APBRmetrics

J.E. wrote:
On team specific HCA, here's a ten year difference in home point differential vs away point differential. I think if there is an effect, it should be the same every year*. Thus you can use more years to reduce noise.

*unless it is somehow player dependent. If it is, the effect is probably minor, I would believe

Oklahoma;2.34
Clippers;3.01
Boston;2.06
Orlando;3.35
Memphis;3.12
Minnesota;2.81
Jersey;3.12
Antonio;3.02
Houston;2.64
Chicago;3.36
Philadelphia;2.23
Detroit;2.301
Lakers;3.565
Toronto;3.091
Denver;4.679
Orleans;2.951
Cleveland;4.182
Dallas;3.180
Portland;3.414
Milwaukee;3.59
Washington;3.589
Miami;2.698
York;2.25
Charlotte;3.761
Golden;4.334
Phoenix;3.437
Utah;4.471
Sacramento;4.109
Atlanta;3.696
Indiana;3.9

Interestingly, GM's think the following have best HCA:

1. Oklahoma City -- 44.4%
2. Utah -- 22.2%
3. Denver -- 14.8%
4. Chicago -- 7.4%
Also receiving votes: Boston, Miami, Portland
Last year: Utah -- 46.4%

http://www.nba.com/news/features/2012-g ... _twitter_L

Denver and Utah make the list, but not GSW. And OKC doesn't actually appear to have such a great HCA, nor does Chicago (which only got 2 votes).

The Thunders are 6-1 at home, avg winning by 7.
They're also 6-1 on the road, avg margin is 3.

Mike G wrote:The Thunders are 6-1 at home, avg winning by 7.
They're also 6-1 on the road, avg margin is 3.

Can we both agree that using 7 games home and away is a small sample size?

Yeah, I just thought it was pretty good or lucky to win 6 of 7 by an avg of +3 .
The T-Wolves have a pt-diff of +1.6, but they're 5-8 : 0-5 in games decided by 6 pts or less.

The 7+7 games wasn't being 'used' to declare anything, it just seemed unusual.
I had wondered if the GM poll might be heavy on this year's results.

Mike G wrote:Yeah, I just thought it was pretty good or lucky to win 6 of 7 by an avg of +3 .
The T-Wolves have a pt-diff of +1.6, but they're 5-8 : 0-5 in games decided by 6 pts or less.

The 7+7 games wasn't being 'used' to declare anything, it just seemed unusual.
I had wondered if the GM poll might be heavy on this year's results.

I assumed they did the poll before the season started. Does anybody know?

I love these....it's like secret Vegas numbers.

Jerry, just for clarity (and you may have already answered this) - but is the RMSE for game-by-game differentials?

bbstats wrote:Jerry, just for clarity (and you may have already answered this) - but is the RMSE for game-by-game differentials?

It's the squared difference of expected point differential and actual point differential for every game, since Jan.1 or so

Here's a new batch of personal market estimates. Gets kind of tricky in spots where a star is injuried (a few teams dealing with that right now), and the market may really punish a shorthanded team in a b2b spot in a way that blows the scale a bit (Jazz were -5 vs. shorthanded Clippers Tuesday Night with the Clips in a b2b and on night 2 of a b-b-b).

Will try to throw in obvious drops for marquee guys being out.

+7: Miami (but only +5 missing a star)
+6: Chicago (but only +4 with Rose out)
+5: OKC
+3: Philadelphia (will lift if they thrive vs. class), Portland, Orlando
+2: Dallas, San Antonio at home (more like +1 on road it seems), LA Clippers when healthy (more like even when Paul is out), New York when healthy (more like even when either Carmelo or Amare is out), Denver (maybe only +1 on the road)
+1: Atlanta
even: Indiana, Boston, Memphis
-1: Houston
-2: Minnesota, Utah
-3: Phoenix, Milwaukee
-4: Toronto (more like -6 when Bargnani is out)
-5: Golden State
-6: Cleveland
-7: Detroit, New Orleans, Sacramento
-8: Charlotte, New Jersey, Washington

Market had Philly -5 vs. Denver tonight, with Denver on night two of a b2b. Interested in that one in trying to pin down where those two really stand in the big picture. So few clean looks these days where both teams are are fresh and healthy, and aren't playing the next night.

Hope everyone feels free to suggest tweaks if they think my estimate of what the market is saying is off with any of the teams...

Did the market downgrade the Spurs too much after losing Ginobili? Maybe not, just asking. The loss to Miami brought their point differential since losing Ginobili down to around +2 but it was about twice that before last night.

Hard to say crow. Some might suggest they underadjusted at first:

GAMES JUST AFTER MINNESOTA (Ginobili hurt vs. Minn)
Spurs by 6 over G. State favored by 7.5
Spurs 22 over Dallas favored by 2.5 (big cover)
Spurs by 4 over Denver favored by 4.5
Spurs lose by 12 at OKC getting +6
Spurs lose by 2 at Milwaukee getting +2
Spurs go OT at home vs. Houston at -7

So, that's 1-5 against expectations in the first six games after the injury (and they didn't cover the game with the injury either...and I think Manu went out in the first half). So, the one cover was by a huge margin...but the median performances within the six-game hunk were misses of 1.5 and 4.

Didn't get a chance to see the Spurs tonight. Watched the first three quarters last night. What a Jeckyll and Hyde sequence that one was. Boxscores showing Coach Popovich being very smart with minutes I think....looking at the big picture...

How do OT games count regarding the spread and the RMSE? Does the regulation only counts or also the OT periods?

mystic wrote:How do OT games count regarding the spread and the RMSE? Does the regulation only counts or also the OT periods?

the final result counts

Thought so, just wasn't sure about it.

So far over the last 137 games, the prediction based on the normal Power Ranking has 12.44 RMSE, 103 correct winners and beat the spread 72 times.
The linear model based on that (Power Ranking Difference * 0.9 - 4.7) has 11.94 RMSE, 99 correct winners and beat the spread 80 times.
Vegas has 11.13 RMSE and 101 correct winners.

Mystic, what was the RMSE for last night's games according to your model(s)? How many did you get correct against the spread?

Using an ensemble of several models, I got 7.96 RMSE for the 11 games last night, and picked 9/11 correct ATS.

EvanZ wrote:Mystic, what was the RMSE for last night's games according to your model(s)? How many did you get correct against the spread?

The linear model had 9.53 and 7 ATS. The standard had 10.8 and 5. The experimental (the former "Vegas") had 9.16 and also 7.

EvanZ wrote: Using an ensemble of several models, I got 7.96 RMSE for the 11 games last night, and picked 9/11 correct ATS.

Wow, that is not bad at all. Especially with OKC losing to the Wizard, which I expected to be a blowout between 12 and 21 for the Thunder. The other big contributer was the Celtics winning by 23 against the Raptors, I had them winning by 3.5 to 5.4. The other two losses against Vegas were the Timberwolves, I had them winning by 8.7 to 9.8, Vegas had them by 8.5 and the Nets, I had them losing by 0.5 to 2.9 (but well, the freaking Warriors ).