2016-17 Team Win Projection Contest / Discussion

Home for all your discussion of basketball statistical analysis.
Post Reply
Mike G
Posts: 4429
Joined: Fri Apr 15, 2011 12:02 am
Location: Asheville, NC

Re: 2016-17 Team Win Projection Contest / Discussion

Post by Mike G » Sun Nov 27, 2016 8:17 pm

Table summarizing our guesses. The first column after the team name is the b-r.com projection today.
Ranked by avg error, top 10 and bottom 10.

Code: Select all

err    0   4.2  4.2  4.3  4.3  4.4  4.4  4.5  4.6  4.6  4.6
tm   proj   MG   RR  amp  snd  lqi  sbs  tao  jgo  tzu  kmd
Atl   46    45   41   41   42   43   41   43   43   44   44
Bos   43    46   50   50   50   51   51   50   51   55   51
Brk   27    24   26   25   26   23   24   28   26   21   24
Cha   45    42   43   45   40   42   42   42   44   41   42
Chi   49    43   39   39   38   38   40   39   39   34   42

Cle   55    54   57   56   54   56   57   61   55   57   56
Dal   26    38   38   40   38   37   38   37   38   36   36
Den   39    38   36   37   35   37   37   35   35   36   36
Det   43    43   43   41   42   43   43   42   40   44   40
GSW   60    66   69   69   67   65   69   65   67   68   71

Hou   48    42   46   47   48   44   45   45   44   49   50
Ind   34    44   40   40   40   42   41   37   41   40   41
LAC   60    51   50   51   52   51   51   50   52   54   51
LAL   38    30   23   23   24   23   24   22   28   24   22
Mem   43    39   39   38   38   42   39   39   37   43   38

tm   proj   MG   RR  amp  snd  lqi  sbs  tao  jgo  tzu  kmd
Mia   37    31   35   35   37   33   35   34   37   35   40
Mil   37    34   34   34   34   33   33   33   33   35   31
Min   36    43   37   37   44   37   40   38   43   38   44
NOP   35    34   36   36   38   39   35   37   40   32   33
NYK   39    34   37   37   38   38   35   36   32   37   35

Okl   43    48   46   45   44   45   45   43   44   44   47
Orl   29    34   36   36   35   37   37   38   38   35   34
Phl   25    25   25   28   28   26   24   26   26   23   24
Phx   29    30   28   25   27   29   29   33   28   27   28
Por   38    43   45   45   45   44   45   44   44   46   45

Sac   36    34   36   34   35   35   35   37   36   35   34
SAS   56    60   57   56   54   58   56   53   57   54   57
Tor   51    50   51   52   51   51   50   50   47   48   52
Uta   47    45   49   48   45   50   48   50   46   49   47
Was   35    40   41   40   41   41   41   43   40   44   39
  
                               
err    0   4.7  4.7  4.7  4.7  4.8  4.8  4.9  5.0  5.1  5.1
tm   proj  nc2  Crow  BD  nc1  sha  ATC  yoo  cal  GK5  nre
Atl   46    43   46   45   40   41   40   44   45   38   41
Bos   43    51   52   50   51   49   54   53   53   48   48
Brk   27    28   23   20   27   28   26   28   26   28   26
Cha   45    37   39   44   38   41   43   39   42   41   43
Chi   49    39   42   41   41   39   39   33   38   43   42

Cle   55    56   57   57   60   59   58   56   56   52   61
Dal   26    37   37   40   36   41   38   35   37   38   42
Den   39    36   36   35   36   36   35   41   36   39   31
Det   43    40   39   40   43   41   41   40   43   40   37
GSW   60    71   70   67   70   66   71   69   72   73   70

Hou   48    48   47   51   47   49   48   50   47   45   44
Ind   34    41   39   43   38   39   37   40   43   39   39
LAC   60    53   55   53   50   50   50   48   52   49   52
LAL   38    22   23   21   21   24   23   27   22   25   23
Mem   43    37   44   43   37   42   40   38   37   37   43

tm   proj  nc2  Crow  BD  nc1  sha  ATC  yoo  cal  GK5  nre
Mia   37    37   35   38   36   38   35   46   41   36   31
Mil   37    34   29   28   33   29   30   31   32   30   30
Min   36    39   37   38   35   33   39   43   39   40   32
NOP   35    38   32   32   35   37   36   28   32   35   39
NYK   39    35   37   35   39   36   34   34   36   31   38

Okl   43    42   47   48   44   46   44   44   45   48   52
Orl   29    38   31   33   38   37   37   30   34   37   32
Phl   25    24   19   19   21   21   25   25   21   29   24
Phx   29    25   37   29   27   28   27   29   26   31   29
Por   38    43   42   44   46   44   45   40   43   45   42

Sac   36    38   34   36   38   40   37   40   33   38   39
SAS   56    59   56   59   58   57   57   58   59   53   60
Tor   51    51   48   52   52   48   52   51   51   52   50
Uta   47    48   48   47   53   49   50   49   48   51   47
Was   35    40   46   42   40   42   42   41   43   39   43
This is thru 20% of the season.

Nathan
Posts: 137
Joined: Sat Jun 22, 2013 4:30 pm

Re: 2016-17 Team Win Projection Contest / Discussion

Post by Nathan » Sun Nov 27, 2016 11:38 pm

Mike G wrote:It's a tough crowd. If you're better than Vegas, but not by much, you may be middle of the pack.

Those oddschecker numbers sum to 1252 wins, 22 more than is possible. Teams avg 41.73 wins.
Dropping the .5 from each one gives an avg error of 4.53
Yeah, I noticed that and thought it was interesting. Shows the effect of homers betting on their own teams, probably.

Mike G
Posts: 4429
Joined: Fri Apr 15, 2011 12:02 am
Location: Asheville, NC

Re: 2016-17 Team Win Projection Contest / Discussion

Post by Mike G » Mon Nov 28, 2016 2:53 pm

Why are the Rockets doing so well? They're on pace for 49 wins, which is just 2.7 better than our avg prediction. But I had them pegged at 42; Vegas said 43; everyone else here was 44 to 51.

Here are Kevin P's suggested minutes, their actual mpg (both in terms of min./team games); this year's rates and last year's.

Code: Select all

Rockets '16-17   KP  mpg   PER   WS/48   BPM  1516 PER  WS/48   BPM
James Harden     34   37   29.6   .290  10.7      25.3   .204   6.7
Trevor Ariza     32   34   15.5   .144   3.6      12.9   .090   1.5
Ryan Anderson    23   31   11.7   .075  -2.4  NOP 17.2   .092  -1.4
Eric Gordon      25   31   13.3   .078  -1.4  NOP 13.5   .069   -.6

Clint Capela     23   26   21.2   .173    .7      18.3   .144    .5
Sam Dekker       10   17   13.4   .112    .0      10.8   .116    .5
Nene Hilario     17   16   14.8   .109  -1.4  Was 16.8   .120    .2
Corey Brewer     19   15    9.9   .072  -2.4       9.9   .035  -3.1

Patrick Beverley 25   10   12.4   .118   1.6      13.0   .102   1.1
K.J. McDaniels   10    9   11.7   .085  -2.3      10.6   .043  -5.0
Tyler Ennis *    14    7    7.4  -.012  -6.3  Mil 10.7   .030  -4.8
Montrezl Harrell  4    5   16.8   .056  -1.9      14.6   .118  -1.9

. total         236  238   16.4   .132   1.26     15.8   .108   .58
* Kevin had Pablo Prigioni in Ennis' role. Dekker's 2016 rates are for 6 minutes of play.

With very few missed games, their top players are getting a lot more minutes than expected. Even those who are not great (or down from last year) are better than the guys who were supposed to be getting the minutes.

J.E.
Posts: 818
Joined: Fri Apr 15, 2011 8:28 am

Re: 2016-17 Team Win Projection Contest / Discussion

Post by J.E. » Tue Nov 29, 2016 8:12 am

Mike G wrote:Why are the Rockets doing so well? They're on pace for 49 wins, which is just 2.7 better than our avg prediction. But I had them pegged at 42; Vegas said 43; everyone else here was 44 to 51.
RAPM/RPM type metrics probably liked them because they ditched Terrence Jones

Mike G
Posts: 4429
Joined: Fri Apr 15, 2011 12:02 am
Location: Asheville, NC

Re: 2016-17 Team Win Projection Contest / Discussion

Post by Mike G » Tue Nov 29, 2016 1:42 pm

Errors continue to plummet.
I've added the Vegas lines submitted by Nathan (previous page).
Also, the 16Py = (2016 pythagorean wins +41)/2 -- i.e., regressed halfway to .500 W%
A 2nd column of numbers is the difference between the participant and that baseline.

Code: Select all

avg error   Py-err    avg error   Py-err
RyRi   3.99   .98      ATCt   4.51   .46
amp3   4.00   .97      vegas  4.51   .45
Mike   4.02   .95      taco   4.54   .43
lnqi   4.22   .74      Crow   4.54   .42
sndi   4.22   .74      shad   4.55   .41
sbs.   4.24   .73      ncsD   4.57   .39
kmed   4.27   .69      yoop   4.67   .29
trzu   4.31   .65      cali   4.68   .29
jg34   4.42   .54      GK5.   4.89   .07
BaDo   4.47   .49    * 16py * 4.96 * .00
ncsB   4.50   .47      nrfo   4.98  -.01
At this time last season, errors were about 60% bigger.
Year before that, they were twice as big as now.
In 2013(14) on this date, almost 2.5 times as large.

shadow
Posts: 220
Joined: Fri May 29, 2015 3:38 am

Re: 2016-17 Team Win Projection Contest / Discussion

Post by shadow » Tue Nov 29, 2016 6:17 pm

Mike G wrote:Why are the Rockets doing so well? They're on pace for 49 wins, which is just 2.7 better than our avg prediction. But I had them pegged at 42; Vegas said 43; everyone else here was 44 to 51.
Their 3PAr is ~7% higher and they are shooting ~3% better on threes compared to last year. Their defensive TO% is down a bit from last year, but their defensive rebounding is vastly improved (30th last year; 10th this season).

It probably also helps they didn't start the season tanking games to get their coach fired like last year.

Mike G
Posts: 4429
Joined: Fri Apr 15, 2011 12:02 am
Location: Asheville, NC

Re: 2016-17 Team Win Projection Contest / Discussion

Post by Mike G » Tue Nov 29, 2016 7:04 pm

Well, maybe my question should have been, How did some here anticipate this improvement from last year?
The players seem to have improved by most measures. Even with Dwight gone and Beverley out, their D is about the same; while they're shooting better and rebounding more?

BasketDork
Posts: 201
Joined: Thu Dec 04, 2014 12:58 pm

Re: 2016-17 Team Win Projection Contest / Discussion

Post by BasketDork » Tue Nov 29, 2016 7:12 pm

Do you think the reason for such a tight field and a much lower margin of error is everyone individually improving their methods over time, or moreso just a 'wisdom of crowds' approach. I believe I saw a couple people say they included the group projection within their blend. It's definitely not all luck. But what IS it ?
Last edited by BasketDork on Tue Nov 29, 2016 9:59 pm, edited 1 time in total.
The Bearded Geek

BasketDork
Posts: 201
Joined: Thu Dec 04, 2014 12:58 pm

Re: 2016-17 Team Win Projection Contest / Discussion

Post by BasketDork » Tue Nov 29, 2016 7:27 pm

2015-2016 on left, 2016-2017 on right
(+) denotes being too high on a team's win total
(--) denotes being too low on a team's win total

With a couple exceptions, it would seem there's just teams I seem to closer on than others.

Miami................0 +1
Minnesota.........0 +2
Atlanta..............+1 -1
Cleveland......... -1 +2
Dallas............... -1 +14
Oklahoma City..+2 +5
Utah...................+2 0
L.A. Clippers.....+3 -7
Memphis...........+3 0
Indiana............... -3 +9
Denver................+4 -4
Washington.......+4 +7
Chicago..............+5 -8
Philadelphia.......+5 -6
Sacramento.......+5 0
Orlando............... -5 +4
Brooklyn.............+8 -7
Phoniex...............+9 0
Portland...............-9 +6
L.A. Lakers.........+10 -17
New York............-10 -4
Boston................ -12 +7
Golden State....... -12 +7
Toronto................-12 +1
Charlotte..............-13 -1
San Antonio.........-13 +3
New Orleans........+14 -3
Detroit...................-14 -3
Houston...............+15 +3
Milwaukee............+15 -9
The Bearded Geek

Mike G
Posts: 4429
Joined: Fri Apr 15, 2011 12:02 am
Location: Asheville, NC

Re: 2016-17 Team Win Projection Contest / Discussion

Post by Mike G » Tue Nov 29, 2016 8:38 pm

BasketDork wrote:Do you think the reason for such a tight field and a much lower margin of error is everyone individually improving their methods over time, or moreso just a 'wisdom of crowds' approach. I believe I saw a couple people say they included the group projection within their blend. It's definitely not luck. But what IS it ?
It's 80 to 90% luck. Along with our 'improving' predictions is a narrowing of the avg team's difference from the previous season. Teams that change more are less predictable.

This isn't always due to players' coming and going. Sometimes there's good or bad chemistry. I haven't noticed anyone here predicting those things very well.

The 4-season 'trend' is therefore just 4 random years that line up from worst to best, re predictability. Thus, luck.
If we are getting better, it might be found in how much better we do than (Py0+41)/2

BasketDork
Posts: 201
Joined: Thu Dec 04, 2014 12:58 pm

Re: 2016-17 Team Win Projection Contest / Discussion

Post by BasketDork » Tue Nov 29, 2016 9:59 pm

Whoops. I left out the word 'all' before luck, in the phrase, definitely not luck. Just reread that.
The Bearded Geek

Crow
Posts: 6248
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2016-17 Team Win Projection Contest / Discussion

Post by Crow » Wed Nov 30, 2016 3:16 am

Props to the leaders.

Still a long way from finish.

Right now the rankings depend heavily on how similar a contest projection is to the measuring stick model plus 20-25% actuals. We don't yet know how good those measuring stick models are so closeness to them is of unclear significance. At halfway point and beyond it will be more likely to be meaningful as the actuals become the majority and the model for the rest of the season presumably also gets better.

The schedules to date will get more similar to the season schedules and each other as well. Right now we didn't know how much of the variance from season projections is coming from schedule related variance.

How good was the BR projection model at the quarter pole last season compare to the full season actuals? Anyone want to check way back machine or other record?

Injury related variance could also be higher now in general but unevenly felt in the contest projections and the measuring stick models as well. Trades will have some affect too. Youth, experience consistency and resiliency of players and coaches is unevenly distributed and degree of improvement from these factors are likely to be as well to some degree.

Playoff drives, coasting, youth development, coaching changes, locker room issues and use of cap & exceptions can also alter performances some. Owners speak up and recede. If offensive efficiency trends up as season progresses as in past that may affect teams differently. If factor level trends deepen or weaken that will too. Is officiating even between top, muddle and bottom teams throughout the season? It might not be. Ejections and suspensions pop up. Travel issues could impact a few outcomes.

The quality of lineup management is likely to pull top and bottom teams further apart.

Mike G
Posts: 4429
Joined: Fri Apr 15, 2011 12:02 am
Location: Asheville, NC

Re: 2016-17 Team Win Projection Contest / Discussion

Post by Mike G » Wed Nov 30, 2016 10:59 am

Don't know how to recover the daily b-r.com projections, but they're pretty heavily regressed at this point in the season. This 'favors' any prediction that is heavily regressed.
At the end of the season, we get tanking, overt and covert. That shakes things down.

Can we standardize some definitions? To me, a 'prediction' is what we guess before any games have been played. A 'projection' is an extrapolation based on games already played.

Granted, we 'project' player performance based on games already played in previous seasons. And from these we predict how teams will do this season. They're somewhat interchangeable terms.
It's sort of like 'ratings' and 'rankings' getting interchanged. But one is a quantification, and the other is an ordering of the (alleged) quantity.

GK5
Posts: 11
Joined: Thu Aug 25, 2016 12:53 am

Re: 2016-17 Team Win Projection Contest / Discussion

Post by GK5 » Wed Nov 30, 2016 1:33 pm

Thank you for taking the time to keep score!

AcrossTheCourt
Posts: 237
Joined: Sat Feb 16, 2013 11:56 am

Re: 2016-17 Team Win Projection Contest / Discussion

Post by AcrossTheCourt » Wed Nov 30, 2016 3:21 pm

Analytics-based projections liked Houston because they saw that Howard was overrated, Nene is always loved by advanced stats, their defensive issues were overblown, and people put too much emphasis on their 2016 season and not 2015.

Post Reply