Comparing BPM and DarkoPM

[Crow]

I may do a number of these, starting with W Kessler.

Worse than -3 on BPM, +1 on Darko.

3/4ths of the difference on defense. BPM is heavily about team on defense. I am more inclined to take near the Darko estimate.

Difference on offense is not much more than a point. BPM appears to heavily punish low assist guys. Kessler on / off basically no difference in team assist rate, which is above league average. Should Kessler take a hit for not being an assistmaker? Maybe, but I'd probably lessen the hit.

One is an evaluation, the other is a projection, though it is also an evaluation on other tables.

I probably rely on all in one metrics more than I should (and should settling for just one) but most of the alternatives (own eye test, consensus or subjective impression from unweighted discrete stats) are probably worse on average.

RAPM was a useful third data point. I hope it comes back in some form beyond RPM. I dunno if I'd use RPM. Has been years since I resorted to that.


Kessler, 15pts, 15 rebs (11 offensive) and 4 blocks last night. Should go up on next day Darko.

[DSMok1]

Interesting comparison!

Kessler is the sort of player that is the hardest to measure with a box-score-only metric like BPM. Blocks are a notoriously poor measure of defensive value... In fact, assists by the big man have a better correlation with defensive value than blocks.

I would trust DARKO over BPM due to the added statistics they are using.

[v-zero]

I have Kessler at -1 at the moment in my now-cast ratings, which use a combination of a (very) extended box score extracted from PBP data, and plus minus type impact data.

There are plenty of players that I differ quite a bit on with DARKO, I have 8 of the same top 10 and 14 of the same top 20, but for instance I have Shai at 5, and DARKO has him at 24... I have him as a +4.4 player, DARKO sees him as +2.5.

Having said that I personally see DARKO is the pinnacle of public one number metrics.

[Crow]

Kessler at -1 is at the mid-point of BPM and DPM estimates.

[Mike G]

Does DARKO update daily? and is there a link to it?
I am reminded it's the work of kmedved, who has not posted here in almost 2 years.

Last year I dug into the prediction contest history as far back as 2016, and "kmed" was in it a few times; finished on top a couple times:

Code: Select all

year  #ent. MAE  RMSE
2016   22    1t    3
2017   23    1     8
2018   20    4     9
2019   18   15    15
as Darko:
2023   19    5     4

#ent is the number of entries, not counting dummies.

[Crow]

Darko hasn't updated since 11-6. I haven't watched that indicator on purpose but it might be once or twice a week at worst, maybe better at times past.

[v-zero]

Does DARKO update daily? and is there a link to it?

It's generally up to date. Go here and choose Current Player Skill Projections from the top menu.

[Crow]

Chet Holmgren, almost +6 on BPM. +1.2 on Darko.

Wembanyama, about -3 now on BPM. +0.3 on Darko.

Chet, +2.4 team +/- raw on court, +1 on / off. Wembanyama team -7 +/- on court, +11 on / off.

[Crow]

Darko was updated today.

[DSMok1]

Chet Holmgren, almost +6 on BPM. +1.2 on Darko.

Wembanyama, about -3 now on BPM. +0.3 on Darko.

Chet, +2.4 team +/- raw on court, +1 on / off. Wembanyama team -7 +/- on court, +11 on / off.

Recall that DARKO is a prediction system, so it will be heavily regressed toward the prior for these players with very few minutes played yet.

[v-zero]

Recall that DARKO is a prediction system, so it will be heavily regressed toward the prior for these players with very few minutes played yet.

Exactly, I have Wemby at about -0.5 right now, but integrate out the priors and he's been around +6....But the priors are there for a very good reason!

[DSMok1]

Recall that DARKO is a prediction system, so it will be heavily regressed toward the prior for these players with very few minutes played yet.

Exactly, I have Wemby at about -0.5 right now, but integrate out the priors and he's been around +6....But the priors are there for a very good reason!

Strong. How about Chet?

[v-zero]

Strong. How about Chet?

He's about a -1, integrating out the priors has him around +5 on the box component but roughly 0 on the plus/minus impact component, which given they are roughly evenly weighted puts him around 2.5 - Wemby's box component on the other hand is slightly below zero, but his plus/minus impact component is massive.

I wouldn't read much into either just yet.

[kmedved]

re: Updates - I run them daily, but sometimes the linkage to the the Shiny App breaks (I am not gifted with front end web dev stuff, and my partner on the app is often busy).

The Google Sheet link here will update daily and is more reliable - just without the fun visuals: https://docs.google.com/spreadsheets/d/ ... 1064086941

[Crow]

Is Darko always and only a projection even under historical career trajectory tab?

Is current skills projection the last data point or the trend line?