2024-25 team win projection contest

[Crow]

Planning to score your projections when final results are in? Post here?

[TeemoTeejay]

What’s the difference between the LEBRON and EPM projections and the advanced model and Vegas advanced model ones?

[kmedved]

What’s the difference between the LEBRON and EPM projections and the advanced model and Vegas advanced model ones?

Sorry about that - those aren't mine. My projections are on the "DARKO wins" tab.

@Knarsu3 asked me to add his projections to the spreadsheet as well for scoring purposes, but I don't know the details of what's in anything on the 'Krishna Stuff' tab.

[kmedved]

Planning to score your projections when final results are in? Post here?

Sure, I'll post them when finished. The spreadsheet has live scoring (pulling from BBR for now) as well, although to reiterate, I agree with the concept that projections that weren't entered here shouldn't "count" (for a bunch of reasons already discussed).

[Mike G]

...I think my projections may have come in a few hours late this year. I promise I didn't use any "in-sample" data, but I would have missed the deadline this year regardless.

'A few hours' is not a problem, I just think a precedent of entering well into the season is inviting turmoil in future.
Having said that, we did late-enter Kevin Pelton and some "vegas" entries, as they have been very traditional; also serving as "standards" to beat.

Looking back I see a kmed entry for 4 seasons, 2016-19 (and possibly earlier) -- tied for 1st, near the middle, 4th best, and 4th worst. Is DARKO directly descended from whatever you used back then? or better at predicting?

[kmedved]

It's the same system along the way, with some incremental improvements here and there. The big thing I've done the last few years is incorporating more preseason data, but there's no other major changes recently. I forget when exactly I launched DARKO. Before DARKO, I was just using RAPM with an aging curve.

[Crow]

If I had regressed every team 1 game to the mean, I calculate I would have been 3 spots higher. 2 games and probably would have been in top group. 3 wouldnt quite make 1st but real close. Yeah, I "should" have regressed. 4 would have been helpful still but not as good as 3. Maybe slightly different than 3 could have been best, this time.

[Mike G]

Code: Select all

.  avg err    rmse   r^2       .  avg err   rmse   r^2
TmTj   7.48    9.9   .45      24pr   8.39   10.8   .33
dtka   7.63    9.7   .46      WShr   8.41   10.2   .39
avgA   7.75   10.0   .41      vegas  8.46   10.6   .39
medi   7.82   11.0   .37      emin   8.49   10.8   .37
DQin   7.86   10.3   .42      24py   8.66   11.7   .33
bpmW   8.04    9.9   .43      bmgm   9.24   12.1   .33
Crow   8.30   10.5   .41      eWin   9.61   11.8   .20
eExp   8.31   10.6   .43      perW   9.86   11.9   .18
KPel   8.35    9.9   .57               

UPDATE Mar. 12

Code: Select all

.   avg err   rmse   r^2       .  avg err   rmse   r^2
TmTj   7.47    9.9   .45      24pr   8.40   10.8   .34
dtka   7.51    9.7   .47      KPel   8.40    9.9   .58
avgA   7.72   10.1   .42      WShr   8.42   10.2   .40
medi   7.76   11.0   .37      vegas  8.42   10.6   .39
DQin   7.76   10.2   .43      emin   8.48   10.9   .38
bpmW   8.09    9.9   .43      24py   8.62   11.6   .34
Crow   8.27   10.5   .41      bmgm   9.10   12.0   .34
eExp   8.37   10.6   .43      eWin   9.63   11.8   .20
.                             perW   9.91   12.0   .18

[Crow]

Past ESPN experts in additional to Pelton and Vegas. Best I got this time.

Can regress too little or too much.


Congratulations to the soon to be contest winner by average error and the various other top performers by other measures and group status.

3 measures, looks like good chance for 3 different bests. Add unlisted but notable names with history here and there will be one or more other bests. Respect to still others close to the top. Pretty well done by most. Even 1 off Vegas is decent. It is tough to predict this with so many factors.

Thanks to Mike G for recordkeeping.

[TeemoTeejay]

hm I think dtka won actually, since Denver should be a lock to beat houstan

[Mike G]

There are 7 teams whose record will fall somewhere between the guesses of TmTj and dtka. So a W or a L will favor one or the other.
Teemo is currently .040 ahead. (Just yesterday the diff was .147!) W or L by these teams favors dtka by the amounts shown:

Code: Select all

tm  dtka  gain   opp
Atl   W   .033   Orl
Mil   W   .027   Det
Was   W   .060   Mia
Den   W   .040   Hou
GSW   L   .033   LAC
Min   W   .007   Uta
SAS   L   .040   Tor

None of these teams will play another, so the total possible 7-0 sweep would be .240. That is unlikely, but there are several avenues for the upset; and probably some that would yield an exact tie.
Note: these are all multiples of .00666..., or 1/150, which is equivalent to 0.1 in the win projection at b-r.com.

[TeemoTeejay]

Well that’s not ideal for me lol

[TeemoTeejay]

Congrats DTKA
wild finish just like the play in lol, teams I wanted to win personally won so Im happy with this too (although would rather some east stuff have gone my way D: )

[Mike G]

FINAL

Code: Select all

.   avg err   rmse   r^2       .  avg err   rmse   r^2
dtka   7.50    9.7   .46      emin   8.47   10.9   .37
TmTj   7.63   10.0   .44      eExp   8.48   10.7   .42
avgA   7.82   10.1   .41      WShr   8.49   10.3   .39
DQin   7.80   10.3   .42      24pr   8.51   10.8   .33
medi   7.87   11.1   .37      KPel   8.52   10.0   .58
bpmW   8.14   10.0   .43      24py   8.66   11.7   .33
Crow   8.33   10.6   .41      bmgm   9.13   12.1   .33
vegas  8.40   10.7   .38      eWin   9.65   11.9   .20
.                             perW   9.96   12.0   .18

KPel wins at e>2.57
dtka takes everything else.

[Crow]

PERw takes last.

24 prior regressed barely beats 24 prior.

avgA was a good blend.

Presumably dtkavana used a craftednba metric blend. Outright CPM or adjusted?