2020-21 team win projection contest

[DSMok1]

Using the values from BBall-Index's average of PIPM, RPM, BPM, and RAPTOR, plus the general projection approach I recently detailed in my Box Plus/Minus thread. I used Kevin Pelton's minutes.

I also did a few manual adjustments on a few key players.

Eastern Conference:
MIL 48.3
BKN 45.8
MIA 43.9
TOR 42.9
PHI 41.9
BOS 40.1
IND 37.7
ATL 37.2
WAS 31.9
ORL 28.5
CHI 27.4
CHA 24.2
NYK 23.0
DET 20.4
CLE 19.8

Western Conference:
LAL 49.6
UTA 47.0
LAC 46.2
DAL 45.8
DEN 41.8
HOU 41.5
POR 39.0
NOP 38.3
MIN 35.1
MEM 34.3
PHX 34.3
SAC 32.8
GSW 29.9
SAS 29.4
OKC 23.6

EDIT: I have made some updates--my revised projections are at: viewtopic.php?p=37518#p37518

[vzografos]

I used Kevin Pelton's minutes.

I saw those too.
Not sure how accurate they are but if you want some additional projections I am going to calculate all the (per-season average) boxscore stats for 2020-21. I expect to have them in the next couple of days. Dont know if they will be any good but I would be interested to see how your team win projections change (if at all).

[Crow]

One huge difference between my first take and DSMok1's set. Then mild to modest differences on a group. Most quite close. 1, 2 actual to projected win differences on teams here and there will add up as decisive for somebody.

[eminence]

Bummer for Hayward, taken too many injuries recently, but probably at least a few tweaks to make now (will make in my original post). Wound up being MIL +1, NYK +1, MIN +1, CHA -3 for me.

Though they're only listing him day to day, maybe I'll have to change back, who knows.

[vzografos]

May I suggest a "Market efficiency" hypothesis?
To whoever is keeping tabs, when you are tracking the contest winner, can you also generate the "market consensus" team win projections by taking the average of all the contest entrants? I am curious to see how this will perform relative to the individual projections.

[eminence]

The folks who usually track have tracked such an average in the past, and it normally seems to do quite well, I consider anything over the average a resounding success for myself if I'm so lucky to hit at that level.

[mtpowell22]

I've just renewed the Win Totals league - https://www.wintotals.com/leagues/b6ca8a6d/ - if we want to use that again this year. Crow won (pretty handily) in the shortened 19-20 season and I totally blanked on posting the results here.

You can't save off decimals but otherwise it should work well.

[vzografos]

The folks who usually track have tracked such an average in the past, and it normally seems to do quite well, I consider anything over the average a resounding success for myself if I'm so lucky to hit at that level.

I think ensemble clustering methods (Which is what averaging is) usually tend to do quite well (assuming no wild outliers).
This gives the illusion of "market efficiency" or that the market knows best. But still, it would be a good baseline to track

[eminence]

I'm going to be annoyed if Harden is traded like a week into the season, lol.

[vzografos]

I used Kevin Pelton's minutes.

I saw those too.
Not sure how accurate they are but if you want some additional projections I am going to calculate all the (per-season average) boxscore stats for 2020-21. I expect to have them in the next couple of days. Dont know if they will be any good but I would be interested to see how your team win projections change (if at all).

If anyone is interested you can find my minutes projections here:
https://drive.google.com/file/d/1juLaPO ... sp=sharing
as an alternative to Kevin Pelton's numbers.

These projections do not incorporate any player injuries before the season, or any changes to player status (i.e. left the team, league, retired etc). Any players that were given 0 minutes from Pelton's method have been highlighted because maybe there is additional information that I dont have so you need to pay attention to those.

[Mike G]

...average of PIPM, RPM, BPM, and RAPTOR, plus the general projection approach I recently detailed in my Box Plus/Minus thread. I used Kevin Pelton's minutes.

I also did a few manual adjustments on a few key players.

Is anyone using ONE stat to predict? As though there might be value in finding which stats work better? At one time, it seems that was kinda sorta the reason for this contest.

Then there are rookie projections (no stats for them); and expected minutes that we would [for a while try to] apply to all.
Where are Pelton's minutes to be found?

[vzografos]

...average of PIPM, RPM, BPM, and RAPTOR, plus the general projection approach I recently detailed in my Box Plus/Minus thread. I used Kevin Pelton's minutes.

I also did a few manual adjustments on a few key players.

Is anyone using ONE stat to predict? As though there might be value in finding which stats work better? At one time, it seems that was kinda sorta the reason for this contest.

There is a new thread about this....
I would definitely be very interested in one-dimensional predictor performance

Where are Pelton's minutes to be found?

Just download my Excel spredsheet. It contains his minutes as well. My data is column G so everything else is his

https://drive.google.com/file/d/1juLaPO ... sp=sharing

[RyanRiot]

Anyone else normally use a simulated season for your projections? That's what I did last year, but I'm not sure if there's any way that makes any sense to do that this season when we only have half the schedule released.

[vzografos]

Anyone else normally use a simulated season for your projections? That's what I did last year, but I'm not sure if there's any way that makes any sense to do that this season when we only have half the schedule released.

Of course. I think that's the only way. And make sure you simulate multiple times using Monte Carlo.
You just need to figure out the rest of the schedule using NBA's rules.

[eminence]

I've never bothered with a simulated season myself, but more power to y'all that do.