My GOAT formula and results (Peak, Prime, Longevity and all time) - 22 players updated (last - Isiah Thomas)

[tsherkin]

I'm not going to include stuff like MVP awards, since not every MVP is equal or even deserved.

Could include top-5 MVP finishes, though. The voters usually don't screw THAT up very often.

[trex_8063]

I've been developing a formula to determine how good was a player in each in the NBA. And ultimately, his career value, peak value and prime value (his best 5 years, don't have to be consecutive).

The Formula is:
Regular season
(PPG+TRB*2+APG*2+STL*2+BLK*2-TOV*2+PER+WS/48/2+TS%/2)/(Missed games coefficient) = RS Value

Post season
(PPG+TRB*2+APG*2+STL*2+BLK*2-TOV*2+PER+WS/48/2+TS%/2)/(Rounds played coefficient) = PS Value

Peak Value - Biggest sum in the same year: RS Value + PS Value
Prime Value - Sum of the top 5 years from the player
Longevity Value - Sum of All RS Value + PS Value
All Time Value = Longevity Value/10*0,3 + Prime Value/5 *0,6 + Peak Value * 0,1

I have a few questions/comments/critiques regarding the formula....
1) How are you using WS/48 and TS%? What I mean is: WS/48 is a decimal (and you're dividing it by 2). Example: '09 Lebron rs WS/48 was .318 (divided by 2 = .159). When you're adding up other variables that are often in the 20's or 30's, adding another 0.159 doesn't exactly shift the needle much (it's virtually insignificant).
So did you shift the decimal point three spots to the right to input WS/48 to the formula (e.g. .318 is put in the formula as 318)? If so, that gives WS/48 approximately 5x the weight of PER.
otoh, if you only move the decimal two spots to the right, WS/48 is only given about half the weight of PER (because you're dividing it by 2).
So I'm curious as to why you're giving WS/48 either so much more or so much less weight.

1b) Same basic question for TS% (input as 56.0 as opposed to .560, for example, yes?). If so, I feel the formula's giving too much weight to shooting efficiency (which is already a BIG factor in WS/48, btw).


2) What exactly are your "Missed games coefficient" and "Rounds played coefficient"?


3) I know you mentioned why you used per game stats rather than per 100 possession numbers, but surely you see how this is going to skew things slightly in favor of players in faster paced eras (essentially skewing against the circa-2000 crowd). You can use per 100 possession numbers and still easily circumvent the concerns you mentioned by simply including mpg in your formula.


4) I gather you want to give a lot of credit to longevity. However, I don't feel your formula construct actually does so. For instance, you're using a set number of seasons to define prime: 5 years. For some players, that just about covers their entire prime. But then there are guys like Kareem, Karl Malone, or Robert Parish, who all had primes that lasted something like 12-15 years. But they're not receiving any extra credit over someone like Sidney Moncrief (short prime) for doing so, because you're still just picking out a 5-year stretch of time for each of them.

You might say that they'll get the credit for that longevity in the "Longevity Value", but unless I'm misinterpreting something, I don't see that that's the case. Because you're still using per game numbers and rate statistics (PER and WS/48) only........nothing that recognizes the length of time they were productive for.
And in fact, it would seem to penalize players for playing addition years in a role-player capacity, as those role-player years are going to depress their career ppg/rpg/PER/WS/48/etc numbers.

Example: Suppose we compare Robert Parish to some hypothetical Player X who played just 6 seasons (5 in his prime, and just one additional role-player year). Player X's career numbers are not going to be depressed that much, because 83+% of his career was prime. Parish, otoh, only played ~60-70% of his career in his prime; so his additional role-player years are having a bigger impact on his career ppg/rpg/PER/etc that you will be using in your formula. And again: no extra credit for his prime which lasted like 12+ years, because you're singling out only a 5-year period.

1) I understand what you mean with the WS/48 thing. The formula is wrong in the OP (I will correct it), I don't divide it by 2. Only ts% is being divided.
In your example - LeBron's WS/48 being 0.318 I'm using the value 31.8. 30 WS/48 is pretty much elite, so it's valued as much as PPG, RPG, TS% and PER.

I see. Well, that's fairly appropriate then.

With ts% I'm using values like 55.1. I'm dividing it by two because 60ts% is elite. So I'm cutting it in half so it doesn't weight more than PER.

I still feel that gives too much valuation to shooting efficiency (which is already factored into both PER and WS/48, and NOT to a small degree where the latter is concerned).

2) Missed Games coefficient is a number I divide the regular season stats. Since I'm operating on per game numbers, I must have a coefficient since I can't give the same amount of value to a guy who plays 40 games in the RS and 82 games in the RS. The coefficient is made in gaps (I think it's irrelevant if a player had 82 games or 79, since some of them might be in the end of the season. However, more than 10 games missed and it starts showing.

The rounds played coefficient acts the same way. I can't value a guy's production when he played 1 post season round as much as a guy who played 4 rounds. One round is a very small sample size, so it kind of eliminates 1 round absurd production. Or two rounds.

Hakeem 88, or Wade 2010 should be good examples on the cut of such value.

3) That's definitely a smart observation.

4) The PPG is calculated by season. Even the "crap seasons" add value. They're not taking away, since I calculate in the end the year value and the sum of that goes into longevity. I don't know if I explained myself right, please tell me if you didn't understand.

I can send you my Excel file if you wish, so you can understand better how it works.

Regarding answer to #2: I understood that was the point of those coefficients; I was actually wanting to see the EXACT numbers used, to see if I agree with how you have things weighted.
If you don't mind, maybe a look at your Excel sheet would be the quickest way to answer all my questions, because I'm still not quite understanding what you're saying in answer to #4, too.

[trex_8063]

I'm not going to include stuff like MVP awards, since not every MVP is equal or even deserved.

Could include top-5 MVP finishes, though. The voters usually don't screw THAT up very often.

Yeah. Or perhaps you could do it in gaps or groupings (something like X-value for a 1st or 2nd place finish in MVP vote, Y-value for 3rd or 4th place, Z-value for 5th or 6th).

[tsherkin]

I'm not going to include stuff like MVP awards, since not every MVP is equal or even deserved.

Could include top-5 MVP finishes, though. The voters usually don't screw THAT up very often.

Yeah. Or perhaps you could do it in gaps or groupings (something like X-value for a 1st or 2nd place finish in MVP vote, Y-value for 3rd or 4th place, Z-value for 5th or 6th).

Yeah, a weighted value for finishes in the top 5 of the MVP race would be cool.

[Joao Saraiva]

I have a few questions/comments/critiques regarding the formula....
1) How are you using WS/48 and TS%? What I mean is: WS/48 is a decimal (and you're dividing it by 2). Example: '09 Lebron rs WS/48 was .318 (divided by 2 = .159). When you're adding up other variables that are often in the 20's or 30's, adding another 0.159 doesn't exactly shift the needle much (it's virtually insignificant).
So did you shift the decimal point three spots to the right to input WS/48 to the formula (e.g. .318 is put in the formula as 318)? If so, that gives WS/48 approximately 5x the weight of PER.
otoh, if you only move the decimal two spots to the right, WS/48 is only given about half the weight of PER (because you're dividing it by 2).
So I'm curious as to why you're giving WS/48 either so much more or so much less weight.

1b) Same basic question for TS% (input as 56.0 as opposed to .560, for example, yes?). If so, I feel the formula's giving too much weight to shooting efficiency (which is already a BIG factor in WS/48, btw).


2) What exactly are your "Missed games coefficient" and "Rounds played coefficient"?


3) I know you mentioned why you used per game stats rather than per 100 possession numbers, but surely you see how this is going to skew things slightly in favor of players in faster paced eras (essentially skewing against the circa-2000 crowd). You can use per 100 possession numbers and still easily circumvent the concerns you mentioned by simply including mpg in your formula.


4) I gather you want to give a lot of credit to longevity. However, I don't feel your formula construct actually does so. For instance, you're using a set number of seasons to define prime: 5 years. For some players, that just about covers their entire prime. But then there are guys like Kareem, Karl Malone, or Robert Parish, who all had primes that lasted something like 12-15 years. But they're not receiving any extra credit over someone like Sidney Moncrief (short prime) for doing so, because you're still just picking out a 5-year stretch of time for each of them.

You might say that they'll get the credit for that longevity in the "Longevity Value", but unless I'm misinterpreting something, I don't see that that's the case. Because you're still using per game numbers and rate statistics (PER and WS/48) only........nothing that recognizes the length of time they were productive for.
And in fact, it would seem to penalize players for playing addition years in a role-player capacity, as those role-player years are going to depress their career ppg/rpg/PER/WS/48/etc numbers.

Example: Suppose we compare Robert Parish to some hypothetical Player X who played just 6 seasons (5 in his prime, and just one additional role-player year). Player X's career numbers are not going to be depressed that much, because 83+% of his career was prime. Parish, otoh, only played ~60-70% of his career in his prime; so his additional role-player years are having a bigger impact on his career ppg/rpg/PER/etc that you will be using in your formula. And again: no extra credit for his prime which lasted like 12+ years, because you're singling out only a 5-year period.

1) I understand what you mean with the WS/48 thing. The formula is wrong in the OP (I will correct it), I don't divide it by 2. Only ts% is being divided.
In your example - LeBron's WS/48 being 0.318 I'm using the value 31.8. 30 WS/48 is pretty much elite, so it's valued as much as PPG, RPG, TS% and PER.

I see. Well, that's fairly appropriate then.

With ts% I'm using values like 55.1. I'm dividing it by two because 60ts% is elite. So I'm cutting it in half so it doesn't weight more than PER.

I still feel that gives too much valuation to shooting efficiency (which is already factored into both PER and WS/48, and NOT to a small degree where the latter is concerned).

2) Missed Games coefficient is a number I divide the regular season stats. Since I'm operating on per game numbers, I must have a coefficient since I can't give the same amount of value to a guy who plays 40 games in the RS and 82 games in the RS. The coefficient is made in gaps (I think it's irrelevant if a player had 82 games or 79, since some of them might be in the end of the season. However, more than 10 games missed and it starts showing.

The rounds played coefficient acts the same way. I can't value a guy's production when he played 1 post season round as much as a guy who played 4 rounds. One round is a very small sample size, so it kind of eliminates 1 round absurd production. Or two rounds.

Hakeem 88, or Wade 2010 should be good examples on the cut of such value.

3) That's definitely a smart observation.

4) The PPG is calculated by season. Even the "crap seasons" add value. They're not taking away, since I calculate in the end the year value and the sum of that goes into longevity. I don't know if I explained myself right, please tell me if you didn't understand.

I can send you my Excel file if you wish, so you can understand better how it works.

Regarding answer to #2: I understood that was the point of those coefficients; I was actually wanting to see the EXACT numbers used, to see if I agree with how you have things weighted.
If you don't mind, maybe a look at your Excel sheet would be the quickest way to answer all my questions, because I'm still not quite understanding what you're saying in answer to #4, too.

Thanks for all the thoughts.

I'm including DBPM right now and actualizing every player's rankings. Someone suggested here that stat would be a good way to approach it. (Idk if it was you, but for who did it thank you).

I'm multiplying it by 4. I think it brings the right amount of weight, and it should reward defense enough. Some defensive aspects are already included in the formula, so I feel it might give it the boost it needs.

Once I'm finished putting DBPM numbers and actualizing rankings I'll send the file to you. Is there any possibility of sending trough here? If not then PM. I'll try to end that ASAP.

[trex_8063]

Once I'm finished putting DBPM numbers and actualizing rankings I'll send the file to you. Is there any possibility of sending trough here? If not then PM. I'll try to end that ASAP.

I don't know if you have a gmail or google account. If so, you can open Excel files in your Google Drive. Then (if you don't mind everyone here having access to it) you should be able to go to the privacy settings for the file ("Share" tab in upper right-hand corner of screen) and change it from "Private" to "Public" (just make sure you also set it such that it's "View-only access"). Then you can just post the link to that page here and all of us can view it.

[Joao Saraiva]

Once I'm finished putting DBPM numbers and actualizing rankings I'll send the file to you. Is there any possibility of sending trough here? If not then PM. I'll try to end that ASAP.

I don't know if you have a gmail or google account. If so, you can open Excel files in your Google Drive. Then (if you don't mind everyone here having access to it) you should be able to go to the privacy settings for the file ("Share" tab in upper right-hand corner of screen) and change it from "Private" to "Public" (just make sure you also set it such that it's "View-only access"). Then you can just post the link to that page here and all of us can view it.

To be honest I'm definitely much of a computer guy

But I'll e-mail it to you or something and you can put it online. When I add a list of 10 players I'll send it again, if you want to keep it going of course (might lose interest for you). I'll still update the thread here.

[Joao Saraiva]

Changed the formula:
- little bit less weight on longevity and more on prime, like you guys suggested. Prime is now at 65%;
- Added DBPM;
- Accodales are now taken into account, but only on all time list (Rings, MVPs, top 10 MVP shares, Finals MVPs).

Introduced new players:
- Karl Malone, Kevin Garnett, Dirk Nowitzki and Charles Barkley.

[trex_8063]

Changed the formula:
- little bit less weight on longevity and more on prime, like you guys suggested. Prime is now at 65%;
- Added DBPM;
- Accodales are now taken into account, but only on all time list (Rings, MVPs, top 10 MVP shares, Finals MVPs).

Introduced new players:
- Karl Malone, Kevin Garnett, Dirk Nowitzki and Charles Barkley.

I look forward to reviewing the changes. I'll get you my email at another date; I'm leaving town in about 36 hours and sort of going "off the grid" for about a week and a half. Will get you my email and review the formula when I get back.

May be a moot point now, but wrt my previously voiced concerns over not incorporating mpg with PER and WS/48 (per game averages partially compensates, yes), and also too much emphasis on TS% (with little recognition of context)......I came up with an example or two to better illustrate what I felt were the formula short-comings......

By your previous incarnation of the formula (before you added accolades and DBPM, etc), and without factoring in the games missed coefficient (let's just assume equal # of games played), '14 Marc Gasol scores out at 98.55 for the rs; his teammate '14 Zach Randolph gets 95.10. By comparison, '14 Chris Andersen (playing barely over 19 mpg) scores a 92.95.

But do we really feel Chris Andersen was that close to Randolph or Gasol that year?

Anyway, looking forward to seeing the changes and reviewing the formula, but it will have to wait a little while. Until then....

[Joao Saraiva]

Changed the formula:
- little bit less weight on longevity and more on prime, like you guys suggested. Prime is now at 65%;
- Added DBPM;
- Accodales are now taken into account, but only on all time list (Rings, MVPs, top 10 MVP shares, Finals MVPs).

Introduced new players:
- Karl Malone, Kevin Garnett, Dirk Nowitzki and Charles Barkley.

I look forward to reviewing the changes. I'll get you my email at another date; I'm leaving town in about 36 hours and sort of going "off the grid" for about a week and a half. Will get you my email and review the formula when I get back.

May be a moot point now, but wrt my previously voiced concerns over not incorporating mpg with PER and WS/48 (per game averages partially compensates, yes), and also too much emphasis on TS% (with little recognition of context)......I came up with an example or two to better illustrate what I felt were the formula short-comings......

By your previous incarnation of the formula (before you added accolades and DBPM, etc), and without factoring in the games missed coefficient (let's just assume equal # of games played), '14 Marc Gasol scores out at 98.55 for the rs; his teammate '14 Zach Randolph gets 95.10. By comparison, '14 Chris Andersen (playing barely over 19 mpg) scores a 92.95.

But do we really feel Chris Andersen was that close to Randolph or Gasol that year?

Anyway, looking forward to seeing the changes and reviewing the formula, but it will have to wait a little while. Until then....

Well, maybe I'll try inserting some role players here later too. Right now I'm foccusing more on the top players. But yeah, like I said it's possible that the formula is not always correct. Any formula usually produces some awkward numbers.

[Joao Saraiva]

Next palyers to be on the lists: John Stockton, Dwyane Wade and Chris Paul!

[Joao Saraiva]

Allen Iverson included.

[2klegend]

Can you show me how you get the number for peak? What do you mean by Rounds played coefficient?

[Joao Saraiva]

Can you show me how you get the number for peak? What do you mean by Rounds played coefficient?

If a player goes to the finals, I divide his production by 1.
If he loses in the conference finals, it divides his production by 1,1.
In the 2nd round, by 1,2.
In the 1st round by 1,3.

It's not the same thing if you put up great performances for an entire playoff or just in one round.

I opted for not dividing for a bigger coefficient because many times the player went as far as he could and it's not his fault.

McGrady 03, Hakeem 88, Kobe 06, LeBron 09, Jordan 90, etc. are good examples of that.

[Joao Saraiva]

I'm not going to include stuff like MVP awards, since not every MVP is equal or even deserved.

Could include top-5 MVP finishes, though. The voters usually don't screw THAT up very often.

I'm including top 10 MVP finishes. That's a cool idea, thanks.

It values longer primes, since in the formula the prime is defined as the best 5 years, and even if the other years are counting to the all time formula they're not counting as much. But giving an extra value for each place on a top 10 finish (20 pts MVP, 18 pts 2nd place, 16 pts 3rd place, etc.) made a nice impact!

[Cycklops]

Without looking too much at the formula, your list is excellent and very interesting.

[2klegend]

Can you show me how you get the number for peak? What do you mean by Rounds played coefficient?

If a player goes to the finals, I divide his production by 1.
If he loses in the conference finals, it divides his production by 1,1.
In the 2nd round, by 1,2.
In the 1st round by 1,3.

It's not the same thing if you put up great performances for an entire playoff or just in one round.

I opted for not dividing for a bigger coefficient because many times the player went as far as he could and it's not his fault.

McGrady 03, Hakeem 88, Kobe 06, LeBron 09, Jordan 90, etc. are good examples of that.

I'm still not getting the right number on CP3 peak.
This is what I enter...
Regular = (21.1+4*2+11.6*2+2.7*2+0.1*2-2.5*2+28.3+0.284+(0.576/2)+0.9)/(2)
Postseason = (24.1+4.9*2+11.3*2+2.3*2+0.2*2-1.8*2+30.7+0.289+(0.565/2)+2.5)/(1.2)

Can you tell me what I did wrong?

[Lucky707]

I really do think this overrates Karl Malone somewhat... Bird has a better peak and prime according to your formula and is still behind him.

[JustNBATingz]

Can you please include Tracy McGrady

[Joao Saraiva]

Can you show me how you get the number for peak? What do you mean by Rounds played coefficient?

If a player goes to the finals, I divide his production by 1.
If he loses in the conference finals, it divides his production by 1,1.
In the 2nd round, by 1,2.
In the 1st round by 1,3.

It's not the same thing if you put up great performances for an entire playoff or just in one round.

I opted for not dividing for a bigger coefficient because many times the player went as far as he could and it's not his fault.

McGrady 03, Hakeem 88, Kobe 06, LeBron 09, Jordan 90, etc. are good examples of that.

I'm still not getting the right number on CP3 peak.
This is what I enter...
Regular = (21.1+4*2+11.6*2+2.7*2+0.1*2-2.5*2+28.3+0.284+(0.576/2)+0.9)/(2)
Postseason = (24.1+4.9*2+11.3*2+2.3*2+0.2*2-1.8*2+30.7+0.289+(0.565/2)+2.5)/(1.2)

Can you tell me what I did wrong?

I don't use those type of numbers in PER, WS/48 and ts%.

PER 28.3
WS/48 = 28.4
TS% = 57.6

Those numbers would be too small to have an impact.

Plus the gmissed value is a coefficient. If the player missed less than 10 games, the coefficient is 1. It means I won't take any value away from him based on that.

Hope I made it clear. If you didn't understand tell me.