Predictions 2013-14

[J.E.]

And here's Vegas atm (04/18/14)

BRK: 1.6 TOR: 2.35
GSW: 3.9 LAC: 1.27
ATL: 5 IND: 1.18
MEM: 4.2 OKC: 1.24
DAL: 6 SAS: 1.14
CHA: 13 MIA: 1.03
WAS: 2.6 CHI: 1.5
POR: 2.7 HOU: 1.5

So, they actually favor Brooklyn. They give Dallas a 15% chance, my method says 8. Most of the others are similar

Comedy option of betting on 4:0 Bobcats gets you 150 to 1

[Bobbofitos]

Bob,

Do you mind sharing what metric you used to so thoroughly dominate? TYIA.

Personal/private metric, sorry

[Mike G]

And here's Vegas atm (04/18/14)

BRK: 1.6 TOR: 2.35
GSW: 3.9 LAC: 1.27
ATL: 5 IND: 1.18
MEM: 4.2 OKC: 1.24
DAL: 6 SAS: 1.14
CHA: 13 MIA: 1.03
WAS: 2.6 CHI: 1.5
POR: 2.7 HOU: 1.5

Can you tell us what these numbers mean?

[J.E.]

It says you get $5 when betting $1 on Atlanta and Atlanta does win the series (you win 4). For Miami you'd only get $1.03 (you win 3c)

You can sort of derive the expected Win% of each team. In 'ATL v. IND's case ATL's chances are 1/5 = 20% and IND's chances are 1/1.14 = ~85%.
20+85 > 100 - that's how Vegas makes money

[Mike G]

Final numbers

Code: Select all

avg error from win totals        ...  relative to 2013 Pyth.
Bobb   6.00      jank   7.13         Bobb   3.34      jank   2.20
hDon   6.47      fpli   7.66         hDon   2.87      fpli   1.67
ATC    6.54      deep   8.22         ATC    2.80      deep   1.11
jBro   6.73       eW    8.26         jBro   2.60       eW    1.07
ncs    6.87      416x   8.70         ncs    2.47      416x   0.64
Yoop   6.87      13Py   9.34         Yoop   2.47      13Py   0.00

Avg avg error was 7.22
Bovada 6.77

I have been giving our "avg avg error" as the avg of the 11 participants' avg errors.
But if we average our guesses (for each team) and use these 30 numbers as a single set of 'predictions' -- then APBRMetrics had an avg error of 6.51
That would seem to be closer to the way ESPN got their collective guess -- 6.47, according to Royce, previous page.
Sorry about the confusion.

[RoyceWebb]

ESPN Forecast (wisdom of crowds panel):

AVERAGE ERROR
6.67 (Summer Forecast; August)
6.67 (Fall Forecast; October)

RMSE
9.10 (Summer Forecast)
9.21 (Fall Forecast)

Lots of very accurate predictions (getting team W-L right or very close), and lots of very inaccurate predictions; strange season that way.

Royce

[Mike G]

Same avg error from Summer and Fall?

I'm not a fan of RMSE in this case. Actually, we have a different winner in that analysis: Bobbo missed Phx by 27; hoopDon missed nobody by more than 19. Squared, that's less than half as bad.

Code: Select all

hDon   8.14      jank    9.17
Bobb   8.28      fpli    9.89
Yoop   8.70      deep   10.14
ncs    8.85      416x   10.53
ATC    8.86       eW    10.72
jBro   8.93      13Py   11.90

Since it sends me to last place, I'm doubly uninterested.

The avg is 9.29, and as a group we're 8.52

[J.E.]

Vegas now has the chances of Brooklyn winning their series at 72%

Pacers are still seen as favorites, but not by a lot. Indiana in 7 as the most likely

Clippers also still the favorite, but at 52%/48%

[wilq]

Final numbers

I added some more entries to the contest... but this year it didn't matter because they couldn't beat you ;-)
Great season for APBR members, congratulations!

Link: http://weaksideawareness.wordpress.com/ ... ojections/

[AcrossTheCourt]

I'm surprised I did as well as that, given my super simple method and how I whiffed on the Bucks and Suns.

I made a couple of dumb subjective tweaks. One was to penalize the Suns a little more (I think I dropped off like three wins from them.) Another was to give Deron Williams the benefit of the doubt and lift Brooklyn even further. Lesson: do not trust Deron Williams or aging scoring point guards in general.

There are a couple areas of improvement I could work on. The first is finding a good, usable value for coaches. That's useful whenever a Doc Rivers is brought in to replace a bad coach. The next is minutes projection/injury prediction. That's the most underrated part of these predictions. Also, I whiffed on Charlotte, and I should have seen that coaching/defensive change coming. That was my fault. I was right in thinking Jefferson's offense wouldn't help them too much.

Looking at my results, I don't think it was a fluke at all. I mean, one of my worst errors was Portland, but I was higher on them than most. I'm also building/testing my own metrics, and I have some margins to mine there. I should do a lot better. That was a test run.

[Crow]

Congratulations to Bobbofitos and others at or near the top by one method or another.

[Mike G]

Somehow I nailed Dallas at 48 (pyth) wins.
This analysis isn't entirely about gloating. A concurrent thread is about how good a coach Carlisle may be, derived from how the players may have improved (or not) under his direction.

Here's my prediction from last Oct., using 2013 eW rates, guessing at this year's minutes; and how I measured them this year:

Code: Select all

2014 Dallas       - expected - - -        - - actual - - -
eW+    Mavs      e484    mpg    eW       e484    Min    eW
4.0   Nowitzki   1.52    26    6.7       1.98    32   10.7
2.0   Dalembert  1.27     8    1.7       1.12    20    3.7
1.8   Ellis      1.16    30    5.9       1.23    37    7.7
0.4   Blair      1.05    14    2.5       1.17    15    2.9
0.2   Larkin        0     0      0        .21     6     .2

0.1   Mekel         0     0      0        .21     4     .1
.05   Ledo          0     0      0        .68     .4   .05
-.2   Crowder     .42    18    1.3        .42    15    1.1
-.3   Wright     1.27    16    3.4       1.41    13    3.2
-.5   Carter     1.23    24    5.0       1.10    24    4.5

-1.1  Calderon   1.37    26    6.0        .96    30    4.9
-1.2  Ellington   .68    14    1.6        .47     5     .4
-1.3  James       .80    10    1.3        .23     2     .1
-2.0  Harris      .89    24    3.6        .96    10    1.6
-2.3  Marion     1.28    28    6.1        .76    29    3.8
                                
-0.1   totals           242   45.0              242   45.0

Marion fell off by an unexpected amount, still got his minutes.
I was way low on Dalembert's minutes, and he had some of his best numbers at age 32.
But Dirk really stepped up, making up for a few disappointments.