2018-19 team win projection contest

[kmedved]

Very early rankings

https://docs.google.com/spreadsheets/d/ ... IwFv_wPes/

Can you add my projections to the contest? Posted here: https://docs.google.com/spreadsheets/d/ ... IwFv_wPes/ (and google docs for easy/copy paste here: https://docs.google.com/spreadsheets/d/ ... edit#gid=0).

Thanks.

[Crow]

When were these projections created? Any documentation of that date? Reason for posting 2 weeks late?

I'd consider based on these answers.

[tarrazu]

.

[tarrazu]

When were these projections created? Any documentation of that date? Reason for posting 2 weeks late?

I'd consider based on these answers.

has Houston at 58, so it checks out!

but this posted/timestamped here, so should be fine to add unless someone has a reasonable objection: https://twitter.com/kmedved/status/1052355167574745088

[eminence]

Seems good to me, welcome

[Jinxed]

Seems I'm off to a hot start. Let's see how long I can hold pole position.

[Crow]

Ok, thanks tarrazu for the publish link.

[kmedved]

When were these projections created? Any documentation of that date? Reason for posting 2 weeks late?

I'd consider based on these answers.

Sorry, I meant to link to the twitter post, which had the timestamp (technically a few minutes after the Celtics/Sixers tipoff). Thanks to tarrazu for clearing up. Appreciate it.

[Mike G]

Well, I have been traveling and not thinking about this much at all; so this year I'll just be an observer.
B-R.com has their Playoff Probabilities Report going now:
https://www.basketball-reference.com/fr ... _prob.html
and relative to their projections, ranked by average 'error':

Code: Select all

avg err       rmse       avg err      rmse
KPel   5.90   7.67      gold   7.19   8.92
AnJo   6.15   8.10      Crow   7.21   9.62
shad   6.40   8.26      J-TF   7.29  10.11
trzu   6.57   7.97      538_   7.51   9.51
cali   6.60   8.72      sndi   7.55   9.40
RyRi   6.68   8.70      Cica   8.15  10.74
emin   6.77   8.79      kmed   8.64  10.66
bbst   6.93   8.42      Rd11   9.45  11.08
Jinx   7.03   9.02            

The Bucks have a 44% chance to win it all!

[mtamada]

The Bucks have a 44% chance to win it all!

Even better, 7 games into the season and they have a 100% chance of making the playoffs!

Those were calculated by Kubatko, right? He does good stuff, but most playoff probability prediction systems fail to regress to the mean enough, and that one does so to an absurd degree. Injuries, such as the concussion that Giannnis recently suffered, can happen at any time to any one.

[Jinxed]

Is the winner of this contest to be judged by MAE or RMSE?

[eminence]

Is the winner of this contest to be judged by MAE or RMSE?

I prefer MAE, but not a big deal either way.

[Crow]

It may still be K's methodology (and may still have equity in company) but he is no longer on staff there. With statmuse.

It became a thing to look at winner via different lenses. MAE is my preference but if people want to look at other things, fine, whatever.

[sndesai1]

i have no pref. let's just say whichever has the most support by sunday is the primary metric for the contest

it's currently MAE-2, RMSE-0



bk-ref projections are ok for regular season wins after a couple months, but there's no informed prior. all that it really seems to use is a team's current srs regressed towards 0 and the remaining schedule, which leads to odd projections early on.

when it comes to the playoff winner projections, there's very little point in continuing to use it. i mean, it's 10+ years old now, so not too surprising that the value is limited

[Mike G]

bk-ref projections are ok for regular season wins after a couple months, but there's no informed prior. all that it really seems to use is a team's current srs regressed towards 0 and the remaining schedule, which leads to odd projections early on.

At what point in the season is current SRS + remaining schedule actually a better predictor than what you thought before the season started? Could we be there already?

I stole your spreadsheet. Thanks! Is Pinnacle what we could call Vegas?