2023-24 team win projection contest

[DarkStar48]

Projections via TeamRankings.

Link: https://www.teamrankings.com/nba/projections/standings/

[DarkStar48]

numberFire win predictions.

Link: https://www.numberfire.com/nba/teams/power-rankings

[DQuinn1575]

okay, my revised entry,

BOS 56.3
PHI 54.0
LAL 51.7
CLE 51.1
MIL 50.7
NOP 48.2
DEN 48.2
PHO 48.2
ATL 45.6
LAC 45.3
NYK 44.5
MEM 44.5
MIA 44.5
CHI 41.7
TOR 41.5
DAL 41.5
MIN 41.3
UTA 41.2
ORL 39.5
SAC 38.8
BRK 37.6
GSW 37.2
OKC 34.9
WAS 33.6
IND 31.7
CHO 30.4
POR 28.9
SAS 27.3
DET 25.3
HOU 24.8

[Mike G]

UTAH should be 40. I changed in my above post from 37 to 40.

(v-zero) I thought I would just double check that my changes since originally posting my numbers have found their way to your spreadsheet.

Got it. Got it.

(DQuinn1575) okay, my revised entry,..

Got it.

All now seem to predict 1230 wins.

[tarrazu]

FWIW Team Rankings looks like they took a "Vegas line" and removed the juice in a way to get to 1230 totals win. Should be a good benchmark entry.

[Mike G]

That was the case last year, too. I considered throwing them out, but left them just to see. And sure enough they tracked my "vegas" entry to the end.
I'm inclined to remain with one called "vegas" since it is the one to beat in the most universal (or superficial) sense. When I find one that seems to fit it (or one is submitted), I have always lopped off the profit margin to make them add to 1230. TeamRankings seemed to do it slightly differently.

Last year there were some differences in the two:

Code: Select all

tm   vegas   TmRk      tm   vegas   TmRk
Den   52.1   50.5      Cha   33.1   35.4
Por   40.1   38.8      Ind   22.1   23.7
Phl   51.1   50.1      Uta   23.1   24.3
GSW   52.1   51.1      Mem   48.1   49.1
Atl   46.1   45.3      LAC   51.1   52.0
Tor   47.1   46.4      Min   48.1   48.6
Bos   53.1   52.5      Det   29.1   29.5
OKC   24.1   23.6      Mil   52.1   52.5

These would only cause separation in the 'standings' when a team is between the two projections.

[DarkStar48]

They have a projection for various Harden scenarios, but I think the Harden plays for PHI case should be the one used in the contest.

With the latest Harden news, it was smart for any projections that accounted for the possibility of him not playing at all, or eventually getting traded. (Betting markets do this indirectly.) May provide a non-trivial systematic advantage in the contest.

[Mike G]

Even if a player is traded, one can assume he's traded for equal value -- in proficiency * minutes -- so it's a wash in the net wins provided.
In the case of Harden, it seems likely his value, by any measure, is not going to be met by another team's offer.

I threw him 600 minutes rather than zero, and didn't go optimistic on anyone else: 2200 for Embiid, after 2284 got him MVP.
So my prediction of 38 wins for Phl is 10 below the avg of the other 7 here.

[nbacouchside]

Even if a player is traded, one can assume he's traded for equal value -- in proficiency * minutes -- so it's a wash in the net wins provided.
In the case of Harden, it seems likely his value, by any measure, is not going to be met by another team's offer.

Why would one assume this? There's lots of trades of present value for future value. It is often not equal value for a single given year.

[Mike G]

It's a default assumption, in lieu of being able to see the future.
At the beginning of the season, most or all teams are hoping to make the playoffs. Closer to the trading deadline, some teams will be tanking for the future.

This year, b-r.com is doing their projections from day 1. I've put them into the spreadsheet and find they sum to just 40 wins on avg. Exactly. So adding 1 W per team, our contest looks like this:

Code: Select all

.  avg err  rmse
eWin   3.8   5.0
trzu   4.7   6.3
ncs.   5.4   7.0
dtka   5.4   6.8
Crow   5.7   7.0
vzro   6.3   8.3
emin   6.4   8.9
DQin   6.6   8.0

With a lot of regression-to-mean, teams project thus:

Code: Select all

best   W       worst   W
Ind   50.9      Hou   30.2
Den   47.0      Uta   34.5
OKC   46.5      SAS   34.8
Phx   46.3      Por   35.1
NOP   45.5      Atl   35.5
Cle   44.0      Chi   36.3
GSW   43.9      Cha   37.8
LAL   43.8      Mia   38.0
Phl   43.4      LAC   38.1
Was   43.1      NYK   39.2
Mil   43.0      Brk   39.7
Bos   42.7      Tor   39.9
Det   42.6      Mem   41.2
Min   42.4      Dal   41.3
Sac   41.6      Orl   41.4

Probably the only time both my team and I are on top here.

[Crow]

Thanks again for scorekeeping.

We'll be after you. I probably won't make too much of the standings until a notable portion of the season is in the books and the regression is lessened.

[DarkStar48]

@Mike G, just a request to include r^2 determination coefficient in the contest analysis.

[Mike G]

Code: Select all

.   avg err   rmse   r^2
eWin   4.55   5.81   .15
trzu   4.69   6.34   .29
Crow   5.01   6.67   .26
dtka   5.44   7.03   .19
ncs.   5.63   7.03   .17
DQin   5.88   7.61   .22
vzro   6.07   7.86   .22
emin   6.09   8.86   .19

When I de-regress this a bit, trzu takes the lead.
Quite a bit more, and Crow takes 2nd.
But everyone's errors increase substantially with this.

[tarrazu]

Because they essentially start with the Vegas lines, these projections do a nice job with limited games played:
https://www.teamrankings.com/nba/projections/standings/

[Mike G]

Almost every East team looks worse, every West team better overnight. West 3-1 vs East.
Averages are now 43 Wins in the West and 39 in the East.
If b-r.com is just having issues, this post may be deleted. Meanwhile here's the contest with APBR on the left and others on the right.

Code: Select all

.   avg err   rmse   r^2      .   avg err   rmse   r^2
eWin   5.06   6.02   .13     vegas   5.57   6.97   .29
trzu   5.95   6.94   .20      EExp   5.65   7.31   .31
dtka   6.11   7.53   .13      EBPI   5.77   6.93   .19
Crow   6.15   7.30   .18      KPel   5.89   7.13   .10
ncs.   6.33   7.31   .14      DRKO   6.06   7.05   .14
emin   6.95   8.70   .22      ChKl   6.08   7.62   .27
DQin   7.08   8.13   .16      IanL   6.11   7.92   .32
medi   7.09   8.49   .21      LEBR   6.21   6.95   .10
vzro   7.15   8.52   .13      NuFi   6.60   8.73   .20

Vegas is an amalgam of the 3 submitted which actually predicted all 30 teams.