2024-25 team win projection contest

[Crow]

Not sure if any other projections explicitly factor in coaching. I doubt there is a way to do it objectively. I consider it subjectively. Top 5s in West and East have well regarded Coches and coaches new to team, better than before.

Top 5s have 6 younger than average teams, 3 slightly older than median and just one clearly older.

[Mike G]

On or about Dec. 16, we all enjoyed our lowest errors of the season. Everyone's gotten worse since then, and a few of us now are at our worst. *

Code: Select all

err   12/16   1/01   dif      err   12/16   1/01   dif
WShr   5.70   6.23   .53      24pr   6.56   7.12   .56
bpmW   5.77   6.30   .54      Crow   6.56   7.14   .58
KPel   5.84   6.51 * .67      medi   6.39   7.15   .76
avgA   5.87   6.51   .64      emin   6.80   7.62   .82
TmTj   5.68   6.57   .88      24py   7.68   7.78   .10
dtka   6.23   6.67   .44      eWin   6.90   7.81 * .91
eExp   6.09   6.68 * .59      perW   7.52   8.35 * .83
vegas  6.43   6.73   .30      bmgm   8.21   8.46   .25
DQin   5.86   7.00  1.14      4141   9.44   9.34  -.10

https://www.basketball-reference.com/fr ... _prob.html
The top to bottom difference is the largest since Nov.

UPDATE Jan. 4

Code: Select all

w  Atl   w  Bos   w  Brk   w  Cha   w  Chi   w  Cle   w  Det   w  Ind
44  pW   62  em   34  WS   34  pW   36 b-r   64 b-r   38 b-r   48  pW
41  WS   61  DQ   33 b-r   33  em   35  WS   52  dt   33  eW   47  mi
41  dt   61  mi   31  eW   31  Cr   35  bW   50  mi   33  pW   47  Cr
41  bW   60  dt   30  pW   31  eW   34  eW   50  Cr   32  TT   46  dt
39 b-r   60 b-r   29  bW   30  DQ   34  pW   49  em   32  bW   44  em
39  DQ   59  Cr   29  DQ   30  WS   31  TT   48  DQ   32  WS   44  TT
39  eW   57  TT   26  dt   29  dt   30  dt   48  TT   28  em   43  bW
38  mi   57  WS   26  mi   28  bW   29  Cr   46  WS   28  dt   43  WS
37  TT   56  bW   24  TT   26  TT   28  DQ   46  bW   27  Cr   42  DQ
35  Cr   52  pW   23  em   25 b-r   27  em   45  eW   26  DQ   41  eW
34  em   52  eW   18  Cr   24  mi   25  mi   44  pW   26  mi   39 b-r
                                             
w  Mia   w  Mil   w  NYK   w  Orl   w  Phl   w  Tor   w  Was      
47  em   50  dt   58  em   49  Cr   55  mi   32  pW   34  eW      
45  eW   49  DQ   54 b-r   48  mi   53  eW   32  WS   33  pW      
44 b-r   48  em   53  Cr   47 b-r   52  pW   32  em   28  WS      
43  bW   48  Cr   50  TT   46  dt   51  TT   32  bW   28  bW      
43  dt   45  pW   50  dt   44  WS   47  DQ   30  eW   21  dt      
43  WS   45  TT   48  mi   43  DQ   47  dt   30  dt   20 b-r      
42  mi   44  mi   44  WS   43  TT   46  WS   29  mi   19  em      
42  DQ   44  bW   43  DQ   42  em   46  Cr   29  DQ   19  Cr      
42  TT   44  WS   43  bW   41  bW   46  bW   29  TT   18  mi      
41  pW   43 b-r   41  eW   39  eW   43  em   26 b-r   15  TT      
40  Cr   43  eW   38  pW   34  pW   34 b-r   25  Cr   14  DQ      
                                            
w  Dal   w  Den   w  GSW   w  Hou   w  LAC   w  LAL   w  Mem   w  Min
56  DQ   53  em   52  Cr   52 b-r   45 b-r   45  TT   53 b-r   57  em
53  pW   51  mi   49  DQ   47  eW   45  DQ   44  Cr   49  Cr   55  DQ
52  eW   51  dt   49  TT   46  mi   44  WS   44  dt   48  em   53  eW
52  TT   51  eW   48  mi   45  bW   40  bW   43  DQ   45  DQ   52  mi
51  Cr   49  pW   47  em   44  TT   40  dt   42  em   43  dt   51  Cr
50  bW   49  DQ   46  bW   44  Cr   36  TT   42 b-r   43  eW   50  WS
48 b-r   49  Cr   46 b-r   44  WS   35  mi   39  mi   43  mi   50  bW
47  WS   47  WS   45  dt   43  pW   35  pW   39  bW   42  pW   50  dt
47  dt   47 b-r   44  WS   38  dt   33  Cr   38  WS   41  bW   48  TT
47  mi   47  bW   41  eW   36  DQ   32  em   37  pW   41  TT   46  pW
43  em   46  TT   40  pW   33  em   31  eW   36  eW   34  WS   44 b-r
                                             
w  NOP   w  OKC   w  Phx   w  Por   w  Sac   w  SAS   w  Uta     
49  mi   65 b-r   51  TT   31  eW   50  Cr   41 b-r   33  pW      
49  em   62  mi   50  mi   28  pW   49  bW   38  em   32  eW      
48  Cr   61  DQ   49  dt   27  bW   48  pW   37  pW   32  WS      
45  DQ   59  TT   48  Cr   26  DQ   47  eW   36  eW   31  mi      
44  bW   58  em   48  pW   26  Cr   47  DQ   36  bW   29  DQ      
44  pW   55  WS   47  bW   26  WS   47  TT   34  WS   29  dt      
44  TT   54  bW   47  WS   26 b-r   47  dt   34  TT   29  bW      
44  WS   54  Cr   46  em   24  TT   46  WS   31  DQ   28  em      
44  eW   54  dt   45  eW   24  mi   44  em   31  dt   27  TT      
42  dt   52  pW   43  DQ   23  em   42  mi   31  mi   26  Cr      
19 b-r   49  eW   38 b-r   22  dt   39 b-r   29  Cr   23 b-r      

There are still 15 teams either above or below all our guesses, and a few more very close to the limits.
The West has 10 of the 15 unpredictables.

UPDATE Jan. 5 -- No more tiers!.. A tie for the lead, and other top 11 all separated by .05 to .10

Code: Select all

.  avg err   rmse   r^2      .  avg err   rmse   r^2
bpmW   6.30   8.3   .49      eExp   6.85   8.8   .52
TmTj   6.30   8.3   .52      Crow   6.92   9.3   .46
avgA   6.37   8.4   .49      24pr   7.12   9.5   .34
WShr   6.42   8.6   .46      emin   7.46   9.7   .42
dtka   6.52   8.2   .52      24py   7.54  10.9   .34
KPel   6.59   8.1   .56      eWin   7.80   9.8   .31
DQin   6.66   8.8   .49      bmgm   8.30  11.3   .35
medi   6.71   9.4   .45      perW   8.40  10.2   .25
vegas  6.78   9.0   .47     

Worst leading number in a month. Worst for KPel and eExp

[Crow]

1/3rd less gap for me to the leader than 4 days ago. Crowded ladder but I'll take progress toward the goal. Next step will be passing Vegas and ESPN experts, sometime soon I hope.

Thanks for the by team grid.

[Mike G]

Leaders at exponent/root other than 1 and 2:

Code: Select all

. e = .10         e = .37         e = .70         e=1.55          e = 4.0
Crow   4.11     WShr   4.83     TmTj   5.56     bpmW   7.29     KPel   10.6
TmTj   4.23     TmTj   4.85     bpmW   5.57     TmTj   7.31     dtka   11.0
WShr   4.25     bpmW   4.92     WShr   5.57     KPel   7.31     eExp   11.7
24py   4.32     medi   5.04     dtka   5.81     dtka   7.43     TmTj   11.8
DQin   4.35     DQin   5.05     DQin   5.86     WShr   7.50     bpmW   11.9
medi   4.35                                    

It's been suggested we 'unify' the leader/winner by an avg of the MAE and RMSE; which can also be done with an exponent around 1.45. And sometimes a contestant leading in that range is neither the leader at e = 1 or e = 2. In which case we could have 3 'winners'.
Pelton may have clinched the RMSE by one prediction: His worst guess is also the best (lowest) guess for the Pelicans. He's only off by 23. Everyone else has a bigger single error, as well as 2nd worst, 3rd, etc.

UPDATE Jan. 7

Code: Select all

.  avg err   rmse   r^2      .  avg err   rmse   r^2
TmTj   6.15   8.2   .53      eExp   6.66   8.7   .54
bpmW   6.16   8.2   .51     vegas   6.67   8.8   .49
WShr   6.23   8.5   .47      Crow   6.81   9.1   .47
avgA   6.25   8.3   .50      24pr   6.89   9.5   .35
dtka   6.33   8.1   .53      emin   7.36   9.6   .42
KPel   6.48   8.0   .58      24py   7.55  10.8   .35
DQin   6.53   8.7   .50      eWin   7.61   9.7   .32
medi   6.57   9.3   .46      bmgm   8.14  11.2   .36
.                            perW   8.17  10.1   .27

[DQuinn1575]

You have my CHI & CHO/CHA win totals mixed up.
It helps me just a smidge for now.
Thanks

Going for most improved, after a poor rookie season ;


UPDATED 10/21

ATL 39
BOS 61
BRK 29
CHI 30
CHO 28
CLE 48
DAL 56
DEN 49
DET 26
GSW 49
HOU 36
IND 42
LAC 45
LAL 43
MEM 45
MIA 42
MIL 49
MIN 55
NOP 45
NYK 43
OKC 61
ORL 43
PHI 47
PHO 43
POR 26
SAC 47
SAS 31
TOR 29
UTA 29
WAS 14

[Mike G]

yeah, sorry about that; but when Charlotte is abbreviated as CHO, it changes its order among teams, alphabetically -- falling after CHI, rather than before.
Cha being worse and Chi being better are both good for you, by .17 -- moving you up one notch in the standings.

I know this is how b-r.com does it, and it's a minor pain that I don't understand.
I've asked them why they have the East on the left (W) and West on the right (E) side of the page.
And why they unnecessarily bloat a column with "W/L%" when it's just W%.
In both cases, "that's just how we've always done it."

I suspect the CHO situation is because CHA was used for the original Hornets who in 2002 traded their entire team to the expansion Pelicans and got nothing in return. Well but they were the NO Hornets for 11 years, and Charlotte re-emerged as the Bobcats, for whom CHA was a perfectly OK abbreviation. So when the Hornets name became available, all the ex-CHA Hornets became proto-Bobcats/Hornets.

[TeemoTeejay]

The stdev of the 4 metric projections is around 8 or less ish, I think the other ones are generally 10-12 or so (13 was last years stdev), although I guess mines a metric prediction too (10.7 I think)

Curious to see what the results would be with the projections regressed the same way to 41 wins for the other predictions as well

[Mike G]

Here is what everyone should have regressed (more or less) to fit best with the current b-r.com projections.
The (additional) regression in the first column works like this:
x is your guess on a team, and R is your optimal current regression to 41.
x2 = x*(1-R) + 41*R

If you guessed Phx will win 51 games, but they are headed for 36, your error for that guess is 15.
If you had regressed your slate of guesses by .17, you have 49 for Phx, and your error is only 13. Roughly.

From least in need of regression* to most.

Code: Select all

regr. to 41    MAE   rmse   r^2
24pr   -.41   6.51   9.59   .35
KPel   -.18   6.34   7.98   .54
eExp   -.17   6.52   8.71   .45
WShr   -.14   6.16   8.52   .47
bpmW   -.04   6.14   8.18   .51
avgA    .07   6.18   8.23   .50
TmTj    .16   5.83   7.97   .53
medi    .16   6.12   8.79   .46
DQin    .19   6.01   8.19   .51
dtka    .24   6.20   8.01   .53
vegas   .24   6.24   8.37   .49
eWin    .27   7.54   9.71   .32
bmgm    .28   6.84   9.68   .36
Crow    .29   6.45   8.45   .47
24py    .29   6.51   9.60   .35
emin    .39   6.52   8.89   .42
perW    .45   7.92  10.18   .27

I just toggled the R to get the best MAE for each entry; RMSE landed wherever.
PER was substantially de-regressed when I submitted their line. Even with this adjustment -- leaving everyone else unchanged -- their rank doesn't change.
* -- disclaimer of sorts: Our avg here is .12 to fit with current b-r.com results. This may be close to the regression they currently apply. And that will fade to zero as the season winds up.

Today a new/old leader:

Code: Select all

.  avg err   rmse   r^2      .  avg err   rmse   r^2
bpmW   6.15   8.3   .51      medi   6.68   9.4   .45
WShr   6.21   8.6   .47     vegas   6.68   8.9   .48
TmTj   6.23   8.3   .53      24pr   6.91   9.6   .34
avgA   6.26   8.4   .50      Crow   6.93   9.2   .47
dtka   6.41   8.2   .53      emin   7.45   9.6   .42
DQin   6.45   8.7   .51      eWin   7.56   9.7   .32
KPel   6.50   8.0   .57      24py   7.73  10.9   .34
eExp   6.65   8.7   .53      perW   8.20  10.1   .27
.                            bmgm   8.25  11.3   .35

[DQuinn1575]

yeah, sorry about that; but when Charlotte is abbreviated as CHO, it changes its order among teams, alphabetically -- falling after CHI, rather than before.
Cha being worse and Chi being better are both good for you, by .17 -- moving you up one notch in the standings.

I know this is how b-r.com does it, and it's a minor pain that I don't understand.
I've asked them why they have the East on the left (W) and West on the right (E) side of the page.
And why they unnecessarily bloat a column with "W/L%" when it's just W%.
In both cases, "that's just how we've always done it."

I suspect the CHO situation is because CHA was used for the original Hornets who in 2002 traded their entire team to the expansion Pelicans and got nothing in return. Well but they were the NO Hornets for 11 years, and Charlotte re-emerged as the Bobcats, for whom CHA was a perfectly OK abbreviation. So when the Hornets name became available, all the ex-CHA Hornets became proto-Bobcats/Hornets.

Thanks, always a pain as different sites use differrent abbreviations.

lol - I've always seen sports standings with East on the left or the top and West on the right or bottom that I never get it a thought that it is the opposite of a map. I guess they do it that way because it's the way people are used to seeing it that way. For me, it would be weird to see the West on the left for standings.

[v-zero]

Thanks, always a pain as different sites use differrent abbreviations.

Hell yeah it is. I've a whole discord bot whose entire purpose is to bother me on my phone whenever it can't automatically marry up player/team etc IDs between differently sourced datasets.

I've got it pretty good now, it is able to make an accurate marriage 95% of the time. The other 5% I get a message and reply with the answer.

Pretty rare these days that I get a message, but sometimes when a player comes up from the G league for a game it has to ask for help.

[Crow]

How about the standings at least once based on actual records to date?

Compared to January 9, 2024 standings, 2 of top 8 in east regressed, 6 stayed about the same. 3 in bottom 7 got worse, 2 better, 2 little change. General regression is a theory that doesn't match with these results.

In west top 8, 6 stayed about the same, 2 got better. The bottom 7 was again a jumble with no dominant direction.

A projection that things would tend to stay about the same from here would appear to fit the experience of last season well. Very little to support general regression from here.

BRef projection is actually for mostly maintaining current performance. 7 cases given major regression to mean. Schedule could be a part of it but I am still generally dubious of major moves as projected. I don't expect the best to falter much if any and I don't expect the bottom cases to all rise.

[Mike G]

OK, just giving teams their current W% times 82:

Code: Select all

.   avg err   rmse   r^2      .   avg err   rmse   r^2
TmTj   7.56   10.0   .50      KPel   8.29   10.2   .51
dtka   7.80   10.0   .50      Crow   8.40   10.8   .44
DQin   7.92   10.6   .45      eExp   8.41   11.0   .49
bpmW   7.95   10.5   .46      24pr   8.68   11.8   .31
avgA   8.10   10.4   .46      emin   8.94   11.4   .38
WShr   8.12   10.9   .42      24py   9.10   12.5   .31
medi   8.15   11.0   .42      eWin   9.41   11.9   .29
vegas  8.17   10.6   .45      bmgm   9.48   12.8   .32
.                             perW   9.60   12.3   .25

And adding current wins to (remaining games * pythagorean W%)

Code: Select all

.  avg err   rmse   r^2      .  avg err    rmse   r^2
TmTj   6.91   9.4   .54      WShr   7.61   10.4   .46
avgA   7.03   9.7   .52      KPel   7.66    9.5   .56
DQin   7.06   9.9   .50      emin   7.80   10.5   .45
vegas  7.22   9.7   .52      eExp   7.98   10.5   .56
dtka   7.26   9.3   .55      24pr   8.63   11.5   .33
bpmW   7.30   9.9   .51      eWin   8.67   11.2   .35
medi   7.42  10.4   .46      24py   8.70   12.1   .33
Crow   7.56   9.9   .51      bmgm   8.92   12.4   .34
.                            perW   9.10   11.7   .29

Pyth wins are just the nearest integer posted at b-r.com, so pretty rough.

How do last season's pyth wins compare to this years projections (current + Pyth)?

Code: Select all

west  24py   impr     east   24py   impr
OKC   57.4    9.6      Bos   64.6   -2.4
Min   56.3  -11.7      NYK   52.0    4.2
Den   53.4   -4.3      Phl   48.3  -14.5
NOP   51.4  -35.2      Ind   47.8   -5.8
LAC   48.8   -5.6      Mil   47.0   -1.2
Phx   48.2  -12.1      Cle   46.9   22.2
GSW   47.1   -4.8      Orl   46.0    1.5
Dal   46.1    4.1      Mia   45.4   -1.9
Sac   45.1   -1.6      Chi   37.5   -2.3
Hou   43.7   12.3      Atl   36.0    2.4
LAL   42.4   -1.9      Brk   34.0   -6.4
Uta   30.0   -6.2      Tor   26.5   -6.3
SAS   26.3   13.6      Was   21.3   -7.3
Mem   24.5   32.3      Det   21.1   18.6
Por   20.7    5.0      Cha   18.7    4.1
correl:   -.53        correl:    -.12

In the west, 3 of the 6 best teams are headed to double-digit dropoffs in wins. And 3 of the bottom 6 doing the opposite; hence the large negative correlation.
League correlation is -.35 at the moment.

[Crow]

Thanks for checking.

[Mike G]

One thing b-r.com does (that likely matters) is strength of schedule, to date and for remaining games.
Knicks' schedule is 2.6 ppg tougher the rest of the way; Cavs 1.8, Grizz 2.4, Kings 1.5
Wiz' sched lightens up by 3.2 ! Raps 2.8, Bulls 1.4, Dubs 1.7, Clips 1.4

UPDATE to Earth Jan. 10

Code: Select all

sol
bpmW   5.99
TmTj   6.08
avgA   6.10
WShr   6.18
dtka   6.27
DQin   6.30
KPel   6.38
m-v-e   
medi   6.50
vegas  6.54
eExp   6.60
mars   
Crow   6.79
24pr   6.87
asteroid   
     belt
jupiter   
emin   7.21
   
saturn   
eWin   7.47
   
24py   7.62
   
   
uranus   
perW   7.99
   
neptune   
bmgm   8.22
	
	

	
	
	
	
pluto	
4141	9.17

UPDATE Jan. 11 -- best overall numbers since before xmas.

Code: Select all

.  avg err   rmse   r^2      .  avg err   rmse   r^2
bpmW   5.89   8.2   .53      eExp   6.50   8.7   .56
TmTj   5.93   8.2   .54      medi   6.51   9.4   .46
avgA   5.96   8.3   .52      Crow   6.79   9.1   .49
WShr   6.14   8.6   .48      24pr   6.80   9.6   .35
dtka   6.26   8.1   .54      emin   7.23   9.5   .44
DQin   6.26   8.6   .52      eWin   7.36   9.6   .35
KPel   6.31   7.9   .58      24py   7.71  10.9   .35
vegas  6.44   8.8   .50      perW   7.78  10.0   .30
.                            bmgm   8.14  11.2   .36

UPDATE Jan. 14

Code: Select all

.  avg err   rmse   r^2      .  avg err   rmse   r^2
bpmW   6.01   8.2   .53      eExp   6.80   8.9   .55
TmTj   6.02   8.2   .55      Crow   6.93   9.1   .48
avgA   6.11   8.3   .52      24pr   6.96   9.6   .35
WShr   6.25   8.6   .49      emin   7.23   9.5   .44
dtka   6.29   8.2   .54      eWin   7.39   9.7   .34
medi   6.41   9.3   .47      24py   7.63  10.9   .35
DQin   6.42   8.7   .51      perW   7.87  10.0   .30
vegas  6.44   8.8   .50      bmgm   8.15  11.3   .36
KPel   6.50   8.0   .58               

[Crow]

Average of APBRmetric sets is doing well.