2025-26 team win projection contest

[Crow]

Fwiw, the composite projection from Nov. 14 I assembled from other entries and 3 projections from mid-Nov. (page 5 of thread) has an average variance from current BRef projection of 4.91 wins.

My personal revised projection from then (page 6) has a variance of only 4.39.

Both closer than any pre-season entry.

A couple weeks of new season data and reflecting on it was quite helpful. Not a surprise, but wanted to see how much it helped.


Might check things again at the end.

[Mike G]

DARKO may be onto some secret sauce; and while they weren't in the contest last year, wasn't it determined they would have won it had they entered?
Winning 2 years in a row would be noteworthy, as for most entries it seems very random -- at least over these last 2 seasons.

Here are everyone's correlations between their predictions and DARKO's:

Code: Select all

corr.  darko  disco
.948   bmgm   .052
.942   Walr   .058
.939   Ebet   .061
.938   Crow   .062
.937   knar   .063
.933   dtka   .067
.929   bpmW   .071
.929   Mgoo   .071
.921   EExp   .079
.901   medi   .099
.891   eWin   .109
.864   WShr   .136
.862   perW   .138
.850   DQin   .150
.849   KPel   .151

The correlations all look at least "pretty high", so I added a dis-correlation column, (1-corr.) to accentuate the dissing between them.
I don't know if DARKO is influenced by any of these higher correlators, or vise versa. I kinda suspect Walrus is in cahoots.

[Crow]

Top 5 correlation with Darko for me is a positive.

I know I have some outliers. They are often my downfall. Can be rewarded but more often hurt. Half the gap for me from lead is due to my massive bet against the Heat. They were so terrible in playoffs and pre-season. However neither of those is the regular season. They are 6W-10L recently but the big early success buffers that so far. 1W-2L in last 3.

Probably should have been more restrained. Hard beat the best metrics at regression though. But also can beat myself. Got to be different to win, but it has to be right amount, right places to win against a strong field.

[Mike G]

The field averages .13 bigger error overnight and is overall worst since Nov. 21.

Code: Select all

.   avg err   rmse   r^2      .   avg err   rmse   r^2
DRKO   6.59   8.20   .48      Mgoo   7.34   8.99   .40
knar   6.84   8.34   .46      dtka   7.42   8.97   .40
bmgm   6.91   8.47   .50      eWin   7.52   9.16   .37
bpmW   6.93   8.54   .45      25pr   7.54   9.66   .32
Walr   6.94   8.54   .45      perW   7.55   9.84   .30
Ebet   7.08   8.52   .50      Crow   7.86   9.65   .37
EExp   7.20   8.66   .43      medi   8.19   9.76   .38
WShr   7.26   9.16   .38      KPel   8.71  10.39   .29
avgA   7.29   8.80   .42      DQin   9.06  11.05   .28

KPel with personal worst for the 4th day in a row, and the top-bottom spread is largest since mid-Nov.

How are we doing with our high and low guesses?

Code: Select all

W-L   who     Highest            Lowest
3-0   Walr   Uta                Dal LAC
3-1   DRKO   Was Uta            LAL Mem
3-1   Ebet   NYK LAL            Chi Uta
2-1   bmgm   Det NYK             Uta   
3-2   WShr   Brk             Mil Phl NOP SAS
3-2   eWin   Mia Orl Tor        Bos Cle
3-2   medi   Cle Phl Min Por     Phx
2-2   knar   Cha NOP            Ind Den
4-4-1 perW   Dal Den SAS   
.                   Cle GSW Hou Min OKC Por
0-0   bpmW
0-0   Mgoo             
2-3   Crow   Chi Det           Atl Mia Uta
0-1   dtka   LAC
4-6-1 DQin  Bos Ind Hou Mem OKC Phx        
.                        Cha Det NYK Orl Tor
1-4   KPel   Mil Tor GSW        Brk Was

Ties are included, defined as within half a win predicted.
OKC currently projects midway between the high and low guesses, so neither a W nor a L for either.
I think all other teams sit clearly closer to high or low end right now.

UPDATE Jan. 10 -- DARKO had their biggest lead of the season yesterday.

Code: Select all

.   avg err   rmse   r^2      .   avg err   rmse   r^2
DRKO   6.53   8.25   .47      Mgoo   7.17   9.02   .40
bpmW   6.77   8.53   .45      dtka   7.36   8.89   .40
knar   6.83   8.31   .46      25pr   7.39   9.58   .33
Walr   6.85   8.54   .44      eWin   7.42   9.19   .37
bmgm   6.94   8.44   .51      perW   7.59   9.91   .29
WShr   7.04   9.11   .38      Crow   7.78   9.63   .37
avgA   7.09   8.77   .42      medi   7.92   9.64   .39
EExp   7.09   8.61   .43      KPel   8.60  10.28   .30
Ebet   7.12   8.53   .50      DQin   8.90  10.93   .29

[Crow]

Heat lose 3rd in last 4 games to Pacers by 24 pts. Next 5 opponents are playoff ranked.

Adebayo loses his 25 minutes by 23, just -1 when he was off.

Worst 3pt shooting of their season at 13%.


And now another loss.

Scoring fell from 125 in October to 110 in January so far. Opponent scoring went from 114 to 119. 20 pt change in margin. Worth noticing to me.

[Crow]

In as continued discussion of my to date biggest miss in the projection contest and whether that miss sees significant reduction.

[Mike G]

Smallest errors of the season for BPM, WS, knarsu, dtkavana, betmgm, and 25pyth-regr.

Code: Select all

.   avg err   rmse   r^2          avg err   rmse   r^2
DRKO   6.06   7.65   .52      EExp   6.75   8.00   .48
bpmW   6.19   7.78   .51      dtka   6.91   8.18   .46
Walr   6.30   7.84   .50      eWin   6.96   8.54   .42
knar   6.35   7.66   .52      perW   7.09   9.25   .34
bmgm   6.45   7.77   .56      25pr   7.13   8.97   .37
avgA   6.51   8.05   .48      Crow   7.23   8.93   .42
WShr   6.58   8.44   .44      medi   7.34   9.00   .44
Mgoo   6.66   8.27   .46      KPel   7.97   9.51   .36
Ebet   6.70   7.85   .55      DQin   8.32  10.35   .33

Smallest top-bottom spread since before xmas; avgA with biggest lead of the month over EExp.

UPDATE Jan. 17 -- after 4 weeks, a new leader.

Code: Select all

.   avg err   rmse   r^2         .avg err   rmse   r^2
bpmW   6.08   7.62   .53      EExp   6.81   7.99   .49
DRKO   6.09   7.49   .54      dtka   6.88   8.11   .48
Walr   6.26   7.75   .51      eWin   6.93   8.42   .44
knar   6.26   7.52   .54      25pr   6.98   8.87   .39
avgA   6.45   7.91   .50      Crow   7.12   8.73   .44
bmgm   6.47   7.70   .57      perW   7.17   9.15   .36
WShr   6.54   8.28   .46      medi   7.27   8.89   .46
Mgoo   6.58   8.13   .48      KPel   7.94   9.38   .38
Ebet   6.72   7.79   .56      DQin   8.18  10.16   .35

Best numbers yet for BPM, knarsu, WS, dtka, Crow, medi, avgA, and 25pr

[Crow]

Your modifications on Mgoo elevate from tier 3 by eWin to tier 2.

Lots of tier 2 and tier 3 with small differences within tier and modest between.

About 50% of season to go. Movements will get slower but they are still coming

[Mike G]

Another season best for Crow and the 4th consecutive days of best for knar and 25pr .

Code: Select all

.   avg err   rmse   r^2         .avg err   rmse   r^2
bpmW   6.16   7.78   .52      25pr   6.90   8.96   .38
DRKO   6.21   7.65   .53      EExp   6.96   8.19   .47
knar   6.26   7.67   .53      dtka   6.98   8.29   .46
Walr   6.47   7.98   .49      eWin   7.00   8.56   .43
avgA   6.51   8.06   .49      Crow   7.07   8.86   .43
bmgm   6.58   7.89   .55      medi   7.29   9.04   .44
WShr   6.60   8.41   .45      perW   7.34   9.30   .35
Mgoo   6.65   8.28   .47      KPel   8.01   9.48   .37
Ebet   6.83   8.00   .54      DQin   8.23  10.26   .34

Our avgA is now leading EExp by .44, and the top-bottom is smallest since Dec. 11.

Noticing the West>East projections have gotten to what may be a season high avg of 42.2-39.8 -- still not as extreme as our consensus 43-39 imbalance, but likely producing the recent drop in most of our errors.
A full week of West outplaying the East:

Code: Select all

W > E      E > W
1-17
Phx-NYK   Mia-OKC
GSW-Cha
Den-Was

1-16
Sac-Was   Ind-NOP
LAC-Tor

1-15
GSW-NYK   Det-Phx
Mem-Orl   Cha-LAL
Por-Atl   Orl-Mem
SAS-Mil

1-14
LAC-Was   Chi-Uta
NOP-Brk
Sac-NYK

1-13
Hou-Chi   Mia-Phx
LAL-Atl
Min-Mil

1-12
Uta-Cle
Dal-Brk
LAC-Cha

1-11
Den-Mil   Atl-GSW
Mem-Brk   Orl-NOP 
OKC-Mia   NYK-Por  
Phx-Was

1-10
SAS-Bos   Chi-Dal
LAC-Det   Cle-Min
.         Cha-Uta

That's a stretch of W 18 - 7 E
https://www.basketball-reference.com/bo ... &year=2026

UPDATE Jan. 19

Code: Select all

.   avg err   rmse   r^2         .avg err   rmse   r^2
bpmW   6.18   7.80   .52      Ebet   6.89   8.05   .53
knar   6.31   7.71   .52      EExp   6.92   8.21   .47
DRKO   6.36   7.74   .52      dtka   6.94   8.27   .46
Walr   6.37   7.98   .49      eWin   6.96   8.61   .42
avgA   6.44   8.08   .48      Crow   7.13   8.87   .43
WShr   6.51   8.40   .45      medi   7.35   9.03   .44
Mgoo   6.61   8.32   .46      perW   7.38   9.36   .34
bmgm   6.64   7.94   .55      KPel   8.00   9.49   .37
25pr   6.89   8.92   .39      DQin   8.27  10.25   .34

Season low for WShr and avgA

Some other leaders at most favorable exponents:

Code: Select all

e = .10        e = .56       e = 1.75      e = 19.9
3.36  Mgoo    5.23  bpmW    7.40  knar    13.0  DRKO
3.97  bpmW    5.44  Walr    7.44  bpmW    13.3  knar
4.10  WShr    5.44  WShr    7.44  DRKO    13.3  bpmW
4.27  knar    5.49  knar    7.62  Walr    13.5  dtka
4.29  Walr    5.50  DRKO    7.66  bmgm    14.0  EExp
4.37  DRKO    5.56  Mgoo    7.79  Ebet    14.3  Ebet
4.73  Crow    5.85  bmgm    7.91  EExp    14.4  Walr
4.86  bmgm    6.07  eWin    7.94  Mgoo    14.8  Mgoo
4.96  eWin    6.08  Crow    7.97  dtka    14.8  bmgm

e 19.9 is the point where BPM overtakes knarsu for 2nd place; after which they close the gap some but never catch DARKO.

UPDATE Jan. 20 -- BPM with their 8th day of improvement vs the field.
Top and bottom are closer yet, and avgA continues to pull away from EExp.

Code: Select all

.   avg err   rmse   r^2          avg err   rmse   r^2
bpmW   6.11   7.78   .52      Ebet   6.93   8.10   .53
Walr   6.35   8.00   .49      eWin   6.95   8.56   .43
knar   6.35   7.70   .52      dtka   6.96   8.28   .46
avgA   6.44   8.06   .49      EExp   6.98   8.25   .46
DRKO   6.46   7.77   .52      Crow   7.11   8.78   .44
WShr   6.50   8.43   .45      medi   7.35   9.02   .45
Mgoo   6.65   8.28   .47      perW   7.47   9.35   .34
bmgm   6.66   7.97   .54      KPel   7.93   9.45   .37
25pr   6.93   8.99   .38      DQin   8.14  10.21   .35

West continues to clobber the East, 7-1 over the last 2 days. Now 42.5 - 39.5 avg projection.
Last year around this time, bpmW was knocked from the top spot by avgA for a few days, then TmTj until nearly the end of the season.

UPDATE Jan. 22 -- major shuffling in the 2nd and 3rd tiers; serious threat to the top.

Code: Select all

.   avg err   rmse   r^2          avg err   rmse   r^2
bpmW   6.38   7.98   .51      dtka   7.01   8.44   .45
Walr   6.40   8.12   .49      EExp   7.04   8.40   .46
knar   6.52   7.89   .51      25pr   7.06   9.15   .37
DRKO   6.60   7.91   .51      eWin   7.10   8.74   .41
avgA   6.63   8.26   .47      Crow   7.20   8.99   .43
WShr   6.68   8.64   .43      medi   7.60   9.21   .43
Mgoo   6.70   8.48   .45      perW   7.61   9.53   .33
bmgm   6.75   8.10   .54      KPel   8.04   9.61   .36
Ebet   6.99   8.23   .52      DQin   8.37  10.42   .33

Two whole days with nobody hitting their season best error -- this hasn't happened in a while:

Code: Select all

date    new lows 
1-11   WS 6.81, 25pr 7.22
1-12   bmgm 6.68, WS 6.75
1-13   bmgm 6.55, WS 6.66
1-14    none
1-15   bpm 6.19, knar 6.35, bm 6.45*
...   WS 6.58, dtka 6.91, 25pr 7.13
1-16   knar 6.34, 25 7.06, medi 7.29
1-17   bpm 6.08*, kna 6.26, avgA 6.45
... WS 6.54, dt 6.88* Cro 7.12, me 7.27*
1-18   kna 6.26*, 25 6.90*, Cro 7.07*
1-19   aA 6.44*, WS 6.51, 25 6.89*
1-20   WS 6.50*

* - still personal best

UPDATE Jan. 23 -- bpmWins lasted 6 days on top.

Code: Select all

.   avg err   rmse   r^2          avg err   rmse   r^2
Walr   6.43   8.13   .48      dtka   7.01   8.44   .45
bpmW   6.45   7.98   .50      EExp   7.04   8.40   .46
knar   6.56   7.88   .50      25pr   7.06   9.15   .37
WShr   6.59   8.65   .42      eWin   7.10   8.74   .41
DRKO   6.60   7.86   .51      Crow   7.20   8.99   .43
avgA   6.66   8.25   .47      medi   7.60   9.21   .43
Mgoo   6.75   8.47   .45      perW   7.61   9.53   .33
bmgm   6.92   8.13   .53      KPel   8.04   9.61   .36
25pr   6.98   9.18   .36      DQin   8.37  10.42   .33

Notice the RMSE leader is at #5; and that top 5 all have the lead at some exponent.

Code: Select all

e = .50-       e = .83       e = 1.26      e = 1.58      e = 20.2+
5.17  WShr    6.07  Walr    6.90  bpmW    7.37  knar    13.7  DRKO
5.27  Walr    6.15  WShr    6.94  knar    7.39  bpmW    13.9  knar
5.51  Mgoo    6.15  bpmW    6.94  Walr    7.39  DRKO    13.9  bpmW
5.53  bpmW    6.30  knar    6.98  DRKO    7.49  Walr    14.1  dtka
5.70  DRKO    6.33  DRKO    7.21  WShr    7.67  bmgm    14.7  EExp
5.75  knar    6.37  Mgoo    7.28  bmgm    7.83  Mgoo    14.9  Ebet
6.00  Crow    6.66  bmgm    7.28  Mgoo    7.86  Ebet    15.0  eWin
< .68          .70-1.04      1.05-1.46     1.48-1.85      > 1.86

Bottom line is the range in which the top entry leads.

UPDATE Jan. 24

Code: Select all

.   avg err   rmse   r^2          avg err   rmse   r^2
bpmW   6.36   7.86   .50      25pr   6.92   9.09   .36
Walr   6.43   8.00   .48      dtka   7.01   8.31   .45
DRKO   6.48   7.68   .52      Ebet   7.01   8.10   .52
WShr   6.48   8.56   .42      eWin   7.09   8.63   .41
knar   6.48   7.75   .50      Crow   7.09   8.79   .43
avgA   6.55   8.10   .47      medi   7.52   9.07   .44
Mgoo   6.65   8.34   .45      perW   7.53   9.28   .34
bmgm   6.77   7.94   .54      KPel   7.92   9.46   .37
EExp   6.92   8.20   .46      DQin   8.28  10.30   .33

[Crow]

Gap from the top narrowing but lots of traffic ahead of me.

Was 4 tiers, I'd say 3 now.

[Mike G]

Code: Select all

.   avg err   rmse   r^2          avg err   rmse   r^2
bpmW   6.22   7.71   .52      Ebet   6.82   7.94   .54
knar   6.34   7.67   .51      dtka   6.84   8.15   .46
Walr   6.40   7.86   .50      25pr   6.88   8.99   .37
avgA   6.41   7.97   .49      Crow   6.90   8.68   .44
DRKO   6.42   7.59   .52      eWin   6.95   8.51   .42
WShr   6.48   8.48   .43      medi   7.37   8.94   .45
Mgoo   6.50   8.21   .46      perW   7.38   9.15   .35
bmgm   6.56   7.74   .56      KPel   7.63   9.21   .39
EExp   6.76   8.03   .48      DQin   8.18  10.24   .33

Smallest errors of the season for WShr, dtka, and Crow; also for avgA and 25pr.

Update on the over- and under-achievement of teams to date, relative to the average of our predictions:

Code: Select all

over   tm   avgA   proj   age      over    tm   avgA   proj   age
14.4  Det   42.8   57.3   25.9    -16.4   Ind   39.3   23.0   26.2
13.3  Phx   33.9   47.2   27.3    -15.7   Sac   38.2   22.5   29.3
10.6  Tor   36.8   47.5   24.7    -11.8   Dal   45.6   33.8   26.3
10.4  SAS   42.5   52.9   25.7     -9.6   LAC   49.3   39.8   30.9
8.7   Mia   37.3   46.1   26.5     -8.7   NOP   32.8   24.1   23.9
8.4   Bos   43.1   51.5   26.7     -7.1   Mil   41.2   34.2   27.5
7.1   Phl   37.2   44.3   25.6     -6.7   Orl   48.9   42.3   24.6
5.9   Cha   29.9   35.9   23.9     -6.3   Mem   42.8   36.6   25.5
5.3   Por   36.1   41.5   24.7     -5.9   Cle   53.0   47.2   26.5
3.1   Chi   35.7   38.8   25.9     -3.7   Atl   43.4   39.8   24.8
2.9   Uta   24.0   26.9   24.5     -2.0   GSW   48.3   46.3   29.3
2.6   Hou   49.7   52.4   26.7     -1.3   Brk   25.7   24.5   23.7
1.2   LAL   44.0   45.2   27.7     -1.2   Was   20.5   19.4   23.8
0.8   OKC   62.7   63.6   25.2                 
0.5   Min   46.7   47.3   28.0                  
0.4   NYK   48.3   48.7   28.3
0.4   Den   50.6   51.0   27.7

Notice the oldest teams are well represented among the Unders. The Clipps are a full 1.5 years older than anyone else, but they are turning it around, for now.
The youngest teams are also doing less well than average-aged teams. In general the age:over correlation is -.23.
The strongest correlation is with abs(age-25.7) at -.36
League avg age is 26.2, so half a year younger than that is most likely to improve -- in this 1/2 year sample.

[DQuinn1575]

Is BPMw simply last year's BPM (per B-Ref) using Pelton's minutes and a default value for rookies?

[Mike G]

Yes.^^
I messed around with age adjustment and decided (again) not to bother.
Also "yes" for WShr, perW, and eWin.

Today's top-to-bottom spread is smallest yet! Down from 2.47 on Jan 9, to 1.91 now.
The whole field has improved since then, by an avg of .60 in MAE:
1.04 KPel
.89 DQin
.88 Crow
.80 Mgoo
.80 avgA
.73 WShr
.68 bpmW
.67 medi
.65 25pr
.63 dtka
.58 eWin
.47 Walr
.47 knar
.40 EExp
.40 bmgm
.37 Ebet
.24 perW
.16 DRKO

[Mike G]

Based on previous 13 seasons of rookies, these are the averages -- heavily "smoothed" -- for draft picks, that I assigned for 4 stats I entered.

Code: Select all

pk   PER   WS/48   BPM  e484
1   17.8   .141    1.8  1.34
2   17.7   .139    1.7  1.32
3   17.0   .131    1.1  1.20
4   16.1   .123    0.6  1.08
5   15.1   .111    0.1   .98
6   14.4   .102   -0.3   .90
7   13.7   .095   -0.7   .84
8   13.6   .093   -0.7   .82
9   13.5   .090   -0.8   .80
10  13.5   .089   -0.9   .80
11  13.3   .088   -0.9   .78
12  13.1   .087   -1.0   .76
13  12.9   .083   -1.1   .73
14  12.9   .082   -1.2   .72
15  12.7   .081   -1.3   .70
16  12.6   .078   -1.4   .69
17  12.2   .073   -1.6   .64
18  12.2   .073   -1.6   .64
19  12.2   .073   -1.6   .64
20  12.2   .072   -1.7   .64

25  12.0   .070   -1.9   .60
30  11.5   .065   -2.2   .56
35  11.1   .061   -2.4   .51
40  10.7   .052   -2.7   .47
45  10.7   .049   -2.8   .45
50   7.8  -.007   -5.3   .18
55   7.1  -.021   -5.4   .12
60   5.6  -.032   -7.3  -.07

After pick #16, there's a serious flattening-out, such that #17 is hardly expected to be better than #25.
After #45 it falls off sharply, and those draftees who never played -- and teams that didn't bother to draft anyone -- would presumably bring the averages down even more.

[Crow]

These are career to date impact estimates or to a certain year or peak?