APBRmetrics

Evan, your ezPM/RAPM blended entries are tied with Crow's for closest to the averages.

The pure ezPM, though, is one of the wilder ones -- just after Vegas and myself (well behind KevinP's) for avg departure.
You are most optimistic on Orl (by 4 W), Atl, Phx (by 6!), and Sac.
Most pessimistic about Por, NY, Min, Det, and Was (by 3) .

DSMok1 has the most favorable view of LAC (tie), Det, and Was (by 2.4).
Least favor for Mia and Chi (narrowly on both).

I seem to like Chi, Ind, Uta (tie), Sac; not so much Okl, Den (by 2.6), Phl (2.4), Mil, Phx, NO.

Vegas loves Okl (by 2.5), Dal, LAL (6.5!), NYK (3.5); not loving Tor (2.5), Cle.

Hollinger likes Den, Mil; not Cha (by 3.5)

J.E. favors SA, Phl (by 2), Hou (tie); disfavor to Ind, Uta, Sac.

Evan's RAPM likes Bos (by 2) and Cle (by 4); dislikes LAC (by 5), SAS, Hou.

dudeimstoked (dis) is very median; propping GSW and dissing NJN.

Crow is favorable to Por, Mem, and Hou (tie); most unfavoring Orl (by 2.5 below vegas and 4 or more below everyone else).

Kevin P leads the 'league' in outlying predictions; he is most hopeful on a bunch of teams with low expectations: Uta (tie), Min (by 2), NJN, NOH (by 4), Tor (by 3), Cha (+4.4); least hopeful for Bos, Dal, LAL (by 2.5), Mem (2.7), Atl, GSW.

With 1/15 of the season played, John Hollinger has made the best predictions relative to current pythagorean (based on 4 to 6 games of point differential) expected wins. He's off an avg of 9.9 per team.
Close by in 2nd-3rd are J.E. and Crow, 10.3 and 10.4

Solidly in last is DSMok1 at 11.7 avg error.
The average of 11 predictions is 10.9 . An 'entry' of last year's pythW also scores 10.9, and 6 of us are worse than that.

My really quick minutes predictions are going to kill my numbers this year.

_________________
APBRmetrics Forum Administrator
GodismyJudgeOK.com/DStats/
Twitter.com/DSMok1

Ha. I'm probably the alltime leader in below-average predictions.
This early report, notably lacking in detail or comprehensiveness, is just a milepost. It's based on little data, strengths of schedule are wildly disparate, etc.

Later we will all feel better that we've all improved from this first report; likely all better than just repeating all teams' 2011 performances. A couple of our entries will beat Vegas'. Someone has to come out on top -- about 90% by luck, perhaps.

Mike, would you be so kind as to refresh my memory on a certain point (what with me being too lazy to look the answer up)? You note that the average prediction error of the 11 participants, 1/15 of the season in, is, well, about 11 games (vs. a pythagorean baseline). Am I off in my recollection of last year (and previous ones as well) that this average error (as well as the spread between best and worst) is approximately what one sees at the end of the season as well?

Edit: I embarrassed myself by my admitted laziness. This is the string summarizing last year's results: viewtopic.php?f=2&t=124&hilit=2011. There are 14 "participants" listed in the table showing results vs. pythogrean. Omitting the entry '41', I calculate the average "error" last year for the remaining 13 to be 6.9.

Recognizing that the same folks aren't playing the game this year and last, for some reason I find it interesting that a reasonable prior is that incorporating the final 14/15 of the season might only be expected to improve average estimate errors by 4 games.

Normal years have tended to be bizarre years, to my immediate memory.

- Last year, 12 entries listed, Hollinger on top with avg error of 5.5, the worst guess was off by 8.4, the median looks like 6.9
- Previous year, 17 entries ranging from 6.67 (back2newbelf) to 8.91 (WinShares). Median 8.1
- 3 years back, just 5 entries, errors from 7.74 (davis21wylie21) to 8.62 (KevinP)

When we look at Pythagorean Wins, we tend to have done better, some more better than others.

After 10 games last night, we all look better.
Errors now range from 9.1 (J.E.) to 10.2 (Vegas).
Just 4 of 11 are better than last year's Pyth.

The season is 10% over (actually 9.8%), and every prognosticator is doing better than the previous year's Py%.
Season records are estimated by games won plus the PythW% of remaining games.
They don't consider strength of schedule to date nor in future.

I don't recall Vegas ever being clearly in the bottom half before, much less in last place. It will probably change but how far?
I gave them a much bigger weight in my preliminary weighted metric than in the past and much more than any other metric. Maybe too much. Looking at 3-5 seasons of past metric performance is pretty thin but that is what I considered. It would be possible to check Vegas' performance back another 5 years at least, with a little more time, but I gave these predictions very little time.

I think I posted those Vegas predictions. I picked those off the web a couple of days before the season started, so those are probably not the ones Vegas had 1 minute before tipoff. I don't know where we can look it up, though

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

Just have to exult in the fact I've surged to the top
of the 2nd division.Season is 12.6% over.

I also hold the distinction of the single worst-looking prediction: That the Sixers will slip to 33-33.
After all, their alleged stars -- Brand and Iguodala -- are just getting older ...An additional 13.8 eWins translates to an additional 27.6 wins, for a 60-6 record.
Their point differential of +15.3 ppg suggests 57-58 wins, given that they've already lost 2.
Philly also represents the worst guess for every contestant here, if their current play holds up.

So to try and summarize and contextualize the horse race, in a typical year (last year apparently not being) we might expect an "average" error of about 8.1 wins vs. Pythagorean predictions. This year, 1/15 of the way in (6.67%), the average deviation of participants was about 11 games. 10% of the season in the bag, the average deviation dropped to 8.7. Add a few more days, with 12.6% of the games complete, and we are already at 8.2.

What's left to anticipate? Asymmetric, unanticipated injuries? Exciting random variation in the last week of the season? It would be interesting to follow, given that the averages seem to be locked in already, whether or not there is much movement in the composition of these averages; whether Philly in Mike G's case stays the outlier or others move in to displace its contribution. Maybe what people say about the NBA game is true about the league at large: all that matters is the very end.

I would think that just like in football (where Buffalo started out super hot and Miami super cold), there are teams right now that are not really as good as we think and others that are not as bad. My guess is Philly comes back down to earth a little bit. And a team like Dallas gets their act together.

_________________
The City: A Golden State Warriors-Centric NBA Blog
NBA WOWY: Find statistics of your favorite team with any arbitrary combination of players on or off the court.

For each team, there exists an average of 10 predictions. Should each team actually win that number of games -- to the nearest integer -- the average error of participants would be 3.65. So I doubt there's any year end distribution of wins that could yield a smaller avg error.

Just as the strengths of schedule have in just a few days all gathered into the +/- single digits, it's reasonable to suppose teams will settle into their season's performance level. And if there were no surprise teams, it would be shocking.

Still, it's a bit dismaying how many regularly bad teams continue to be bad -- or even worse -- even those you'd think could hardly get any worse. Right now, Cha is headed for 13-53, Det the same, GS 19-47, NJ 12-54, Sac 17-49, Was 8-58 !