Predictions 2013-14

[Mike G]

While such things are on my desktop, a team that has over-performed relative to what I expected.

Code: Select all

eWins        predicted         actual/projected
.Wiz      Min   e484   eW      Min   e484   eW
Wall       30   1.69   8.6      37   1.71  10.7
Nene'      24   1.39   5.6      25   1.30   5.5
Gortat     28   1.07   5.1      33   1.27   7.1
Webster    26    .87   3.8      28    .58   2.7
Beal       24    .90   3.7      29   1.08   5.2

Ariza      22    .87   3.3      33   1.15   6.4
Seraphin   20    .61   2.1       9    .84   1.3
Booker     12    .87   1.8      17    .93   2.6
Harrington  6    .74    .8       3    .40    .2
Maynor     12    .37    .7       4    .25    .2

Singleton  12    .33    .7       3    .63    .4
Temple     12    .27    .5      10    .07    .1
Vesely     10    .26    .4       8    .41    .6
Childress   4    .01    .0       0      

. totals  242         37.0     238         43.0
                     
. new guys    Miller,Andre       1    .86    .2
.                Rice,Glen       2    .25    .1
.              Porter,Otto       4   -.17   -.1
.                              244         43.1

I measure Wall to be just as good as last year -- e484 is assumed to remain the same -- but he's played a lot more minutes.
Net result is +2.1 eWins by year's end.
Nene' is cool; Gortat is +2.0 by playing more and better; Webster is down -1.1; Beal is +1.5, mostly with more minutes.
And Ariza is thru the roof. From decent 6th man to solid starter. Another +3.1 from expected. His eFG% is up .075, and other numbers are generally the best of his career, suddenly, in his 10th season.

So I had picked this team to win 32 -- a couple games worse than last year and lowest of all our predictions -- and they're headed for 43 or so.
(Bobbo guessed 41, and in a stronger East, that would be right on.)

[Bobbofitos]

Did you expect Jack to fall off? His shooting is off by .060. PER and WS/48 say he was a good player last year (15.9/.115) and terrible this year (10.9/.038). From mostly the best numbers of his career to the worst.

No, I have very similar numbers for Jack as you.

Meanwhile, if you thought Earl Clark was terrible, why would you expect him to get as many minutes as Thompson?
Thompson improved immensely from yr 1 to yr 2, and this year, he's stalled out.

I didn't expect the Cavs to play their best players the most minutes, I just expected Clark to get a lot of minutes based on the contract he signed

[Mike G]

Current avg errors vs b-r.com's forecast.

Code: Select all

Bobb    5.95        Yoop    7.17
hDon    6.44        fpli    7.76
jBro    6.47        deep    8.03
ATC     6.51        416x    8.37
ncs     6.65         eW     8.40
jank    7.16        13Py    9.22

See team by team player improvements and slides here: viewtopic.php?f=2&t=8393&p=18928#p18928

[Mike G]

In just a few days, improvement in the group error, now at just 7.01

Code: Select all

Bobb   5.81      Yoop   6.97
ATC    6.24      fpli   7.62
hDon   6.28      deep   7.88
jBro   6.43       eW    8.12
ncs    6.54      416x   8.25
jank   6.94      13Py   9.04

While most of us improved by .20 or so, the 2013 Pyth entry also did so (by .22)
Relative to that standard, the leaders are 3-4 times better than us stragglers:

Code: Select all

Bobb   3.24      Yoop   2.08
ATC    2.80      fpli   1.42
hDon   2.76      deep   1.16
jBro   2.62       eW    0.92
ncs    2.50      416x   0.79
jank   2.10      13Py   0.00

After the Wolves, who should be 40-23 (by pt-diff; but they're 32-31), I notice the Kings are also under-performing.
They should be 27-38, but they're 4 wins shy of that. With SRS of -1.78, in the East they might be vying with Charlotte (-1.80) for the 8th playoff spot.

[AcrossTheCourt]

It looks like I moved up into second place.

So I wanted to compare our results to other sets of predictions. Using b-ref's projection, SCHOENE (ESPN) is at 8.85 RMSE (6.77 MAE), Teamrankings is at 8.50 (6.67), Bovada (Vegas) is at 8.83 (6.71), and poor Boxscoregeeks is at 10.2 (8.3). So compared to the professionals/major sites some of us are doing really well.

Also, are these lines the best to use for Bovada? I wasn't sure:
http://nba.si.com/2013/10/18/nba-over-under-win-totals/

[DSMok1]

It looks like I moved up into second place.

So I wanted to compare our results to other sets of predictions. Using b-ref's projection, SCHOENE (ESPN) is at 8.85 RMSE (6.77 MAE), Teamrankings is at 8.50 (6.67), Bovada (Vegas) is at 8.83 (6.71), and poor Boxscoregeeks is at 10.2 (8.3). So compared to the professionals/major sites some of us are doing really well.

Also, are these lines the best to use for Bovada? I wasn't sure:
http://nba.si.com/2013/10/18/nba-over-under-win-totals/

You'd expect a set of predictions to have some scatter--some will randomly beat the true "optimum" projection just out of random luck.

[Mike G]

... just out of random luck.

You can't say "random" without saying "dumb".

[AcrossTheCourt]

It looks like I moved up into second place.

So I wanted to compare our results to other sets of predictions. Using b-ref's projection, SCHOENE (ESPN) is at 8.85 RMSE (6.77 MAE), Teamrankings is at 8.50 (6.67), Bovada (Vegas) is at 8.83 (6.71), and poor Boxscoregeeks is at 10.2 (8.3). So compared to the professionals/major sites some of us are doing really well.

Also, are these lines the best to use for Bovada? I wasn't sure:
http://nba.si.com/2013/10/18/nba-over-under-win-totals/

You'd expect a set of predictions to have some scatter--some will randomly beat the true "optimum" projection just out of random luck.

That's a bit dismissive of our hard work here, but it's worth noting a group of five of us are beating those major predictions. It's not like we're a group of 1000 predictions. It's a small group here.

Beating Bovada is no small feat, though maybe I didn't use the fully updated win totals (I'd be happy to be corrected here, as I'm not too familiar with Vegas.) An 82 game (well, getting closer to 82 now) sample with 30 teams is sufficient enough to test predictions.

[Mike G]

A week later and I believe the order is unchanged.

Code: Select all

avg diff. from b-r.com forecast           relative to 2013 Pythagorean
Bobb   5.76      Yoop    6.93            Bobb   3.26      Yoop    2.09
ATC    6.25      fpli    7.68            ATC    2.76      fpli    1.33
hDon   6.30      deep    7.85            hDon   2.72      deep    1.17
jBro   6.41        eW    8.02            jBro   2.60        eW    1.00
ncs    6.59      416x    8.13            ncs    2.43      416x    0.89
jank   6.88      13Py    9.02            jank   2.14      13Py    0.00

Our avg avg error is now 6.98

Also, are these lines the best to use for Bovada? I wasn't sure:
http://nba.si.com/2013/10/18/nba-over-under-win-totals/

I entered these and they add up to 1234 wins -- 4 too many.
Factoring their entries by (1230/1234), predictions are reduced by 0.1 to 0.2 wins per team; and their avg error is 6.53, which is beat by 4 of our members.

[Mike G]

At roughly 89% of the season:

Code: Select all

Bobb  5.82     jank   6.81         Bobb  3.40     jank   2.41
hDon  6.28     fpli   7.35         hDon  2.94     fpli   1.87
ATC   6.45     deep   7.91         ATC   2.77     deep   1.30
jBro  6.52      eW    8.09         jBro  2.70      eW    1.13
ncs   6.67     416x   8.29         ncs   2.54     416x   0.92
Yoop  6.72     13Py   9.22         Yoop  2.50     13Py   0.00

Avg error = 6.99, Bovada at 6.55

If we 'weight' the avg error (columns at left) for each of our participants by the values on the right, we have a composite avg error of 6.73
Of course, that's not legit for this year's prediction; but we could do it next year (using this year's weights).

[Mike G]

With just a few games left -

Code: Select all

relative to b-r.com forecast        ...  relative to 2013 Pyth.
Bobb   6.04      jank   6.98         Bobb   3.40      jank   2.46
ATC    6.51      fpli   7.54          ATC   2.93      fpli   1.90
hDon   6.52      deep   8.06         hDon   2.92      deep   1.38
jBro   6.72       eW    8.33         jBro   2.72       eW    1.11
Yoop   6.84      416x   8.51         Yoop   2.60      416x   0.93
ncs    6.90      13Py   9.44          ncs   2.54      13Py   0.00

At one point, I was almost out of the 2nd division. Had I made it over that hump, anything might have happened.
Community avg error lurched up to 7.18. Bovada now off by 6.76

[nbacouchside]

Bob,

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

[RoyceWebb]

With just a few games left -

Code: Select all

relative to b-r.com forecast        ...  relative to 2013 Pyth.
Bobb   6.04      jank   6.98         Bobb   3.40      jank   2.46
ATC    6.51      fpli   7.54          ATC   2.93      fpli   1.90
hDon   6.52      deep   8.06         hDon   2.92      deep   1.38
jBro   6.72       eW    8.33         jBro   2.72       eW    1.11
Yoop   6.84      416x   8.51         Yoop   2.60      416x   0.93
ncs    6.90      13Py   9.44          ncs   2.54      13Py   0.00

At one point, I was almost out of the 2nd division. Had I made it over that hump, anything might have happened.
Community avg error lurched up to 7.18. Brovada now off by 6.76

I like to see how our ESPN Forecast "wisdom of crowds" project is matching up with the more sophisticated statistical systems, and here's the latest (from Sunday):

AVG ERROR
Fall Forecast (October): 6.47
Summer Forecast (Aug.): 6.77

Strange season in which we might be within 0 or 1 win(s) on 10 teams and 10+ games off on eight teams.

Royce

[J.E.]

Here are playoff predictions using W Pyth with an exponent of 11, Sagarin's ratings (with +8 added to each rating because his, for whatever reason, don't match up with league average) and a homecourt advantage of 3.5

Listing most likely outcome

Spurs vs. Mavs: Spurs in 5 (27%). Chance of Dallas winning the series: 20%
Rockets vs. Blazers: Rockets in 7 (19.5%). Chance of Blazers winning the series: 40%
Thunder vs. Grizzlies: Thunder in 5 (23%). Chance of Memphis winning the series: 26%
Clippers vs Warriors: Clippers in 7 (20%). Chance of Warriors winning the series: 35%

Pacers vs. Hawks: Pacers in 5(24%). Chance of Atlanta winning the series: 27%
Bulls vs. Wizards: Bulls in 7(20%). Chance of Washington winning the series: 40%
Heat vs. Bobcats: Heat in 5(26%). Chance of Charlotte winning the series: 22%
Raptors vs. Nets: Raptors in 5(24%). Chance of Nets winning the series: 27%

These all seem to favor the underdog a little too much, although I seem to remember that an exponent in the 10-12 range was best to predict future games. Changing the exponent so my series-win% match better with vegas (in this case, german online sports book bwin.com) we get

Spurs vs. Mavs: Spurs in 5 (36%). Chance of Dallas winning the series: 8%
Rockets vs. Blazers: Rockets in 7 (22%). Chance of Blazers winning the series: 34%
Thunder vs. Grizzlies: Thunder in 5 (32%). Chance of Memphis winning the series: 15%
Clippers vs Warriors: Clippers in 5 (25%). Chance of Warriors winning the series: 25%

Pacers vs. Hawks: Pacers in 5(32%). Chance of Atlanta winning the series: 15%
Bulls vs. Wizards: Bulls in 7(23%). Chance of Washington winning the series: 33%
Heat vs. Bobcats: Heat in 5(35%). Chance of Charlotte winning the series: 10%
Raptors vs. Nets: Raptors in 5(31%). Chance of Nets winning the series: 16%

Looks pretty good to me. Only series I heavily disagree with is Raptors - I'm guessing the Nets have a better than 16% chance of winning it.
Rockets v Blazers, Bulls v Wizards and Warriors v Clippers seem like the series to watch

[Mike G]

... Sagarin's ratings (with +8 added to each rating because his, for whatever reason, don't match up with league average)...

An average team is 90 on his scale.

When you change your exponent, the only series (most likely outcome) that changes is LAC-GSW -- from Clipps in 7, to Clipps in 5.

According to b-r.com (and Sagarin) the 6 best teams (SRS) are in the West.
If the Pacers make the Finals (.321), they have a .283 chance of toppling the Western representative.
If it's the Heat, they are given (112/351) a .319 likelihood.

The Spurs are given a 34% chance of getting out of the West; if they do, 83% likely they win it all.
A 1/3 chance they face neither the Heat or Pacers.
http://www.basketball-reference.com/fri ... f_prob.cgi

... Raptors - I'm guessing the Nets have a better than 16% chance of winning it.

They went 34-17 since Jan. 1; or 34-15, then tanked their last game or two. Because they'd rather face Tor than Chi?