2024-25 team win projection contest

[Crow]

A primary reason LEBRON and DARKO projections weren't included is because they were not entered here by authors, checked or found.

If projections were made and published before season, please provide link.

Last season, Darko finished 13th of 20 listings, LeBron 18th.

viewtopic.php?t=10034&start=165

4th and 5th in 2022-23:
viewtopic.php?t=9946&start=165#p39853

Darko 1st in expanded check for 2021-22, LeBron 19th of 30:
viewtopic.php?t=9903&start=105


3+ year averages might be interesting. There is some "randomness" involved I'd say.

'20-21: viewtopic.php?t=9842&start=105

'19-20 (my most recent win):
viewtopic.php?t=9715&start=150

[Crow]

I don't know if any want to use findings in first 60ish games to project the remaining games or re-predict the totals, but utility of in-season metrics for predictions is of interest to me, avoiding some of the issues with aging, team and role changes, minute allocations.

The composites I suggested could be used for such effort.



Value of regular season win totals in predicting playoff outcomes vs SRS or win% against top 10 or top 16 or pre-season projections or current minutes weighted and compiled individual metrics or other criteria would be an interesting topic.

[Crow]

For last 5 years, the APBR average was an average of a bit less than 0.2 better on average error than Vegas line. I was a bit better than Vegas, a little short of APBR average.

There might be others who competed all 5 years or maybe 3 or 4. Use the links to add up your cumulative error totals and find an average if you want. APBR average had 31.37 avg. error over 5 seasons. 6 avg error is usually good but not always achieved.



(Fwiw, I had lower average error than Kevin Pelton in all 4 years he was listed, significantly in 3. I also has fairly wide leads on LEBRON and Darko over last 2 years where they had an average error calculated. Of these, only APBR average was better, thru last season. Being averaged over a broad set is helpful for average performance as long as most of the components are decent.)

[TeemoTeejay]

Pretty sure Darko would be winning and LEBRON is about where Vegas is this year or a tad better

I think the LEBRON or EPM over-unders cooked, altho I guess it’s like a predictive version of both more than just the metrics themselves iirc right?

[Crow]

Either some version of these metrics were published before season and can be found with help of authors or proponents even this late to be fairly scored or they are not available for assesment and I am not going to guesstimate or assume quality in general or given erratic past performance.


Including other prediction sets not officially entered here is a gift. May be interesting. How far to go with it, is up to those willing to find or compute.

In my view, if someone doesn't enter the contest, I am not that interested in their predictions or giving them potential shine here. They can report or not, possibly shine somewhere else. Mike G finds and computes others but if he doesn't find or compute or others don't actively help, then they might not show up. I provided the Vegas info because that is tradition.

[Mike G]

Code: Select all

.  avg err   rmse   r^2      .  avg err   rmse   r^2
TmTj   6.54   8.8   .52      Crow   7.40   9.5   .47
dtka   6.73   8.7   .52      emin   7.52   9.7   .45
avgA   6.84   8.9   .48      WShr   7.59   9.5   .41
KPel   6.95   8.5   .56      24pr   7.82  10.2   .33
medi   7.03   9.9   .44      eWin   8.33  10.5   .28
vegas  7.17   9.4   .46      24py   8.44  11.4   .33
bpmW   7.20   9.0   .47      perW   8.64  10.8   .25
DQin   7.21   9.2   .48      bmgm   8.90  11.8   .33
eExp   7.29   9.5   .53     

[Crow]

Not metric based, but I think BRef current projecrios are

too high for Rockets, Timberwolves, Spurs, Pelicans

too low for Lakers, Blazers.

Not sure if that would help or hurt me. Just fwiw.

[Mike G]

I wonder how much work it would be to inject recency into a team 'strength' measure.
The Mavs were pretty good when they had 2 great centers. With none, not so great; and no Luka.
Spurs are 1-5 since Wemby was shelved. They're still projected to go 10-13 the rest of the way.

Lakers are 12-2 since Jan. 27, Blazers 14-6 since Jan. 13.
Giving equal weight to games before these dates, is not what I would call exceptional analysis; it's just easier.

[Crow]

Time decay team strength calculation, similar to with players. And rest of season SOS layered on too.

In the end BRef or any other estimate doesn't matter.

Pretty close to the end. Will find out.

[bruddog]

Always a fair amount of randomness in predicting team wins.

One of the biggest sources of noise being trades and injuries

Some from this season.
1. Pelicans catastrophically injured.
2. Wemby out for remainder of season
3. Luka trade
Etc.

Predicting the pelicans wins this year provided almost zero signal into how good your preseason prediction was.

[Mike G]

In the last month, the Knicks are 8-3 with avg scoring margin of -3.54. And 3-0 in overtimes.
Are they progressing in a positive or negative direction?

I've taken W-L and MOV in half-month intervals for each team and weighted them progressively heavier toward the present.
October numbers are weighted x1, first half of Nov x2, etc. The latest, 2nd half of Feb (with a couple March dates) are x9.
https://www.basketball-reference.com/fr ... &lg_id=NBA

Those ^ pages show team W-L, Pythagorean W-L and MOV thru the given date.
I've subtracted the previous reading (from a half month before) to get the latest half month, for these numbers in just that interval, and weighted and summed these 9 intervals to date. They are then divided by (1+2+...+9) 54 to get a recency-weighted number.

As the Knicks show, wins are not always related to MOV (or SRS etc). The correlation to MOV is .79, and with Pyth wins it is .93.
Possibly a team is just better in close games, or they may just be lucky.
Here's the summary of the differences between current and recency-adjusted MOV, Risers on the left, fallers on the right.

Code: Select all

up      W%    pW%   mov      down     W%     pW%    mov
Det   .086   .115   2.8      NYK    .019   -.087   -2.4
Was   .031   .054   2.7      Mem   -.049   -.040   -2.0
Por   .020   .059   1.8      Dal   -.021   -.024   -1.8
LAL   .039   .055   1.8      Hou   -.037   -.017   -1.8
Ind   .043   .045   1.7      Cha   -.017   -.049   -1.7
Den   .033   .036   1.5      Orl   -.046   -.053   -1.6
Sac   .063   .081   1.5      Phx   -.069   -.040   -1.4
NOP   .044   .038   1.4      SAS   -.041   -.018   -1.0
Uta   .030   .029   1.0      LAC   -.048   -.034    -.9
GSW   .005   .066   1.0      Bos   -.015   -.008    -.7
Mil   .045   .025    .7      Mia   -.023   -.011    -.6
Cle   .001   .022    .5      Brk    .005   -.006    -.4
Atl  -.017   .023    .3      Tor    .015   -.011    -.3
OKC   .009   .038    .2      Phl   -.014    .000    -.3
Min  -.007  -.003    .1      Chi   -.024   -.005     .0

Here teams are in their conference, ranked by recency-weighted MOV.
Win% and MOV (without the R) are season-total as of yesterday.

Code: Select all

tm      W%   W%R    mov   movR      tm     W%    W%R    mov   movR
OKC   .817  .826   12.5   12.7      Cle   .833  .834   11.5   12.0
Den   .639  .672    5.1    6.6      Bos   .705  .690    8.6    7.9
Mem   .633  .584    6.3    4.3      Det   .557  .643    1.3    4.1
LAL   .644  .683    1.6    3.4      Ind   .576  .619    1.3    3.0
Min   .532  .525    3.1    3.2      Mil   .576  .621    1.9    2.6

Sac   .525  .588    1.5    3.0      NYK   .667  .686    4.7    2.3
GSW   .533  .538    1.7    2.7      Mia   .475  .452    -.5   -1.1
Hou   .617  .580    3.8    2.0      Atl   .450  .433   -2.8   -2.5
LAC   .533  .485    2.0    1.1      Orl   .468  .422   -1.7   -3.3
Dal   .525  .504    1.5    -.3      Chi   .393  .369   -3.9   -3.9

Por   .443  .463   -3.9   -2.1      Phl   .356  .342   -4.3   -4.6
SAS   .424  .383   -2.2   -3.2      Tor   .311  .326   -5.9   -6.2
Phx   .459  .390   -2.1   -3.5      Brk   .350  .355   -6.5   -6.9
Uta   .250  .280   -6.9   -5.9      Cha   .237  .220   -7.5   -9.2
NOP   .279  .323   -7.7   -6.3      Was   .186  .217  -12.2   -9.5

Pistons now appear to be #3 in the East (and would be roughly tied for 3rd in the West!) due to recent performance. B-R.com expects them to go 11-10 the rest of the way and finish just behind Mil and Ind for the #6 spot. That would very likely place them in a 1st round vs the #3 Knicks; and win in 6.

[Crow]

40% see about 4% win% change or more, at this look.

4% isn't that much but it could make minor changes to contest results.

It could change several times, maybe back to previous trends. Change at the end, to be determined.

[Mike G]

I think BRef current projecrios are

too high for Rockets, Timberwolves, Spurs, Pelicans

too low for Lakers, Blazers.

Por and LAL are #3 and 4 in the 'risers' list.
Hou and SAS are among the down-turners; Min is neutral.

Pels look better of late, and don't seem be given any b-r.com 'future regression' phenomenon. Current Win% is .281, and their 6-15 projected future is .286.
Their recency-adjusted MOV is -6.3 (season -7.7), which is equivalent to a .337 Win%. Subject to resting/tanking strategies, they might now be expected to win 7 of 21.

Blazers over the last 1.5 months are +4.1 mov, and that would be ~5th in the West over the whole season.
Over the last half-month, they are +11.7! Making the play-in would require utter collapse from Sac and Dal, who are missing their best players; and no surge by Phx or SAS. All are possible.
B-r.com gives them 1.3% chance of it.

[Crow]

Change in win % recently is one thing, the BRef win% projections for rest of season are quite another, far more dramatic and probably often unrealistic.

[Mike G]

Often I wonder about what exponent/root applied to midseason errors signify, regarding their prediction value for the final results. So last year's contest records are examined, and for each posting, everyone's MAE, RMSE, and r-squared are correlated to the final (Apr. 15) MAE and RMSE values.
Thru 7-8 games, the standings were quite different from how we would finish the season (negative correlations). Yet r^2 had the end results mostly right.

Code: Select all

mean absolute err. '23-24   root mean sq err
mae   rmse    r^2    date   mae   rmse   r^2
-.61  -.52    .37   11.02   -.56  -.46   .41
-.20  -.21    .23   11.09   -.18  -.17   .24
.43    .36    .65   11.16    .43   .38   .66
.53    .68    .82   11.23    .53   .71   .83
.63    .75    .83   12.03    .63   .77   .85

.94    .86    .85   12.18    .94   .87   .86
.98    .92    .89   12.31    .97   .93   .90

By mid-Dec, MAE was best at predicting final MAE; and also better at final RMSE.

Then Jan-Feb had RMSE as the better predictor at both exponents:

Code: Select all

mean absolute err. '23-24   root mean sq err
mae   rmse    r^2    date   mae   rmse   r^2
.91    .92    .88    1.14    .92   .93   .89
.92    .96    .91    2.04    .92   .96   .93
.96    .97    .91    2.15    .95   .97   .93

.982   .981   .92    3.11    .98   .99   .94
.993   .989   .94    4.02    .99   .998  .96
1.00   .991   .95    4.15    .99   1.00  .97

Down the stretch (Mar/Apr) they predicted themselves best.

Today's standings, not much change but to be worse and worst.

Code: Select all

.  avg err   rmse   r^2      .  avg err   rmse   r^2
TmTj   6.57   8.9   .51      Crow   7.44   9.6   .47
dtka   6.73   8.8   .51      eExp   7.46   9.7   .53
avgA   6.82   9.0   .48      emin   7.49   9.7   .45
KPel   7.03   8.6   .56      WShr   7.71   9.6   .41
medi   7.08  10.0   .43      24pr   7.93  10.3   .32
bpmW   7.19   9.1   .47      eWin   8.40  10.7   .27
DQin   7.25   9.3   .48      24py   8.49  11.5   .32
vegas  7.29   9.6   .45      perW   8.63  10.9   .24
.                            bmgm   8.95  11.9   .33

Projections from b-r.com relative to the avg of all our predictions:

Code: Select all

over   tm   proj   avgA      over    tm   proj   avgA
19.4  Cle   67.1   47.7     -22.7   NOP   22.7   45.4
15.9  Det   45.4   29.5     -19.0   Phl   29.6   48.6
10.4  OKC   66.3   55.9     -11.5   Phx   35.9   47.4
9.8   Por   35.5   25.7      -9.4   Cha   20.1   29.5
9.3   LAL   50.1   40.8      -8.2   Uta   21.5   29.7
8.1   Mem   50.9   42.8      -6.8   Dal   43.0   49.8
7.9   Hou   49.9   42.0      -6.8   Was   16.1   22.9
6.0   LAC   43.2   37.2      -5.3   Orl   37.5   42.8
5.6   NYK   52.4   46.8      -4.6   Min   46.7   51.3
2.6   Ind   47.0   44.4      -3.7   Sac   43.1   46.8
2.1   Den   51.4   49.3      -2.6   Mia   40.2   42.8
1.8   SAS   35.5   33.7      -0.9   Atl   37.9   38.8
1.2   Bos   58.9   57.7      -0.5   GSW   45.6   46.1
1.2   Chi   32.0   30.8               
1.0   Brk   28.0   27.0               
0.4   Mil   46.3   45.9               
0.2   Tor   30.2   30.0