What are all the new and best advanced metrics?

Moderator: Doctor MJ

NoParticular
Ballboy
Posts: 11
And1: 3
Joined: Oct 18, 2021
 

What are all the new and best advanced metrics? 

Post#1 » by NoParticular » Thu Jan 13, 2022 2:18 am

I'm new here for context. I was wondering if someone could provide a list of all the best advanced metrics. I know about RAPTOR, LEBRON, RAPM, EPM, etc etc, but there are some really good more unknown ones. Does anyone have a database of APM? That's an example.
blabla
Sophomore
Posts: 156
And1: 76
Joined: May 23, 2012

Re: What are all the new and best advanced metrics? 

Post#2 » by blabla » Mon Jan 17, 2022 6:32 pm

A lot of the current metrics (try to) do the same thing. Most of them do it badly, though that's something one wouldn't know until one built one of these themselves
More generally, there appears to be a glass ceiling: no significant progress has been made in regards to player metrics since 2013.
Some creators might have hyped certain features, but pretty much all of them are useless.

I believe there is a google doc somewhere that shows multiple metrics, and the average for every player. I'm sure if you look around on Twitter you can find it
jambalaya
Sixth Man
Posts: 1,668
And1: 286
Joined: Feb 01, 2005

Re: What are all the new and best advanced metrics? 

Post#3 » by jambalaya » Fri May 20, 2022 9:46 pm

One study:https://dunksandthrees.com/blog/metric-comparison

Author of DARKO found it to be best in general in another study. Was best this year in projecting 2021-22 season team results.

Explanation of past and projection are different purposes.
SNPA
Head Coach
Posts: 7,432
And1: 7,153
Joined: Apr 15, 2020

Re: What are all the new and best advanced metrics? 

Post#4 » by SNPA » Tue Jun 7, 2022 4:39 am

blabla wrote:A lot of the current metrics (try to) do the same thing. Most of them do it badly, though that's something one wouldn't know until one built one of these themselves
More generally, there appears to be a glass ceiling: no significant progress has been made in regards to player metrics since 2013.
Some creators might have hyped certain features, but pretty much all of them are useless.

I believe there is a google doc somewhere that shows multiple metrics, and the average for every player. I'm sure if you look around on Twitter you can find it

Delete the bold part and it right, there’s no need to try and create one. The idea of an all in one (or however it’s described) metric is dumb and no one that understands basketball would put any stock into it. It’s for stat nerds to nerd out on, and hey…if they enjoy it by all means. Have a blast. It’s just not relevant to understanding how basketball played by humans actually works.
tsherkin
Retired Mod
Retired Mod
Posts: 78,421
And1: 19,970
Joined: Oct 14, 2003
 

Re: What are all the new and best advanced metrics? 

Post#5 » by tsherkin » Wed Jun 22, 2022 4:24 pm

SNPA wrote:
blabla wrote:A lot of the current metrics (try to) do the same thing. Most of them do it badly, though that's something one wouldn't know until one built one of these themselves
More generally, there appears to be a glass ceiling: no significant progress has been made in regards to player metrics since 2013.
Some creators might have hyped certain features, but pretty much all of them are useless.

I believe there is a google doc somewhere that shows multiple metrics, and the average for every player. I'm sure if you look around on Twitter you can find it

Delete the bold part and it right, there’s no need to try and create one. The idea of an all in one (or however it’s described) metric is dumb and no one that understands basketball would put any stock into it. It’s for stat nerds to nerd out on, and hey…if they enjoy it by all means. Have a blast. It’s just not relevant to understanding how basketball played by humans actually works.


This is a fairly limited view of the value of unified metrics. They have weaknesses and need to be taken with a grain of salt, but they do have some value. Each of them operates in a given way and will have bias, no different than a subjective human eye test. A lot of them end up heavily rewarding box score production, particularly scoring, so you'll see certain people represented that way. The +/- stuff is very role-specific, so it is challenging to use for comparisons because each number is contextually framed. But as you stack a bunch of them next to one another, you start seeing certain trends which can help support or undermine the picture of a given player's value. They can also be used to find potential "diamonds in the rough," and have historically had some success at doing so. It's all in understanding the pros and cons of the tools involved instead of just categorically dismissing them without particularly good cause.
SNPA
Head Coach
Posts: 7,432
And1: 7,153
Joined: Apr 15, 2020

Re: What are all the new and best advanced metrics? 

Post#6 » by SNPA » Wed Jun 22, 2022 6:19 pm

tsherkin wrote:
SNPA wrote:
blabla wrote:A lot of the current metrics (try to) do the same thing. Most of them do it badly, though that's something one wouldn't know until one built one of these themselves
More generally, there appears to be a glass ceiling: no significant progress has been made in regards to player metrics since 2013.
Some creators might have hyped certain features, but pretty much all of them are useless.

I believe there is a google doc somewhere that shows multiple metrics, and the average for every player. I'm sure if you look around on Twitter you can find it

Delete the bold part and it right, there’s no need to try and create one. The idea of an all in one (or however it’s described) metric is dumb and no one that understands basketball would put any stock into it. It’s for stat nerds to nerd out on, and hey…if they enjoy it by all means. Have a blast. It’s just not relevant to understanding how basketball played by humans actually works.


This is a fairly limited view of the value of unified metrics. They have weaknesses and need to be taken with a grain of salt, but they do have some value. Each of them operates in a given way and will have bias, no different than a subjective human eye test. A lot of them end up heavily rewarding box score production, particularly scoring, so you'll see certain people represented that way. The +/- stuff is very role-specific, so it is challenging to use for comparisons because each number is contextually framed. But as you stack a bunch of them next to one another, you start seeing certain trends which can help support or undermine the picture of a given player's value. They can also be used to find potential "diamonds in the rough," and have historically had some success at doing so. It's all in understanding the pros and cons of the tools involved instead of just categorically dismissing them without particularly good cause.

You highlight why they can be generally dismissed.

There is bias in their construction. There is bias in their outputs. There is bias in their interpretation. At some point it’s just arguing what you want using them as justification and hiding behind numbers to appear unbiased.
tsherkin
Retired Mod
Retired Mod
Posts: 78,421
And1: 19,970
Joined: Oct 14, 2003
 

Re: What are all the new and best advanced metrics? 

Post#7 » by tsherkin » Thu Jun 23, 2022 6:42 am

SNPA wrote:You highlight why they can be generally dismissed.

There is bias in their construction. There is bias in their outputs. There is bias in their interpretation. At some point it’s just arguing what you want using them as justification and hiding behind numbers to appear unbiased.


That's certainly one way to look at them. But if bias is what concerns you, then you've got basically nothing to work with at all, subjectively or objectively. At some point, it becomes an exercise in piecing together information and looking for commonalities.
dantas
Sophomore
Posts: 166
And1: 123
Joined: Jun 22, 2018

Re: What are all the new and best advanced metrics? 

Post#8 » by dantas » Wed Feb 1, 2023 6:04 pm

I think the hustle stats are very underrated

https://www.nba.com/stats/players/hustle-leaders

Return to Statistical Analysis