2016-17 Team Win Projection Contest / Discussion

[ampersand5]

PTPM + RPM + BPM blend

GSW 69
Cle 56
Sas 56
Tor 52
Lac 51
Bos 50
Uta 48
Hou 47
Cha 45
Okc 45
Por 45
Atl 41
Det 41
Dal 40
Ind 40
Was 40
Chi 39
Mem 38
Den 37
Min 37
NYK 37
Nop 36
Orl 36
Mia 35
Mil 34
Sac 34
Phi 28
Brk 25
Phx 25
Lal 23

[Mike G]

avg of eWins and fauxRPM, using Kevin Pelton's player minutes

Code: Select all

Atl   45
Bos   46
Brk   24
Cha   42
Chi   43
Cle   54
Dal   38
Den   38
Det   43
GSW   66
Hou   42
Ind   44
LAC   51
LAL   30
Mem   39
Mia   31
Mil   34
Min   43
NOP   34
NYK   34
Okl   48
Orl   34
Phl   25
Phx   30
Por   43
Sac   34
SAS   60
Tor   50
Uta   45
Was   40

EDITED to relieve rounding errors.

[tarrazu]

Got my submission in before last year's fwiw.
Player based projection inspired by many of the public systems out there, with an emphasis in utilizing SportsVU data.

Code: Select all

team wins		  team wins	
GSW	68			ATL	44
CLE	57			BKN	21
BOS	55			BOS	55
SAS	54			CHA	41
LAC	54			CHI	34
HOU	49			CLE	57
UTA	49			DAL	36
TOR	48			DEN	36
POR	46			DET	44
WAS	44			GSW	68
OKC	44			HOU	49
DET	44			IND	40
ATL	44			LAC	54
MEM	43			LAL	24
CHA	41			MEM	43
IND	40			MIA	35
MIN	38			MIL	35
NYK	37			MIN	38
DAL	36			NOP	32
DEN	36			NYK	37
SAC	35			OKC	44
MIL	35			ORL	35
MIA	35			PHI	23
ORL	35			PHX	27
CHI	34			POR	46
NOP	32			SAC	35
PHX	27			SAS	54
LAL	24			TOR	48
PHI	23			UTA	49
BKN	21			WAS	44

(edit: failed on my initial 2nd column alphabetization)

[lanqiu]

I am a little late, however, these projections only use data that was available prior to the start of the season. It took a little longer than expected to finish the models, if it is too late to enter I understand.

GSW 65.4
SAS 58.2
CLE 56.0
BOS 51.3
LAC 50.8
TOR 50.6
UTA 49.5
OKC 44.5
HOU 44.3
POR 44.1
ATL 43.1
DET 42.8
CHA 42.1
IND 41.7
MEM 41.6
WAS 41.0
NOP 38.7
CHI 37.5
NYK 37.5
MIN 37.2
DAL 36.9
ORL 36.8
DEN 36.6
SAC 35.4
MIL 33.2
MIA 33.1
PHO 28.9
PHI 26.2
BKN 22.6
LAL 22.5

[Crow]

Entered, barring major uproar.

After tomorrow, probably too late.

[BasketDork]

I took a look back at a year ago, where I found myself finishing right in the middle of the pack. 13th of 22 in average error, and 14th out of 22 in RMSE. Out of the 40 predictions presented from apbr members, magazines, Westgate..I finished 26th in avg. error and 25th in RMSE. I figured I'd post my biggest successes and failures from a year ago in an easy spot so that I could see who I "got right" and who I'm still way off on.

FROM 2015-2016
(+) denotes being too high on a team's win total
(--) denotes being too low on a team's win total

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

Underestimated both Spurs and Dubs. Teams like Houston, New Orleans, and Charlotte made everyone look a little dumber. I dont even remember why I was so high on the Bucks, except for some grittiness in the '14-'15 playoffs. Way underestimated some teams, namely Boston, New York, and Toronto. Gave the Lakers way too much credit. Within 3 wins on ten (10) teams. Not too bad. Nailed Miami, Minnesota, Atlanta. Cleveland, and Dallas. Should have reversed Phoniex and Portland, the Lakers & Knicks, and the Kings & Magic. Hopefully 2016-17 treats me even better.

[AcrossTheCourt]

Anyone want to do a summary of the entries and early numbers on the leaders?

[Mike G]

Now that b-r.com has their Playoff Probabilities Report going, here are the avg errors/departures from that forecast.

Code: Select all

MG   5.19      RR   5.79      ncs2 6.13
snd  5.42      yoop 5.88      BD   6.26
jgo  5.48      tzu  5.91      shad 6.36
sbs  5.65      lqi  5.93      ATC  6.45
kmed 5.67      cal  5.94      ncs1 6.56
amp  5.70      GK5  5.98      nre  6.91
taco 5.73      Crow 6.01      16py 7.41

http://www.basketball-reference.com/fri ... f_prob.cgi

square root of avg of squared errors:

Code: Select all

MG   6.42      RR   7.69      cal  7.98
jgo  7.05      yoop 7.72      ncs2 8.00
snd  7.28      lqi  7.73      BD   8.00
taco 7.53      shad 7.90      ATC  8.30
sbs  7.57      Crow 7.92      ncs1 8.39
amp  7.59      tzu  7.94      nre  8.47
kmed 7.59      GK5  7.95      16py 9.28

Big stories are how weak the GSW look and how well the Lakers are doing.
From 1/3 to 1/2 of our squared errors are due to just these 2.

[Mike G]

Ranked by projected surplus wins beyond what we predicted.
Error column is just b-r.com projected wins minus our avg prediction.

Code: Select all

err    tm    avg    proj
19.2   LAL   23.4   42.6
10.2   Chi   39.2   49.4
9.9    Atl   42.9   52.8
9.0    LAC   51.3   60.3
9.0    Brk   25.1   34.1
8.3    Cha   41.9   50.2
5.0    Mil   31.8   36.8
4.5    Mia   36.7   41.2
3.9    Min   38.3   42.2
3.2    Phx   28.3   31.5
1.2    Phl   23.7   24.9
0.4    Mem   39.3   39.7
-0.1   Den   36.0   35.9
-0.1   Okl   46.2   46.1
-0.3   Hou   46.5   46.2
-0.3   NYK   35.5   35.2
-0.8   Det   41.3   40.5
-1.0   Sac   36.1   35.1
-1.2   Tor   50.6   49.4
-2.9   Uta   48.2   45.3
-4.0   SAS   57.4   53.4
-4.8   Cle   56.8   52.0
-5.4   Was   41.3   35.9
-5.5   Ind   40.5   35.0
-5.7   Orl   35.4   29.7
-6.2   NOP   35.0   28.8
-8.8   Dal   37.9   29.1
-9.0   Por   43.9   34.9
-11.2  Bos   50.7   39.5
-16.5  GSW   68.6   52.1

The Dubs are 10-2 with SRS 6.52, expected to go 42-28 (.600) the rest of the way.
Lakers at 7-6 and 0.92 presume to go 36-33 hereafter. Their worst case scenario is 27 wins, 10 better than last year.
GSW best case is 66-16, a dropoff of 7.

East teams are currently 27-25 vs the West; apparently due to schedule advantage, since projections are for West teams to avg 41.5 wins vs 40.5 in the east.

[Mike G]

things change quickly

Code: Select all

avg error
MG   5.27      BD   5.80      lqi  6.01
kmed 5.40      amp  5.83      ncs2 6.01
snd  5.50      taco 5.84      shad 6.09
jgo  5.53      tzu  5.88      ATC  6.23
sbs  5.74      RR   5.90      nre  6.67
yoop 5.79      GK5  5.93      ncs1 6.71
Crow 5.79      cal  5.93      16py 7.32

RMSE
MG   6.48      amp  7.51      cal  7.80
jgo  6.89      RR   7.54      tzu  7.81
snd  7.16      Crow 7.55      GK5  7.81
kmed 7.29      shad 7.58      ATC  8.03
taco 7.37      lqi  7.59      nre  8.12
yoop 7.43      BD   7.65      ncs1 8.21
sbs  7.46      ncs2 7.73      16py 9.02

sum of sq.rts. of errors , squared
kmed 4.34      Crow 4.72      ncs2 4.99
snd  4.46      GK5  4.77      RR   4.99
MG   4.54      yoop 4.86      shad 5.06
jgo  4.59      cal  4.91      ATC  5.21
sbs  4.60      taco 4.91      nre  5.81
BD   4.68      amp  4.95      ncs1 5.91
tzu  4.71      lqi  4.96      16py 6.33

This last block -- SMRE? -- has the opposite skewing effect that RMSE has: Rather than exaggerate larger errors, they're compressed. Exacting predictions gain distinction over near misses.

[Mike G]

Relative to the power ratings shown here: viewtopic.php?f=2&t=9228

Code: Select all

MG   5.67      BD   6.35      RR   6.51
kmed 5.74      cal  6.37      shad 6.53
yoop 5.88      Crow 6.39      lqi  6.67
jgo  5.93      taco 6.45      ATC  6.70
snd  6.04      amp  6.45      ncs1 7.12
tzu  6.23      GK5  6.50      nre  7.44
sbs  6.25      ncs2 6.51      16py 7.73

avg absolute errors

[Mike G]

avg error

Code: Select all

MG   5.08      taco 5.47      ncs2 5.69
snd  5.10      BD   5.56      shad 5.80
jgo  5.32      lqi  5.57      cal  5.82
tzu  5.38      sbs  5.58      ATC  5.96
Crow 5.38      yoop 5.58      ncs1 6.24
amp  5.41      RR   5.60      nre  6.28
kmed 5.45      GK5  5.61      16py 7.61

[Mike G]

Amazing how everyone's errors are approaching zero. Teams seem to be doing just what they're supposed to be doing.

Code: Select all

Mike   4.32      trzu   4.62      shad   4.86
RyRi   4.36      jg34   4.68      ATCt   4.92
sndi   4.43      ncsD   4.71      cali   5.00
ampd   4.45      kmed   4.72      yoop   5.01
lnqi   4.45      BaDo   4.80      GK5.   5.15
sbs.   4.53      Crow   4.84      nrfo   5.27
taco   4.61      ncsB   4.84      16py   7.05

In just 2 days of games, the largest error among us has improved by 1.01
http://www.basketball-reference.com/fri ... f_prob.cgi

[Nathan]

I did over/under picks for fun just before the season started, so I have the preseason vegas lines for all the teams, which i took to be the median line on oddschecker a few hours before tip off of the season opener:

tm proj oddschecker
Atl 46 43.5
Bos 43 52.5
Brk 27 20.5
Cha 45 41.5
Chi 49 39.5

Cle 55 57.5
Dal 26 39.5
Den 39 36.5
Det 43 45.5
GSW 60 67.5

Hou 48 43.5
Ind 34 44.5
LAC 60 53.5
LAL 38 25.5
Mem 43 43.5

Mia 37 35.5
Mil 37 35.5
Min 36 41.5
NOP 35 36.5
NYK 39 39.5

Okl 43 44.5
Orl 29 36.5
Phl 25 23.5
Phx 29 28.5
Por 38 44.5

Sac 36 33.5
SAS 56 57.5
Tor 51 50.5
Uta 47 47.5
Was 35 42.5

Average error 4.5

As expected the vegas lines are doing well, but only slightly better than the median prediction here.

[Mike G]

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