2022-23 team win projection contest

[DarkStar48]

For Sportsbook comparison, here are the DraftKings predictions as of Mon, Oct. 17 at 11:45 a.m. ET.

Link: https://www.vegasinsider.com/nba/odds/win-totals/

Boston Celtics 53.5
Milwaukee Bucks 52.5
Phoenix Suns 52.5
Los Angeles Clippers 52.5
Golden State Warriors 51.5
Philadelphia 76ers 50.5
Brooklyn Nets 50.5
Denver Nuggets 49.5
Minnesota Timberwolves 49.5
Memphis Grizzlies 48.5
Dallas Mavericks 48.5
Miami Heat 48.5
Cleveland Cavaliers 47.5
Atlanta Hawks 45.5
Los Angeles Lakers 45.5
Toronto Raptors 45.5
New Orleans Pelicans 44.5
Chicago Bulls 41.5
Portland Trail Blazers 39.5
New York Knicks 38.5
Washington Wizards 35.5
Sacramento Kings 33.5
Charlotte Hornets 33.5
Detroit Pistons 29.5
Orlando Magic 26.5
Utah Jazz 24.5
Oklahoma City Thunder 23.5
Indiana Pacers 23.5
Houston Rockets 23.5
San Antonio Spurs 22.5

[DarkStar48]

Projections via TeamRankings.

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

[DarkStar48]

DARKO model forecast by Kostya Medvedovsky.

Link: https://docs.google.com/spreadsheets/d/ ... 1290726872

[DarkStar48]

numberFire win predictions.

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

[Crow]

More media references than individual entries.

Two divisions to track.

[DarkStar48]

Team record forecast by Mike Prada.

Link: https://twitter.com/mikepradanba/status ... 99232?s=21

[Mike G]

https://fivethirtyeight.com/features/wh ... -disagree/

[DarkStar48]

BBIPM (yet unreleased) and LEBRON win estimates. (Values are preseason despite Oct. 20 date.)

Link: https://www.bball-index.com/season-team-projections/

[v-zero]

BBIPM (yet unreleased) and LEBRON win estimates. (Values are preseason despite Oct. 20 date.)

Link: https://www.bball-index.com/season-team-projections/

Those BBIPM predictions are so extreme. I'll be quite surprised if they perform strongly.

Out of interest is everybody else using minutes projections and individual player value metrics to make their predictions?

[Crow]

I impression is that most do, though there typically isn't a lot if chatter about exact methods.

[v-zero]

A shame, I think, that lack of discourse. I'd be genuinely interested to hear what methods and models people are using, and how they arrived at them, and willing to share in kind.

[eminence]

I use team level data (just wins/and MOV really) combined with a subjective* discrete variable that I assign based on which 'direction' I think the team is moving (getting worse, improving, staying the same). Really it only has the 3 inputs, I've played around with adding an 'in the playoffs' vs 'out of the playoffs' variable to pull projections up or down, but the addition of the play-in has thrown me off on that, so not this season.

Basically my projections are a lot more gut and a lot less good/sound model.

*=I pull it out of my ass, though obviously I've watched a lot of basketball and look at a lot of the individual stats quite regularly

[Crow]

My approach has usually involved weighted blending known and private metrics and adjusting for coaching quality, "goals", new players, aging and perceived changes in fit.

I have had different levels of thoroughness / complexity thru the years. I don't spend as much time recently as I have in past years.

[Mike G]

Relative to projections shown here:
https://www.basketball-reference.com/fr ... _prob.html

Mean (average) Absolute Errors, and Root Mean Squared Errors:

Code: Select all

.  avg err   rmse        avg err  rmse
ncs.   6.4   7.7      TmRk   7.8   9.5
vzro   6.5   7.3      vegas  7.9   9.7
538E   6.8   8.8      EExp   8.0   9.7
drko   6.9   8.2      veg2   8.0   9.7
EBPI   7.1   8.1      emin   8.0   9.3
dtka   7.3   8.1      nuFi   8.1   9.7
538R   7.5   8.8      MPra   8.9  10.9
trzu   7.5   8.8      LEBR   9.0  10.5
Crow   7.6   9.0            

Does not include last night's games. Later I may try to "de-regress" the b-r.com projections.
Without regressing every team toward 41 wins, it looks like East teams are on pace to avg 43.6 wins vs 38.4 out West.

[Mike G]

A lot of games last night but just 2 between East and West: Mem 130-99 Cha, and Dal 111-110 Tor.
Dal-Tor was about as expected, so just one lopsided game shifted the whole East/West imbalance; with rSOS (remaining-games strength of schedule) a factor, b-r.com now estimates the conferences are equal.

I added a couple of dummy "contestants": Last season's wins (2022) and Pythagorean-expected wins (22Py). Yes, it's possible to do worse than these.

Code: Select all

.   avg err   rmse          avg err   rmse
ncs.   6.23   7.82      TmRk   7.63   9.44
drko   6.68   8.18      vegas  7.63   9.63
vzro   6.75   7.79      emin   7.70   9.20
538E   6.78   9.00      veg2   7.70   9.57
dtka   6.94   8.13      nuFi   7.93   9.60
trzu   7.12   8.71      EExp   8.02   9.84
Crow   7.17   8.70      22Py   8.07  10.11
EBPI   7.19   8.27      MPra   8.51  10.64
538R   7.30   8.60      LEBR   8.76  10.09
2022   7.38   9.32         

vegas was submitted by DarkStar atop page 2; vegas2 appears as 'Caesars' along with the ESPN experts' predictions.