538 Releases RAPTOR - All in one metric

Moderators: Domejandro, infinite11285, Harry Garris, ken6199, Dirk, bisme37, KingDavid, bwgood77, zimpy27, cupcakesnake

giberish
RealGM
Posts: 15,824
And1: 5,809
Joined: Mar 30, 2006
Location: Whereever you go - there you are

Re: 538 Releases RAPTOR - All in one metric 

Post#101 » by giberish » Mon Oct 14, 2019 3:10 am

Doctor MJ wrote:
Ambrose wrote:Advanced stats sure love Chris Paul. It's hard to take seriously when it puts him over guys like LeBron, who is simply on another level as a player.


I'll be interested to see discussion on what this stat says about LeBron in the time to come. It feels like a slap in the face that hasn't yet connected.

Now, peak LeBron does quite well in the metric, but over the past 6 years - the stuff with the data actually driving the algorithm - LeBron is not getting rated highly. Why?

My guess is that a lot of this amounts to LeBron and his teams tending to coast in the regular season in more recent years.

Consider '16-17, the year where I felt like LeBron was more unstoppable than in any year of his career. The Cavs only won 51 games that year and had an utterly mediocre SRS. And then they followed it up with their most dominant run through the Eastern playoffs and put up a better fight against the Warriors than anyone else.

This wasn't anything like a 2.87 SRS team, but according to regular season data, it was. Perhaps makes sense then that LeBron doesn't come across as the GOAT there.


That's one thing that's hard to model - but would be vitally important in trying to project playoff performance. The 'give a damn' metric. Most teams and players have a full standard level of ''give a damn' throughout a season.

LeBron clearly operated at less then full 'give a damn' during his last Cleveland run. IMO the rest of the team had a lower intensity because of it as well (following his lead).

Golden State operated at full 'give a damn' during the 2014-5 and 2015-6 seasons but then noticeably slacked off in intensity in 2016-7 and even further the last two years.

Kwahi operated at less then full 'give a damn' last year, though it didn't seem to effect the rest of the Raptors the way that LeBron did in Cleveland.

When players or full teams play at less then full intensity during the regular season it's easy for them to improve for the playoffs (while their regular season stats will be worse then their overall quality).
Doctor MJ
Senior Mod
Senior Mod
Posts: 50,604
And1: 19,355
Joined: Mar 10, 2005
Location: Cali
     

Re: 538 Releases RAPTOR - All in one metric 

Post#102 » by Doctor MJ » Mon Oct 14, 2019 4:19 am

giberish wrote:
Doctor MJ wrote:
Ambrose wrote:Advanced stats sure love Chris Paul. It's hard to take seriously when it puts him over guys like LeBron, who is simply on another level as a player.


I'll be interested to see discussion on what this stat says about LeBron in the time to come. It feels like a slap in the face that hasn't yet connected.

Now, peak LeBron does quite well in the metric, but over the past 6 years - the stuff with the data actually driving the algorithm - LeBron is not getting rated highly. Why?

My guess is that a lot of this amounts to LeBron and his teams tending to coast in the regular season in more recent years.

Consider '16-17, the year where I felt like LeBron was more unstoppable than in any year of his career. The Cavs only won 51 games that year and had an utterly mediocre SRS. And then they followed it up with their most dominant run through the Eastern playoffs and put up a better fight against the Warriors than anyone else.

This wasn't anything like a 2.87 SRS team, but according to regular season data, it was. Perhaps makes sense then that LeBron doesn't come across as the GOAT there.


That's one thing that's hard to model - but would be vitally important in trying to project playoff performance. The 'give a damn' metric. Most teams and players have a full standard level of ''give a damn' throughout a season.

LeBron clearly operated at less then full 'give a damn' during his last Cleveland run. IMO the rest of the team had a lower intensity because of it as well (following his lead).

Golden State operated at full 'give a damn' during the 2014-5 and 2015-6 seasons but then noticeably slacked off in intensity in 2016-7 and even further the last two years.

Kwahi operated at less then full 'give a damn' last year, though it didn't seem to effect the rest of the Raptors the way that LeBron did in Cleveland.

When players or full teams play at less then full intensity during the regular season it's easy for them to improve for the playoffs (while their regular season stats will be worse then their overall quality).


Well, this is a reasonable thing to bring up, but to me it's not something we should really be looking to model, but of course, I tend to disagree with basketball statisticians about the purpose of basketball statistics.

They tend to think the goal is to create things that perform as well as possible without any human analysis, I believe the goal should be to aid human analysis.

This is why I've always lamented the dominance of stats in the XRAPM family. The box score and the +/- stats tell different parts of the same story and you can't get maximum insight simply by adding them together. Hence while I'm not against making an all-in-one stat using both, the fact that this led to a loss of advanced +/- only stats made it a net loss in my book.

Going back to the present concern: I don't need a stat to tell me a player or team is slacking off, it's obvious. Putting a great deal of effort into making a stat for this that will likely be worse than my own judgment for the foreseeable future is not the best use of statisticians imho.
Getting ready for the RealGM 100 on the PC Board

Come join the WNBA Board if you're a fan!
ShotCreator
Analyst
Posts: 3,479
And1: 2,328
Joined: May 18, 2014
Location: CF
     

Re: 538 Releases RAPTOR - All in one metric 

Post#103 » by ShotCreator » Mon Oct 14, 2019 4:36 am

ClipsFanSince98 wrote:Someone help me understand. If Paul is so Godly in this stat, why did Houston go to clear title favorites and 57 wins despite trading him for WB he ranks nowhere near as good? So before trade were Rockets expected to win 65 with 50% title odds or something??

The Rockets absolutely destroyed the NBA for about 57 straight games with the #1 offense and #2 defense. Once they cut dead weight, Paul got back and the defensive scheme settled they turned it up to a pretty high level. Over a quarter into the season they were in the lottery. It took some extreme dominance to damn near take the #2 seed. If they ran it back a 60 win team is not surprising at all, barring health concerns...
Swinging for the fences.
KqWIN
RealGM
Posts: 15,520
And1: 6,360
Joined: May 15, 2014
 

Re: 538 Releases RAPTOR - All in one metric 

Post#104 » by KqWIN » Mon Oct 14, 2019 4:44 am

Doctor MJ wrote:
Well, this is a reasonable thing to bring up, but to me it's not something we should really be looking to model, but of course, I tend to disagree with basketball statisticians about the purpose of basketball statistics.

They tend to think the goal is to create things that perform as well as possible without any human analysis, I believe the goal should be to aid human analysis.

This is why I've always lamented the dominance of stats in the XRAPM family. The box score and the +/- stats tell different parts of the same story and you can't get maximum insight simply by adding them together. Hence while I'm not against making an all-in-one stat using both, the fact that this led to a loss of advanced +/- only stats made it a net loss in my book.

Going back to the present concern: I don't need a stat to tell me a player or team is slacking off, it's obvious. Putting a great deal of effort into making a stat for this that will likely be worse than my own judgment for the foreseeable future is not the best use of statisticians imho.



I didn't know this was a common thought, but you're the second person who has this opinion. While I can see how one would can appreciate a "pure" +/- approach, it is not mathematically sound to exclude box score statistics. It would be similar to ignoring point differential in favor of W/L percentage. We know that W/L is prone to variance and point differential is a better estimator of true strength and future W/L itself. The same applies here. The addition of box score stats into these models is not to add them for the sake of adding them, it's to stabilize something we know is variable in smaller sample sizes (RAPM).
Ambrose
Assistant Coach
Posts: 3,816
And1: 4,227
Joined: Jul 05, 2014

Re: 538 Releases RAPTOR - All in one metric 

Post#105 » by Ambrose » Mon Oct 14, 2019 8:10 pm

I_Like_Dirt wrote:
Ambrose wrote:Advanced stats sure love Chris Paul. It's hard to take seriously when it puts him over guys like LeBron, who is simply on another level as a player.


It's hard to take you seriously when you have no idea what you're just making stuff up. This stat doesn't say Paul is better than Lebron. It actually says the top 12 individual seasons since 1976-77 were either Michael Jordan (7 seasons), Lebron (3 seasons), or Steph Curry (2 seasons). Sure doesn't seem like it puts Paul ahead of Lebron to me. It does seem to suggest that Lebron has coasted on defense a fair bit in the regular season and Paul hasn't as much. Whether or not that passes your eye test is up to you.


The chart that shows up in the link provided has 4-5 Paul seasons before any LeBron in the last few years. That's what I was referencing. Chill out brother.

Return to The General Board