2011 Win Predictions

[J.E.]

If Crow already copied the old thread I can't find it, search tells me I cannot use the words "wins" or "predictions" as a search term.

Code: Select all

b2n	6.8
Vegas	5.5
JH	5.47
KP	7.67
KD	6.6
Dsmok1	6.29
Crow	6.27
schtevie	7.53
WoW	7.41
WS	6.78
SPM(bbr)	7.07
SRS	8.38
“41”	10.8
(lastyear+41)/2	8.4

Congratulations to John Hollinger for winning the contest and beating Vegas.
Congratulations to Crow for best predictions in this thread, although I feel as long as he doesn't post which metrics he blended this won't help too much. Then again, he said he blended ~12 metrics? That probably means that pretty much every metric is in there
Congratulations to Dsmok1 for lowest error with a single metric

Looking for what "went wrong" with my predictions, compared to Vegas:
-Overestimated Cleveland. Some of the bad play can probably be tracked to the injury of Varejao, and Scott not being as good of a coach as MB was, according to RAPM with coaching, but even then I had them way too high
-Overestimated Minnesota.
-Expected more games missed or general regression due to injury for Chris Paul
-Severely underestimated Philadelphia and overestimated Sacramento. Overall a bad assessment of the Dalembert/Nocioni trade. Also, coach RAPM says Doug Collins is +2 above average
-Phoenix, overestimated Hakim Warrick and their depth in general

p.s. I get logged out everytime I close the page. That shouldn't happen, should it?

[Jeff Fogle]

I happened to log some some of the predictions by hand heading into the last day of the season so I could do some comparisons in won/lost record vs. Vegas. Anything by hand might have a miscue or two. Using differentials of two points or more between prediction and what was listed as "Vegas" in the original thread. I show:

Pelton: 15-11
Hollinger: 11-9
KD: 12-12
Crow 10-11
DSMok: 11-13

Wasn't clear on what the specific "selections" might have been for the other groups/people listed. KP was better with direction than scope. For example, Golden State didn't go Over by 17 wins like he had guessed, but they did go Over. That shows up as a win one way, but a bad miss if you're just looking at margins. Atlanta Under 46 is a winner though he had undershot by a bit. So, congrats to KP for having a good read on directions...and to JH for besting Vegas both ways. Nothing to be ashamed of from the group. Football Outsiders was 6-16 two years ago with their NFL projections, so it's possible for a stat-based approach to have an off-year. They did get much of that back in 2010.

Hope Greyberger can re-post that link that he had in the other thread with the write-ups. Recall that guy doing very well, but I didn't log down what he had said.

Beating expectations: Chicago, San Antonio, LA Lakers, Boston, Denver, Oklahoma City, New Orleans, Dallas, Golden State, Philadelphia, New York, Indiana, Memphis

Missing expectations: Miami, Portland, Orlando, Atlanta, Utah, Phoenix, Milwaukee, Houston, Charlotte, Cleveland, New Jersey, Detroit, Sacramento, Washington, LA Clippers, Toronto, Minnesota

[Mike G]

I get logged out, too, but at least it remembers my password.

Assuming these are errors from actual wins. Actual wins 'erred' by about 2.6 from Pythagorean.
Could you report the differences between predictions and Pythagorean?

[Jeff Fogle]

Based on the market prices listed in original thread:

Biggest Surprise: Chicago (+16)
Next Biggest: San Antonio (+11)

Worst Surprises: Milwaukee, Washington, Cleveland (-10)
Next Worst: Utah (-9)

Other Positive Surprises:
Memphis (+8), New York, Dallas, Denver (+7), New Orleans, Philadelphia (+6)

Other Negative Suprises:
Miami, Minnesota (-6)

[J.E.]

Could you report the differences between predictions and Pythagorean?

Code: Select all

b2n	6.8
Vegas	5.23
JH	4.93
KP	7.47
KD	6.8
Dsmok1	6.62
Crow	6.27
schtevie	7.33
WoW	7.6
WS	6.59
SPM(bbr)	6.91
SRS	8.44
“41”	10.8
(lastyear+41)/2	8.6

[Crow]

I did re-assemble this thread (currently on page 3 of 6)
http://sonicscentral.com/apbrmetrics/vi ... p?f=2&t=64

yes the search tool will occasionally balk at searching for "common words" (seems pretty ridiculous, but that is the way it is)

a manual search of the 6 pages is an option if the search tool is balking


I don't recall being asked to name the metrics blended before, but now that I have, I will say they were:

Vegas, Hollinger, back2newbelf, DSMok1, KP, KD, Hickory High (WP based), the ESPN panel average, Win Shares, Neil's SPM, Arturo, Simple Rating and Schtevie. I will probably decline to give the variable weights because I don't want to offend some of those authors. After finding the weighted score, I added by own judgment / adjustment, with some varying amount of regression to the mean.

A revised formula that I tinkered with maybe two months ago did better than this larger meta-metric. Based on several years of tracking other metrics, I might go with a smaller set of best performers in the future. It uses only 5 of these.

[Kevin Pelton]

KP was better with direction than scope. For example, Golden State didn't go Over by 17 wins like he had guessed, but they did go Over. That shows up as a win one way, but a bad miss if you're just looking at margins.

Interesting. My sense was that was the case, but it's nice to have it quantified. Still lots more work to do on SCHOENE ... .

[J.E.]

I did re-assemble this thread (currently on page 3 of 6)
http://sonicscentral.com/apbrmetrics/vi ... p?f=2&t=64

Should we paste the old stuff in here?

After finding the weighted score

How did you find the weights?

[Crow]

I don't think there is any further need to paste stuff either direction. A link was made to the old thread if any want to refer to it.

The metric weights were based on a brief review on metric performance in past years, where that data was available, as well to some degree my evaluation of and comfort with the metrics themselves (the formulas and the stat weights they used and what was missing from the metric).

[greyberger]

The blogger I referred to the first time around, and Jeff Fogle mentions above, was M. Haubs at the painted area. Here are his pre-season predictions. I dunno how he would rank against our Quant Bachelors but his reasoning for the picks make for a good read.

Edit: Haubs nails it. He missed by picking Nets to go over 25.5 (they won 24) and Portland to go over 51.5 (they won 48). The other five picks were correct and he picked San Antonio and Chicago, obviously to great success.

Second edit to quote this: "NYK Over 36 (29): Pelton's numbers love this one, but trading Randolph and Gallinari for Melo could be a "be careful what you wish for" kind of thing." Nailed it!

[Crow]

bump

[Crow]

bump

[Crow]

bump