APBRmetrics

EvanZ wrote: I'm just looking at GSW as an example. They're power rating (1) is -0.4 and their RAPM-minutes rating (2) or whatever you're calling it is -3.8. Doesn't the fact that the first rating is higher than the second one mean they are *overachieving*?

Yes. How is that different from what I wrote?

Jeff Fogle wrote:Is it okay if I post that in this thread as a way to compare it to other power ratings?

Definitely okay with me.

In all likelihood the market is probably the best team evaluation tool there is. Right now, the market and my rankings seem to disagree the most on Atlanta and Philly

There's a NFL power rating that does regression with HFA for each team. That might be even more useful in basketball

I tried that a couple of years ago and it didn't help at all. Back then I used SRS though (it doesn't square the difference).
Why are those HCAs in the list so high? League average is 3, how can only two teams have less than that?

J.E. wrote: According to RAPM, Indiana, New Orleans, Cleveland, Toronto, Minnesota, Charlotte, Golden State, Sacramento have overachieved. Miami, Oklahoma, Chicago, Boston and Dallas are the biggest underachievers

Eight average-to-bad teams are overachieving, and five expected elite teams are underachieving?

J.E. wrote: Why are those HCAs in the list so high? League average is 3, how can only two teams have less than that?

I've gone ahead and done the ridge-regressed ratings with HCA. The average HCA according to this is 2.83 for last season, which seems reasonable. Maybe my calculations before were off by a factor of 2 for some reason? Not sure. Anyway, I will calculate this for 2012 once more games are played. I found lambda = 160 using 10-fold c.v.
Denver has a huge HCA, which makes sense.

J.E. wrote:

EvanZ wrote: I'm just looking at GSW as an example. They're power rating (1) is -0.4 and their RAPM-minutes rating (2) or whatever you're calling it is -3.8. Doesn't the fact that the first rating is higher than the second one mean they are *overachieving*?

Yes. How is that different from what I wrote?

It's not. My mistake. I mis-read it earlier.

http://stats-for-the-nba.appspot.com/power_rankings

The power rankings now also list (homecourt adjusted) point differential and SRS. BBRs SRS is not homecourt adjusted.
I also list MSE for every metric, starting with games on the 30th

Miami and Oklahoma are underachieving (on a point differential-level). New York, Boston and Dallas are obvious disappointments. Atlanta, Cleveland and Golden State are doing far better than expected.

All metrics agree on New Jersey, Detroit and Washington being horrible, which is probably not a good sign for those 3.
Also, their ridge regressed team rating is probably "too high" because there haven't been many games yet; it's probably gonna get worse over time.
Washington is dead last pretty much everywhere. Wall has shot 10 for 37 with 4.7 TO/G

J.E. wrote:.. Atlanta, Cleveland and Golden State are doing far better than expected.
...

The Cavs are 1-2, projecting to 22 wins, just about what we expected.
But seriously. They've outscored 3 opponents by one point, and those (Tor, Det, Ind) have a combined SRS of -18.

Here's what I predicted (to merge a few threads) and what I now measure in eWins:Sessions off the bench is playing out of his head. Alonzo Gee is suddenly a player. Both Thompson and Irving should get more minutes.

Thanks J.E. I'll try to update these every few days if I get a chance. These are the market estimates at the moment after watching the initial moves in Sunday's games. Hope everyone following the markets feels free to offer up any suggestions if they think I'm off with some of these...

+7: Miami
+6: Oklahoma City, Chicago
+5: San Antonio
+3: LA Lakers
+2: Dallas, Orlando, Denver, Boston
+1: LA Clippers, Memphis, Portland, New York
0: Atlanta, Philadelphia, Indiana
-1: Houston
-3: Milwaukee, Golden State, Phoenix
-4: Minnesota
-5: New Orleans, Charlotte, Sacramento
-6: Detroit, Toronto
-7: Cleveland, New Jersey, Utah
-8: Washington

Still estimating on the Lakers with Bynum back. Denver was bet from -3 up to -5 today, with both teams in a b2b against each other. Considering Lakers were -5 in G1...it looks like oddsmakers initially had the Lakers slightly better...but then bettors stepped in on Denver Sunday because the visitors were at altitude and the home team was trying to earn a split. If Bynum keeps playing like he did Saturday, +3 might stick...

Maybe bettors read my post on HCA:

http://thecity2.com/2011/12/30/ridge-re ... advantage/

Miami and Chicago are head and shoulders above everyone else
The 76ers now sit pretty at 3rd place, leading the league in homecourt adjusted point differential (they haven't played a single home game yet). RAPM also likes them, thanks to lots of minutes for Young in the recent games.
Boston is getting things together since Pierce is back, against rather bad teams though
No metric has the Clippers above average, they're dead last in DRTG. Opponent teams scored 120 points per 100 possessions when Paul and Griffin were playing. They allow the second best eFG% and the second most opponent OReb%. I have a feeling that their low post players suffer from the Javalee McGee syndrome: Not giving an eff about basketball fundamentals because you can jumper higher than everyone else
Cleveland is above or equal 0 in almost every metric. RAPM and Vegas seem to expect them to bounce back to earth. Minnesota is similar. Rubio has looked strong so far
Dallas is still in recovery. Marion looks horrible in +/-, just like he did last year. Odom admitted to being out of shape. It looks like either Kidd or Terry have fallen off a cliff, defensively
Memphis looks disappointing, Conley has been injured

For tonight Vegas has:

-Cleveland by 4 at home vs Charlotte. So Vegas must think Cleveland is worse than what they've done so far, or the Bobcats are better than it seems. The latter is very unlikely, I think. I'm not sure I would expect the Bobcats to lose by less than 4 points against anyone in an away game, except maybe the Wizards.

-Portland losing by 4.5 in Oklahoma (OKC is on a b2b). Portland has looked very strong. My rankings don't expect OKC to win by that large of a margin, even without it being a b2b for them.

Great stuff Jerry.

As a reference, do you know the RMSE of simply using (standard) home-court-advantage as the PD predictor (i.e. not having any knowledge about either team).

Early Cleveland line for Wednesday is +5 at Toronto in a b2b for the Cavs. Still struggling to pin down exactly what the market is thinking about the bottom rung teams in terms of how they relate to each other. I'm honestly not having much time to check out injuries down there, so it could be that some teams drop a point or two if a starter is going to be out.

Most recent guesstimates at market ratings:

+7: Miami
+6: Chicago
+5: Oklahoma City
+3: LA Lakers
+2: Orlando, Dallas, Denver, Boston, SA (dropped without Ginobili)
+1: Portland, Atlanta, LA Clippers, Memphis
even: Indiana, Philadelphia, NY (without Amare, probably +2 with him)
-1: Houston
-2: Phoenix
-3: Minnesota, Milwaukee, Golden State
-5: Detroit, Toronto, Sacramento
-6: New Orleans, Utah, Cleveland
-7: Charlotte
-8: Washington, New Jersey

Spurs opened -8 Wednesday vs. Golden State, so I knocked them down to where that registered on the scale after factoring in 3 for home court. Don't feel good saying all of these rock solid...but I think most everyone is within one plus or minus...and I'd be surprised if too many were 2 or more away from where the market currently has them. A little goofy because of all the b2b's. And, I feel like I've been chasing my tail with Phoenix and NO a bit. Still fluidity in this early stage, but the generalities are mapped out pretty well I think...

bbstats wrote:As a reference, do you know the RMSE of simply using (standard) home-court-advantage as the PD predictor (i.e. not having any knowledge about either team).

That is indeed an important thing to look at, because when any metric is worse than standard HCA, you should probably not use it.

As of Jan 4th, RMSE is calculated for games starting from Dec 31., which is a sample size of 32 games. Each metric gets to readjust their ratings using the latest games. RMSE is as follows:
RAPM derived team strength: 13.3
Ridge Regressed team strength: 13.9
Pure HCA: 14.3
Point differential: 15.5
SRS: 17.5

I'm 99% sure that PD and SRS will be better than HCA once more games are played


Comparing my ratings with Jeff's vegas estimates, it seems Vegas is high on OKC and low on POR/ATL/PHI/CLE

Performing below pre-season expectation, in a point differential sense, is OKC, BOS, DAL and to some extent Memphis (RAPM thinks Memphis is performing as expected, but I think they were generally expected to be above average).
Dallas has obviously been underwhelming, but RAPM says it's not just a fluke or the age of their core players, but also because they have some horrible players in their rotation (RAPM hates Marion and Delonte). Thus, it doesn't expect Dallas to get back to elite level in the West

Performing above pre-season expectation is PHI, ATL, POR, CLE, MIN, TOR, NOH(?!). PHI is a young team with good chemistry, which is probably a good recipe. Gerald Wallace has been great for POR, and even Camby has been healthy so far. Rubio seems to be doing good things in MIN. Casey has Toronto performing at a 105 DRtg, which is good for #21. Considering what they did the years before, this is great. I have no explanation for why Cleveland is so good

The Pacers are outperforming RAPM expectations but everyone else expected them to be that good, I think

In other news, I went 2-0 with my predictions. yay. Glancing over point spreads for tonight, I don't expect SAC, which is on a b2b, to "only" lose by 10.5 in Denver.

J.E. wrote:I have no explanation for why Cleveland is so good
.

They have 2 really good rookies?
At current 48 mpg between them, I have Irving projected at 7 wins and Thompson with 3.
So they're about 10 wins better than they'd be without these 2.

Well, I started yesterday to test my Power Ranking: http://bbmetrics.wordpress.com/power-ranking-2/

So far 18 games are played and the RMSE is 11.82, if I use the point differential given by the rating points in my Power Ranking and an adjustment for HCA (+3). Without HCA adjustment the RMSE is 13.05, only HCA it is 15.19.