2020-21 team win projection contest

[Crow]

ATL 36
BOS 41
BRK 47
CHA 25
CHI 24
CLE 21
DAL 42
DEN 46
DET 21
GSW 35
HOU 40
IND 37
LAC 44
LAL 51
MEM 40
MIA 47
MIL 49
MIN 24
NOP 41
NYK 25
OKC 25
ORL 29
PHI 42
PHX 35
POR 36
SAC 28
SAS 31
TOR 41
UTA 45
WAS 32

[DSMok1]

Here were some of the more official system projections:

Kevin Pelton's RPM Projections:
https://www.espn.com/basketball/story/_ ... -nba-teams

ESPN:
ESPN Forecast (Expert Panel)
BPI
FiveThirtyEight
Also includes Vegas odds.
https://www.espn.com/nba/story/_/id/305 ... ning-night

[Crow]

Imo, Kevin Pelton's projections look overly regressed to the mean. No team projected over 44.5? Only 4 with 25 wins and under? It is a quite different take than mine.

44.5 wins is 61.8% win. Last season 6 teams beat that mark and 4 were very close (over 61%).

I have 6 at 45 and over wins then 5 more with at least 41 wins (most of those with just slightly less than 60% win). Pelton has 5 total at or over 41 wins.

I expect a few more at or under 25 wins than he does.


I didn't check everyone listed here but everyone I did check has a number with 41 plus wins near mine and way above Pelton's.

41 wins this season is about 57% win. There is no season in last 11 years that did not actually have 10 or more teams at or over 57% win (or at least double Pelton's 5). I stopped looking after that.

[DSMok1]

Conversely, BPI's projections look under regressed--much wider spread than any others I've seen.

As a reminder--just because the spread is narrow doesn't mean we don't think there will be elite win totals. It just means that the mean projection for each team is down in that narrower range. We don't know which teams will end up with the elite win totals.

[Crow]

"We don't know which teams will end up with the elite win totals."

Yes. But the point of this contest is to make the best point estimates.

The article offers "average wins". Either that is being offered as the best guess of projected wins or it is not definitively answering the question of projected wins. The headline phrases it as providing win projections.

[tarrazu]

Conversely, BPI's projections look under regressed--much wider spread than any others I've seen.

I'll predict BPI is among the bottom 10% entries.

[Crow]

Using win projections at
https://projects.fivethirtyeight.com/20 ... edictions/

I calculate my average error right now at 3.33. Better in East (under 3) than West (almost 4). Worst 4 are off by 6-7 wins. Within 0-1 on 7.

[Crow]

ESPN expert panel at 2.87.

Kevin Pelton's RPM, 4.49.

[Crow]

If a table is built, it may be wise to make the projection baseline a blend of 538, ESPN and B-Ref.

Til then, calculate and publish your average error as you wish.

(The Vegas line will need to be checked to see if wins projected equal wins available. Pro-rated if necessary.)

[liminal_space]

i'll try to create a spreadsheet with everyone's entries today

[eminence]

That'd be much appreciated, thanks

[Crow]

I'm worse now relative to 538 Elo. (Think I used their player based projections last time and not sure if that gets updated.)

[vzografos]

I'm worse now relative to 538 Elo. (Think I used their player based projections last time and not sure if that gets updated.)

Actually dont compare (or use) the 538 elo or carmelo predictions. They are pretty poor, consistently predicting below the market. I have been tracking them for the last few years.
Their new Raptor stuff (where do the come up with those names...) is somewhat better so far but we will see how it performs by the end of the season, since it is the first time they apply it.

[Crow]

https://www.espn.com/nba/story/_/page/B ... ayoff-odds

I am dubious of some of these projections. Toronto, Brooklyn, Houston, OKC, Utah...

[vzografos]

https://www.espn.com/nba/story/_/page/B ... ayoff-odds

I am dubious of some of these projections. Toronto, Brooklyn, Houston, OKC, Utah...

I am not ESPN predictions are even worse than 538's

I d like to paste a table of expert predictors from the last few years so you can see where ESPN and 538s and others are placed, but this site does not seem to allow html in the message? (and I cannot be bothered with Tableau).

Any other options for displaying a table apart from sending an image over?