APBRmetrics

Crow wrote: ↑Mon Apr 14, 2025 1:24 am Who / how many regressed as a separate step? Based on guess or formal analysis? By a standard amount or informed by team specific experience relative to estimated strength?

Any other adjustments people want to share / show-off / ask about / discuss?

Anybody include coach and / or front office as separate variables in a model or guesstimate? I include those considerations in my final subjective adjustments but not in a formal or data driven way.


If someone was really motivated, could compare results to predictions and estimate how much was performance estimate error, minutes distribution error, other coach / front office variance (especially trades and other roster adjustments but not limited tovthat), luck variance, injuries, SOS shift from average or expected...

If you could reasonably estimate impacts of everything else, could any residual be ascribed to front office / coach? Which did best / worst?

For those with player level age / development adjustments, how close did it track on average? Outlier cases?

TeemoTeejay wrote: ↑Wed Apr 09, 2025 9:41 pm

kmedved wrote: ↑Wed Apr 09, 2025 6:35 pm

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

TeemoTeejay wrote: ↑Mon Apr 14, 2025 5:49 pm Sounds like an age adjustment was the missing piece for me
I was doing this as a preliminary test of the metric, so I was more so concerned to see where it would rank by all star break before trade deadline shenanigans happened. Looking through my results it felt like age adjustment stuff + it being very high on the mavs/suns/pels who all had injury issues or age decline issues was what separated it from how Darko did + it predicting relatively widely in terms of range at least, so im pretty happy with the results even though a bit disappointed in the final day loss (Although as a fan the results on the west were how I wanted them to go and I'll gladly take that over a win).

Pretty happy with a second place finish over CraftedDPM considering how good it had done in recent years, The MAMBA metric itself I basically had to take it down since I got a job so it hasnt been updated in awhile (I dont think its available anymore since I took it down too), but im happy it beat out the predictive versions of EPM and LEBRON.

I am curious how you guys went about projecting from a chosen metric though, all I did was lm(Mamba ~ Wins) just since it was a test run of the metric, but seeing what Darko and the LEBRON/EPM BBI projections were (In terms of preseason adjustments and vegas stuff) it sounds like that might be just as complicated as a step

knarsu3 wrote: ↑Tue Apr 15, 2025 4:47 pm Normally I've seen people using whatever metric- MAMBA, LEBRON, EPM, DARKO etc. and take that to generate win probabilities based on the minutes and simulate the season (or you can just calculate and add up based on the schedule). But I honestly think that is maybe unnecessarily complicated and that your method (and mine in that other win projection model) is the best way- just do lm(Wins ~ Mamba (or LEBRON/DARKO etc.) and you can include other variables like preseason net rating etc. It's also just so much easier now with the uncertainty with the schedule due to the IST games. Schedules aren't really balanced and so imo, I do think your way is the best way.

Mike G wrote: ↑Sun Apr 13, 2025 11:19 pm 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
medi   7.87   11.1   .37      24pr   8.51   10.8   .33
DQin   7.93   10.3   .42      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.

Mike G wrote: ↑Thu Apr 17, 2025 6:19 am A very strange thing happened. In mid Jan. I opened my Excel file to update, and it had reverted to a version saved sometime in Nov. This after updating (and saving) almost every day in the meantime.

I searched all thru my computer for another version of the file but never found one. So I blame microsoft.
One page in the file (named "log") is a daily record of everyone's MAE that is nice to be able to reference. I patched in entries for the missing weeks by copying from this thread; it's every few days instead of daily.

So yeah, I guess we fixed the Cha/Chi issue in that 'lost' interval. Sorry about that.
Maybe mediocre will swap his 3rd-place prize money for yours.

Crow wrote: ↑Tue Apr 15, 2025 5:45 pm KMedved, what ws the final error for Darko?

Crow wrote: ↑Tue Apr 29, 2025 7:28 pm It was a good performance.

Pre-season adjustments by player and / or team?