APBRmetrics

ampersand5 wrote: It would be awesome if you could calculate that and post the results; I'm really interested in analyzing that data.

The one problem I remember having (and will probably be amplified using BPM) is that I'm not exactly sure how to deal with negative numbers. I think for WS I just looked at the top 10 players on a team (which were all pretty much positive or zero) but with BPM I would assume a far larger percentage are below zero. I might use VORP, I don't know what the distribution of that is. Maybe SD of BPM among top 10 players? Any ideas?

This might be a case (among others) where further explanation of the findings in layman's terms would help pull people in, along, up. I double-checked how gini coefficients worked. Many may not have had to but some others probably would need to but might not. If part of this board's purpose is education, it might be useful for posters to go an extra half-mile or mile on explanations. Or post an explanatory link, like this one: http://en.m.wikipedia.org/wiki/Gini_coefficient

italia13calcio wrote:

ampersand5 wrote: It would be awesome if you could calculate that and post the results; I'm really interested in analyzing that data.

The one problem I remember having (and will probably be amplified using BPM) is that I'm not exactly sure how to deal with negative numbers. I think for WS I just looked at the top 10 players on a team (which were all pretty much positive or zero) but with BPM I would assume a far larger percentage are below zero. I might use VORP, I don't know what the distribution of that is. Maybe SD of BPM among top 10 players? Any ideas?

I think using VORP would be preferable because its a better measure of impact, which is what we are trying to isolate and evaluate.

further to Crow's point - I would be happy to write up an explanatory analysis of your numbers to help present the material to a wider audience in order to promote the info (only if you did not want to).

Require a plus 6 or better on RPM and a winning record and the number of superstars with a 10% chance or greater of a title this year falls to six, basically in line with colts18's other methods.

Calculating the gini on VORP isn't working as nicely - some teams negative or near zero which greatly throws off the calculations. Another method that I looked at is sd divided by the mean of the absolutes, this can be applied to anything - BPM, VORP, ws. Might be better. Correlation between this and gini (for the sensical gini scores) is pretty high. I currently have all the calculations back to 1974 or whenever bref first has BPM. Haven't filtered it down to just champions or finalists yet.

italia13calcio wrote:Calculating the gini on VORP isn't working as nicely - some teams negative or near zero which greatly throws off the calculations. Another method that I looked at is sd divided by the mean of the absolutes, this can be applied to anything - BPM, VORP, ws. Might be better. Correlation between this and gini (for the sensical gini scores) is pretty high. I currently have all the calculations back to 1974 or whenever bref first has BPM. Haven't filtered it down to just champions or finalists yet.

Can you post this numbers on a google sheet - I'd love to have a look at them.


To get around negative VORP, I would just look at the top 8 players on the team with the highest vorp. With that criteria for NBA championship winning teams, I don't think there will be any players with a negative VORP.

If you look for a top 5 player on RPM on a team that has both an offensive efficiency above 108 and defensive efficiency lower than 104, only Curry remains. Everyone else misses by at least a little on these 3 criteria. What does the record show? Haven't looked yet for precisely all 3.

Using the slightly different criteria of top 5 player and 2 or more points better than league average offensive and defensive efficiencies, 10 of the last 15 champs essentially met all 3 criteria (counting one 0.2 pt miss as a make). Only the 2010 Lakers missed on 2 of 3.

How many teams met 2 of 3 criteria this season? 6- GSW, HOU, SAS, CLE, OKC and ATL. LAC misses on RPM leader slightly for a 2nd strike against.

I've posted about this a few times. I believe the "you need a superstar to win a title" claim is overrated for two main reasons:

- The league is constantly evolving. There is a lot of NBA titles, but barely a handful in the analytics-driven era of GMing and coaching and this level of knowledge of the value of spacing, efficiency, not as much emphasis on ppg, etc. Conceivably analytics and smart teams alone could make the sample size of superstar-driven titles now an untrustworthy barometer of what it takes to win a title because teams now have a cheat code. Asking whether the 2015 Atlanta Hawks can win a title has little comparable because in all the decades of superstar-less title teams coming up short, they didn't have the resources or frankly commitment to trying whatever it takes to break through, that sharp teams have now. This is before considering tons of other changes such as a vastly different CBA from the late 90s on, rules changes like the hand check, very different drafting than when everyone stayed 3-4 years in college, etc. The farther you go back the more useless NBA titles are to predict modern day. The 60s are useless because in an 8-9 team league you were basically as likely to have Russell, Wilt, West, or Oscar as you are to have any all-star today. The 80s title distribution was made specifically by a Lakers 45 W pretender with Kareem getting 2 #1 picks in 3 years to take Magic and Worthy, and a 60 win Celtics after Bird's rookie year, also having a #1 pick in their back pocket to move for McHale and Parish. Considering how likely it is Magic, Worthy, McHale all go to bad teams in modern day instead of ones with superstars on it, it changes how well that decade can be used as a predictor for modern day. To me between the growth in style of play and the CBA even 20 seasons may be too long to matter in terms of predicting who can win a title now. NBA has gone from checkers to chess in 20 years. From serve and volley to baseliners.

- I believe teams who came very very close to winning the title should count as much last the ones who do and it so happens that the last decade and a half most of the teams that came closest to winning a title without getting one, were ensemble-style teams. The 2013 Heat winning the title instead of the 2013 Spurs means nothing in terms of whether you need an MVP candidate to win a title when the Spurs win 99 times/100 with 30 seconds left and superstar Lebron badly bricked a shot that could've ended the series. Likewise the 2010 Lakers go into the category of a superstar title instead of the ensemble Celtics (ignoring for now many APBR members could consider KG a superstar this year) when Kobe goes 6 for 24 in Game 7 and the Celtics have a healthy double digit lead in the second half? The 2005 Pistons have a lead at the end of the 3rd quarter in Game 7 where Duncan shoots poorly, and may have finished the series in 6 if they had won the Robert Horry game. 2000 Blazers have all time collapse and probably just bad luck in terms of good looking shots bouncing out, a collapse that would again happen <5% of the time, or else them or Pacers are champion this year. 2002 Kings are debatable in terms of whether they're a star-led team but they also could've easily won the title that year and should've if not Donaghy. So you take these handful of teams that proved they were as title caliber as the teams they lost to and it changes the numbers the last decade and a half. If the Hawks put themselves one more good quarter away from winning the title like all the above losers, they're just as likely as the superstar-led opponent to win the title

I am fine with your general point.

By RAPM, the 2008-9 and 2009-10 Lakers were not led by a top 5 guy.

ampersand5 wrote: Can you post this numbers on a google sheet - I'd love to have a look at them.


To get around negative VORP, I would just look at the top 8 players on the team with the highest vorp. With that criteria for NBA championship winning teams, I don't think there will be any players with a negative VORP.

heres the spread sheet:
https://docs.google.com/spreadsheets/d/ ... edit#gid=0

If your just interested in the championship winning/contending teams then you can look at the top 10 players, which is what I did. The only teams that have wierd Gini are bad teams, that's why there are scores not between 0 and 1. I also included SD/mean of the absolutes, which may be a more appropriate measure. Let me know if you want any changes, like top 8 or whatever.

Regarding writing it up, feel free to do so. I will hopefully use this in a larger post at some point, but feel free to use this in any way also.

italia13calcio wrote:

ampersand5 wrote: Can you post this numbers on a google sheet - I'd love to have a look at them.


To get around negative VORP, I would just look at the top 8 players on the team with the highest vorp. With that criteria for NBA championship winning teams, I don't think there will be any players with a negative VORP.

heres the spread sheet:
https://docs.google.com/spreadsheets/d/ ... edit#gid=0

If your just interested in the championship winning/contending teams then you can look at the top 10 players, which is what I did. The only teams that have wierd Gini are bad teams, that's why there are scores not between 0 and 1. I also included SD/mean of the absolutes, which may be a more appropriate measure. Let me know if you want any changes, like top 8 or whatever.

Regarding writing it up, feel free to do so. I will hopefully use this in a larger post at some point, but feel free to use this in any way also.

Thank you a ton for putting up your data. Would the correlation between the Gini and SRS/winning pct be meaningful? or is team construction too varied for the number to be coherent?

Here are the Gini coefficients (inequality of impact) for all the NBA finalists over the last decade.
Spurs are the most equal teams while Lebron-lead teams are the least equal. Not entirely surprising but seeing DET near the bottom goes to show how the 'superstar' thing misses the mark.

Couple of different pieces in this thread (and at least one that could have been on team inequality).
To probably wrap up my piece, I'll say that a superstar who wins or could win a title is a superstar in a strong team efficiency context and usually a two-way strong context. Just being a RPM superstar is not enough. It has to come in and be compatible with strong team efficiency.

Eleven years is a pretty short sample and there are other ways to group the data, but 5 of the 8 most "equal" won titles, 3 of the middle five and 3 of 9 for the least equal. This breakdown creates at least the appearance of a bias in favor of equality, overall.

2 of the higher equality title winners were in last 5 years, 3 in previous 6 years. All 3 of the middling equality winners were in the first 6 years. All 3 of the higher inequality winners came in the last 5 years. Maybe I am making a pattern as much as finding one, but I found / made it to think about it and track it from here.

Combine the middle group with the more equal or more unequal and the apparent story changes. Maybe there isn't a coherent, lasting story / trend.