2025-26 team win projection contest

[Crow]

This adjustment was fast enough to deserve acceptance, imo.

The detailed explanation was sufficient.

[Mike G]

As usual I've taken the averages of APBR guesstimates for each team and will include that (avgA) in the contest updates.
The average guess for the Sixers is 37.4 wins, with an enormous range from 31.7 (WinShares) to 47.4 (mediocre).
That +10.0 difference for medi is the biggest one in the field. Here are the entries and the teams on which they have the highest or lowest guesses, with distance from avgA:

Code: Select all

DQin   Hou+8  Ind+7  Bos+7  OKC+6  Phx+5  Mem+5      
.                       NY-5  Orl-7  Cha-7  Det-8  Tor-9
medi   Phl+10  Por+7  Min+5  Cle+5          Sac-5  Phx-6
EBET   NYK+5  LAL+4                         Chi-4  Uta-5
Crow   Chi+6  Det+3                         Atl-4  Mia-8
KPel   GS+8  Sac+5  Tor+4  Mil+3            Was-6  Brk-8
DRKO   Was+5  Uta+4  Cle+4                  Mem-5  LAL-5
EExp   Atl+5                                       OKC-5
Mgoo                        
dtka   LAC+4                     
eWin   Orl+6  Tor+4  Mia+4     Cle-4  Ind-4  Den-4  Bos-6
bpmW                        
WShr   Brk+8                    Mil-4  NO-4  SA-4  Phl-6
perW   Dal+8  SA+7  Den+4  NO+2            
.                Cle-4  Hou-4  OKC-5  Min-5  Por-6  GS-6
Walr   Cha+6 Uta+4                          LAC-3  Dal-6

Again sorted by avg deviation from 41.
There are a handful of ties -- defined as within 0.1 -- but all are rounded to integers.
Using Pelton minutes, WShr is 8 wins above avgA on the Nets, while KPel is 8 below. Amazing?

[knarsu3]

I don't know if it's too late but my win projections are here:

https://x.com/knarsu3/status/1980768011549061572

spreadsheet link: https://docs.google.com/spreadsheets/d/ ... 1870481632

I did actually post them before the season started (twitter link has the time) but I know Mike G is strict on submissions.

[Crow]

I saw the on-time publication and given that, this request, previous participation in contest and board and leadership in analytic community in general and the intended friendliness of this contest, I consider the request reasonable and acceptable.

Participation is encouraged and would only be discouraged or refused in the case of improper behavior.

[DQuinn1575]

I agree

[Mike G]

knar is excluded from this update only
After 1 or 2 games, advantage is clearly to the timid.

Code: Select all

. avg err          avg err
7.17   Walr      7.93   WShr
7.34   eWin      8.25   KPel
7.59   perW      8.68   Crow
7.65   bpmW      9.05   EBET
7.67   Mgoo      9.42   medi
7.69   DRKO     10.19   25Py
7.73   EExp     10.25   DQin
7.81   dtka     11.12   2025

b-r.com projections range from 47.9 (Mil) down to 30.8 (Dal).
These correspond to alleged SRS of +15.4 and -23.2
The current dummies are unregressed, straight 2025 wins and Pythag. The 50% regressed versions would still be in the bottom half.

UPDATE Oct. 26
2-3 games in, Mia and Por, both 1-1, lead their conferences in SRS.
https://www.basketball-reference.com/fr ... _prob.html

Code: Select all

.  avg err   rmse          avg err   rmse
Walr   6.93   8.8      25pr   8.01    9.4
knar   7.03   8.7      perW   8.16    9.9
EExp   7.32   9.0      WShr   8.17    9.8
eWin   7.38   9.3      Crow   8.36   10.4
bpmW   7.65   9.4      KPel   8.55   10.3
DRKO   7.68   9.2     vegas   8.57   10.7
Mgoo   7.73   9.6      25re   8.72   10.5
avgA   7.74   9.5      medi   8.94   10.7
dtka   7.80   9.3      DQin  10.53   12.3

knarsu replaces Walrus with smallest ave-dev.
avgA does not include dummies, EExp, or vegas.
'25 regressed and '25 pythagorean regressed are 75% last year and 25% =41.

UPDATE Oct 27 because it's fascinating the daily changes at this point in the season.
Heat vs Spurs, most likely Finals matchup -- currently 9%

Code: Select all

.  avg err   rmse          avg err   rmse
knar   7.75   9.2      bpmW   8.67   10.2
Walr   7.85   9.5      25pr   8.78   10.3
DRKO   8.13   9.6      WShr   8.91   10.5
EExp   8.15   9.8      Crow   9.10   11.2
perW   8.36  10.0      medi   9.44   11.7
avgA   8.45  10.1     vegas   9.44   11.5
eWin   8.49   9.9      25re   9.67   11.4
dtka   8.64  10.2      KPel   9.72   11.5
Mgoo   8.65  10.2      DQin  11.23   13.1

At 2-3 games, correlations all are under .20, r^2 are nil. One corr. is subzero.

note: The East is currently 20-23 vs the West. This W% (.465) is close to the avg of our E/W predictions (.476)
Every entry likes the West by at least a 42-40 avg -- the avg avg is 43.0 - 39.0
Somehow, b-r.com projections avg 42.5 - 39.5 in favor of the East.

[DQuinn1575]

hi,
can you list who abbreaviation is for?
Thanks

[Mike G]

Key to Abbreviations:
knar - knarsu3
Walr - Walrus
DRKO - DARKO (by kmedved)
- next 5 submitted by myself:
WShr - Win Shares
perW - PER Wins
bpmW - BPM Wins
eWin - eWins by yours truly
Mgoo - me again, subjective/intuitive
dtka - dtkavana (Crafted NBA)
Crow - Crow
medi - mediocre
KPel - ESPN's Kevin Pelton, founder of this site
DQin - DQuinn1575
avgA - average of all entries but EExp, vegas, and 'dummies' (currently 25re and 25pr)
vegas - aka ESPN BET, see the Kevin Pelton link
EExp - ESPN Experts
25re - 2025 wins regressed
25pr - 2025 Pythagorean-expected wins, regressed

[Mike G]

Big improvements overnight.

Code: Select all

.  avg err   rmse         avg err    rmse
knar   6.92   8.8      WShr   7.83    9.8
DRKO   7.02   8.8      bpmW   7.87    9.6
Walr   7.13   8.9      Mgoo   7.88    9.7
EExp   7.43   9.1      Crow   8.00   10.5
avgA   7.64   9.4      25re   8.52   10.6
25pr   7.66   9.4      medi   8.66   10.7
perW   7.69   9.7      KPel   8.72   10.7
eWin   7.72   9.6     vegas   8.79   10.7
dtka   7.80   9.3      DQin  10.25   12.3

"Errors" are relative to b-r.com's current projections. Comparing these to our average predictions, ranked by overachievement to date -- thru just 3 or 4 games:

Code: Select all

over   tm   avgA   b-r       over    tm   avgA   b-r
19.3  Uta   24.0   43.3     -19.9   OKC   62.7   42.8
18.0  Cha   30.2   48.2     -13.8   Dal   45.6   31.8
15.0  Was   20.5   35.5     -11.0   Min   46.7   35.7
11.9  Phl   37.2   49.1     -10.7   Atl   43.4   32.7
11.7  Mia   37.3   49.0      -9.8   Orl   48.9   39.1
9.1   Chi   35.4   44.5      -9.5   Ind   39.3   29.8
8.4   Por   36.1   44.5      -6.6   Den   50.6   44.0
8.3   Brk   25.7   34.0      -5.8   Mem   42.8   37.0
5.6   SAS   42.5   48.1      -5.6   LAC   49.3   43.7
2.8   Bos   43.1   45.9      -5.3   Hou   49.7   44.4
1.6   Mil   41.2   42.8      -4.6   Cle   53.0   48.4
1.2   NOP   32.8   34.0      -4.2   GSW   48.3   44.1
0.8   NYK   48.3   49.1      -3.7   LAL   44.0   40.3
0.7   Det   42.8   43.5      -1.5   Tor   36.8   35.3
.                            -1.4   Sac   38.2   36.8
.                            -1.3   Phx   33.9   32.6
over  avg   35.5   43.7      under  avg   45.8   38.7

[Crow]

BRef projections look highly regressed this early. None projected over 50 and none under about 30? Final distribution is likely to be considerably wider.

[Mike G]

BRef projections look highly regressed this early. None projected over 50 and none under about 30? Final distribution is likely to be considerably wider.

Yes, it's always this way, and I think b-r.com projections have gotten more consistent in the last couple years.
The Hornets are winning by 12.6 ppg, which would put them on track to win 64 if they keep it up. The 48 seen above is probably a lot closer to where they will close.

This year I have made a conscious effort to keep those entries under my control close to the avg distribution. The regressed 2025 wins and Pyth wins are 75% last year's numbers + 25% =41. Their ave-dev falls into the 7-8 range this way, roughly where everyone else averages. If it were 50-50, they'd be among the leaders at this stage of the season.

Most of those holding down the bottom of the 'standings' now are perennial contenders who will be expected to improve steadily as the league reaches equilibrium. As they say, the cream rises.

[Mike G]

Not much change in the standings or the overall errors, so here is a first installment of leaders at various exponents.
Normal updates show e = 1 (MAE) and e = 2 (RMSE)

Code: Select all

e = .25--        e = 2         e = 2.8         e = 10        e = 60++
5.38  knar     8.69  Walr     9.78  DRKO     13.6  EExp     17.0  dtka
5.40  avgA     8.79  knar     9.84  EExp     14.0  dtka     17.1  EExp
5.74  Walr     8.84  DRKO     9.84  knar     14.1  eWin     17.2  eWin
6.23  EExp     8.91  EExp     9.88  Walr     14.2  knar     17.9  Mgoo
6.27  Crow     9.23  dtka    10.14  dtka     14.6  Mgoo     18.3  knar
6.34  DRKO     9.29  avgA    10.31  avgA     14.6  DRKO     18.4  avgA

These are maximum separations between #1 and #2 in the interval where #1 is #1.
Below .25 and above 60, the gap just gets bigger; hence the -- and ++

Dummy entries are not included in this ^ survey. The one where every team is predicted at 41 wins is currently the leader at every exponent, and by wide margins.

UPDATE Oct. 30

Code: Select all

.  avg err   rmse          avg err   rmse
Walr   7.30   9.0      bpmW   7.83    9.6
DRKO   7.34   9.1      eWin   7.88    9.9
25pr   7.43   9.1      dtka   7.92    9.4
EExp   7.56   9.3      25re   8.27   10.3
avgA   7.73   9.6      perW   8.29   10.1
knar   7.74   9.2      medi   8.35   10.5
WShr   7.74   9.7      KPel   8.48   10.4
Crow   7.78  10.5     vegas   8.63   10.7
Mgoo   7.82   9.9      DQin   9.90   12.1

Last year's log started on this date, with WinShares on top at 5.70 MAE

[Mike G]

Overnight improvements of .10 to .52, averaging .36 lower.

Code: Select all

.   avg err   rmse   r^2      .   avg err   rmse   r^2
Walr   6.87   8.39   .20      eWin   7.51   9.18   .11
DRKO   6.89   8.29   .21      bpmW   7.57   8.95   .16
25pr   7.05   8.62   .18      dtka   7.66   8.85   .16
EExp   7.14   8.59   .19      perW   7.71   9.39   .10
knar   7.23   8.51   .14      25re   7.90   9.70   .14
avgA   7.44   8.86   .00     vegas   8.21   9.89   .20
Mgoo   7.46   9.10   .15      medi   8.26   9.89   .22
Crow   7.47   9.73   .17      KPel   8.28   9.75   .17
WShr   7.50   9.16   .13      DQin   9.75  11.49   .08

Four games last night: Wins by Orl and OKC helped, as did losses by Cha, Was, and Mia.
The other winners, SAS and Mil, and loss by GSW, were only bad for some.

UPDATE - New month, new leader

Code: Select all

.  avg err   rmse   r^2      .  avg err   rmse   r^2
DRKO   6.99   8.4   .20      WShr   7.83   9.4   .11
Walr   7.17   8.7   .18      dtka   7.90   9.1   .14
25pr   7.28   9.0   .14      Crow   7.94   9.7   .17
knar   7.29   8.5   .15      perW   7.99   9.4   .11
EExp   7.37   8.7   .18      25re   8.11  10.1   .11
Mgoo   7.62   9.1   .16     vegas   8.43   9.9   .20
eWin   7.63   9.2   .12      KPel   8.44   9.8   .16
avgA   7.72   9.0   .16      medi   8.69  10.1   .19
bpmW   7.72   9.0   .16      DQin   9.93  11.7   .07

Last year at this time, there was 1 entry within 1.00 of the leader and 5 within 2.00.
This year it's 10 @<1 and all-but-one within 2. Not counting the dummies.

UPDATE Nov. 2 -- huge improvements overnight

Code: Select all

.  avg err   rmse   r^2      .  avg err   rmse   r^2
DRKO   6.25   7.9   .27      perW   7.10   8.8   .18
knar   6.47   7.8   .23      dtka   7.16   8.5   .21
Walr   6.56   8.1   .25      WShr   7.24   8.6   .20
EExp   6.77   8.1   .25      Crow   7.26   9.2   .24
avgA   6.84   8.3   .25     vegas   7.60   9.2   .29
bpmW   6.84   8.2   .26      25re   7.69   9.4   .17
25pr   6.86   8.4   .22      KPel   7.77   9.1   .25
Mgoo   6.88   8.4   .25      medi   8.16   9.4   .28
eWin   6.92   8.5   .20      DQin   9.36  10.9   .14

The East was cooperative last night: Mil, Bos, Cha, and Was all lost, while Ind and Det won. Only the Det and Bos results hurt anyone.
NOTE: I just noticed knarsu was not included in the avgA (APBR average) error. Now they are.
K Pelton is in avgA and presumably also in ESPNExperts; a dual citizen.

UPDATE Nov. 3

Code: Select all

.  avg err   rmse   r^2      .  avg err   rmse   r^2
DRKO   6.22   7.7   .31      dtka   6.86   8.1   .27
EExp   6.42   7.7   .33      perW   7.12   8.6   .21
Walr   6.47   7.7   .31      WShr   7.16   8.4   .24
knar   6.58   7.7   .26      Crow   7.22   8.7   .31
eWin   6.64   8.3   .24      25re   7.23   9.1   .23
bpmW   6.69   7.9   .31      vegas  7.36   8.6   .37
Mgoo   6.71   8.1   .30      KPel   7.50   8.7   .30
25pr   6.75   8.2   .26      medi   7.99   9.0   .34
avgA   6.76   7.9   .30      DQin   9.35  10.7   .17