76
Orlando Magic 	5.64
Los Angeles Lakers 	4.8
San Antonio Spurs 	4.52
Dallas Mavericks 	3.94
Houston Rockets 	3.4
Boston Celtics 	3.34
Denver Nuggets 	2.78
Memphis Grizzlies 	2.68
Oklahoma City Thunder 	2.46
New Orleans Hornets 	1.52
Portland Trail Blazers 	1.24
Philadelphia 76ers 	1.04
Phoenix Suns 	-0
Milwaukee Bucks 	-0.14
Atlanta Hawks 	-0.6
New York Knickerbockers 	-0.86
Utah Jazz 	-1.56
Golden State Warriors 	-1.74
Indiana Pacers 	-1.98
Los Angeles Clippers 	-2.18
Sacramento Kings 	-2.86
Minnesota Timberwolves 	-4.54
Charlotte Bobcats 	-4.64
Toronto Raptors 	-4.96
Washington Wizards 	-4.98
Detroit Pistons 	-5.42
New Jersey Nets 	-5.72
Cleveland Cavaliers 	-6.76

[Crow]

I assume the sum of player RAPM should ideally equal Regularized Adjusted Team Rating. How much variation is there? Would help in finding out if there were team labels on players to allow sorting. And perhaps a listing of minutes played too.

I assume the other players besides Miami's big 3 sum to enough of a negative to pull down the team to about just +6.

Looks like the big 3's sum of 1 year RAPM lost about 40-45% off their 4 year RAPM.

[bbstats]

Interesting...this puts the Celtics way lower than my lineup-level playoff estimate based on RAPM(they're 2nd)

[Mike G]

The one that jumps out to me is Orlando, ranked 3rd, just .62 below #1 Miami.
What are the odds that a team loses 4 of 6 to a team (Atlanta) that's 6.24 worse?

Sagarin has Atl about the same, .72 below avg. But Orl is 7th, +4.49
http://www.usatoday.com/sports/sagarin/nba1011.htm
Includes playoffs.

Houston is #11 according to Sagarin, and #7 here. Best team in a long while to miss the playoffs?

[J.E.]

I assume the sum of player RAPM should ideally equal Regularized Adjusted Team Rating.

That would depend on injuries of players and opponent players, trades etc.

Looks like the big 3's sum of 1 year RAPM lost about 40-45% off their 4 year RAPM.

The ratings just don't reach the same height because there's less data. You will always have that effect on top players if you're working with less data, whether they're playing well now, or not

Sagarin

I don't know how he gets his ratings, so I cannot compare his to mine in terms of accuracy. All I know that the ones here are better than SRS. Also, these ratings tries to minimize error in terms of forecasted point differential, not wins. And while Orlando lost the series, they have a +2 point differential per game for the series

[DavidPerez]

Nice catch!