Predictions 2014-2015

[Mike G]

SIXERS WIN!
Phl 85 - 77 Min

Two nights after losing by 6 to a Spurs team that rested Duncan and Parker, Philly outscored a Wolves team that was without Rubio, Martin, and Pekovic -- 3 of their top 6 players.
And the win was in Minny.

B-R.com now expects Sixers (1-17) to win 19 games. Wolves (4-13) at 30 and in free fall.

[permaximum]

Injuries hurt Minnesota really bad. I thought they could push 35-40 wins at the beginning of the season.

[Crow]

I hope the metric comparison gets done and even done in another way if there is disagreement on approach and validity. There was a retrodiction thread in past with a long build up and I don't recall an outcome. It would be a shame if that happened now. I am encouraged that several capable metricians have engaged and are willing to share to some degree of their work.

[Mike G]

This is not the betting thread, nor is it the BPM/comparison testing thread; but it may be a convergence of sorts.
A person betting on the season, simply by averaging (with equal weight) all the apbrmetrics predictions, would be vying for the lead here. And ahead of Vegas.

AJ(1 of 2), Crow, and mystic are ahead of that average, but not by much -- according to the B-R.com forecast.
We can weigh predictions according to the previous seasons' results, and perhaps get even better 'blends'.

[Crow]

If you feel like it, add an equal weighted blend of all entries to your scorecard. Might be worth seeing if I added any value above that.

[sndesai1]

i believe you do (when i checked a week ago, i remember you beating that prediction by a not insignificant amount, although that entry would still be one of the better ones)

i am happy to add the entry if there is some agreement on which entries to include in the average (all, all other than vegas, all that aren't from major websites like espn/538?)

[Mike G]

We're at roughly 27% of the season.

Code: Select all

avg error  14py-err      avg error  14py-err
AJb1   5.7    3.5        AJb2   6.8    2.3
crow   6.0    3.1        bobb   7.0    2.1
myst   6.2    2.9        itca   7.0    2.1
atc    6.4    2.7        DrP    7.1    2.0
bbs    6.5    2.6         eW    7.2    2.0
v-0    6.5    2.6        fpli   7.5    1.6
ncs    6.7    2.4        snd1   7.5    1.6
HDon   6.8    2.3        14py   9.1     .0

In the 8 days since last posted, very little has changed in the order or the magnitudes of projected errors.
Yet we have all added some separation from last year's pythagorean (14py).
The group average now projects an avg absolute error of 6.24

[AcrossTheCourt]

crow's still holding a strong lead
the pyth + 0.500 entry is dropping/has dropped down to the bottom
as a group, 5 of the biggest misses so far are: OKC, SAC, CHA, TOR, MIN

Update?

[sndesai1]

as of 12/22


crow still with strong lead, although he's lost some of it


edit: i'll add in the average later within a couple of days

[Crow]

7 ahead of Vegas. Last season 4 finished ahead, the most in previous 3 years.

Compared to 538, I think I will do better in the west for full season. East, which I've admitted knowing and care less about, might be toss-up to slight edge to 538. (Couple others in mix too, 538 was easiest to compare against at this moment for me).

If you add a simple average, and I hope you still do, make it as broad as you want. Including all tests the theory that a simple average of all is powerful.

[sndesai1]

added average, not including xmas day games:


for the average entry, i only did not include the two "benchmarks" (vegas and pyth + .500)

[Crow]

Thank you. So my tweaks and weightings improved performance about as 3/4ths as much as the broad average improved on Vegas. An important boost. Median prediction was a bit below Vegas, so the total value of blending was a bit greater.

[Crow]

How well is this guy's model doing?

http://nathanbrixius.wordpress.com/2014 ... ba-season/

[Crow]

Predictions from a guy stating apbrmetric registration issues. (EDIT: see the comment at end of Rob's article)
http://pointsperpossession.com/nba/34-n ... umentation

He is registered but don't know if he knows that, can use that. No posts yet.

[Bobbofitos]

Crow, I'm PPPbasketball. Those are just my predictions