PatsSoxKnicks

07-13-2011, 06:45 PM

I had originally posted this in the #8 thread but I know Swash had suggested that I post this in the NBA Stats forum. I'd be open to suggestions. In fact, I already have a few minor tweaks in mind that I'm planning on implementing.

Edit: I just realized, I really need to thank basketball reference. They have such a great site and a lot of these statistical debates wouldn't be possible without bball-ref.

Also, probably need to cite this article, since this is Pelton's method:

http://basketballprospectus.com/article.php?articleid=415

Hey Swash, so I've completed my analysis using Win Shares. I'll get right down to it and explain what I did.

I mostly used Pelton's method but adjusted it a bit to weight the playoffs more.

Anyways, first, I found the career total win shares for each of the 5 players on the list. Next, I found the best 3 seasons for each player, regardless of year.

For example, doing this for Duncan you get 17.8 WS in 01-02, 16.5 in 02-03 and 13.2 in 00-01 and the total is 47.5. The fact that they came in 3 consecutive years is irrelevant. For some of the other players, that wasn't the case.

After, I calculated the best 5 consecutive seasons for each player. This is basically the maximum total of any 5 years in a row. It doesn't necessarily have to be ages 25-30 but I would imagine most of the time, those seasons fall in that age area.

This gave me the 3 regular season components.

After that, I looked at their career playoff numbers (Simply looking at any 1 season is going to skew the results because of the limited sample size of playoff games). I took their career WS in the playoffs and divided that by games played. I could've used WS/48 but I wanted minutes played to be factored in. So anyways, with career WS / games played, you have Win Shares per game. Then I multiplied that by 82 to get essentially a season worth of playoff games.

The following is the results of all these calculations:

career best 3 5 consc playoff WS

Hakeem 162.8 42.5 60.3 12.78

Barkley 177.2 50.1 75.8 13.00

Duncan 170.0 47.5 73.6 13.42

Kobe 156.3 44.0 64.2 10.57

KG 174.7 50.0 77.7 10.00

However, these are all uneven totals and their needs to be some form of weights applied.

So I divided the career total by 10 (yes, its an arbitrary # but it produces a comparable result to the other numbers in the table). I divided the best 3 seasons by 3 and the 5 consecutive best seasons by 5.

This gets you comparable #'s, although slightly different. I then added those 3 numbers up (career/10, best 3/3 and 5 consc/5) and divided by 3 to find the average. This number is the regular season total.

Then it's simply about deciding how much weight you want to place on the playoffs vs. the regular season. Obviously, you have to keep in mind that in the regular season, you have a much larger sample size. At the same time, the playoffs are the ultimate test and show how you do under pressure and against good teams.

I weighted it 40% regular season and 60% playoffs. This was done by

reg season*.4 + playoff WS*.6 = Overall

Here were the results:

career best 3 5 consc reg season Overall

Hakeem 16.28 14.17 12.06 14.17 13.34

Barkley 17.72 16.70 15.16 16.53 14.41

Duncan 17.00 15.83 14.72 15.85 14.39

Kobe 15.63 14.67 12.84 14.38 12.09

KG 17.47 16.67 15.54 16.56 12.62

It's interesting to see Barkley on top, although barely above Duncan. But if you fudge around with how much weight you give to the playoffs- maybe instead of 60-40, you do 70-30- than Duncan will come out on top. It's easy enough to do if you want to mess around with the weights.

There could be some issues with my whole method here. For example, it'd probably make it more accurate if I factored in the % of performance loss from the regular season. KG would be the biggest loser in that regard. And then as Chronz said, if I included 7 seasons and 9 seasons, it could change the entire results around. Another issue is that I've taken career playoff #'s and extrapolated them out, instead of taking some individual playoff runs and giving them any weight. Personally, I don't like the idea of doing that since its such a small sample size.

Anyways, enjoy. Let me know your thoughts. For me personally, I enjoyed doing this and it certainly makes me feel better about picking Duncan here.

I do want to do this out for EWA (PER's cousin) and WARP to see if anything is different.

And Chronz, I'd love it if you weighed in on what I did here.

PS- This only furthers my opinion that Barkley is one of the most underrated players ever. If he had a championship, people's opinion of him would be drastically changed. And thats a shame that a team accomplishment is holding back Chuck in most people's eyes.

I'll continue to do this out for some of the all-time greats. I've already done it for MJ, Shaq and Magic.

In fact, it was when I was doing this out for Shaq that I noticed some few things I'd like to change. Shaq's playoff numbers get watered down from his older days and unfortunately, that gets reflected in his Win Share numbers.

So I plan on rectifying that my talking each players 3 best playoff runs with 500+ minutes and weighting it at 10% each. So the playoff numbers will end up being 70% career playoff numbers and 30% 3 best playoff runs.

Edit: I just realized, I really need to thank basketball reference. They have such a great site and a lot of these statistical debates wouldn't be possible without bball-ref.

Also, probably need to cite this article, since this is Pelton's method:

http://basketballprospectus.com/article.php?articleid=415

Hey Swash, so I've completed my analysis using Win Shares. I'll get right down to it and explain what I did.

I mostly used Pelton's method but adjusted it a bit to weight the playoffs more.

Anyways, first, I found the career total win shares for each of the 5 players on the list. Next, I found the best 3 seasons for each player, regardless of year.

For example, doing this for Duncan you get 17.8 WS in 01-02, 16.5 in 02-03 and 13.2 in 00-01 and the total is 47.5. The fact that they came in 3 consecutive years is irrelevant. For some of the other players, that wasn't the case.

After, I calculated the best 5 consecutive seasons for each player. This is basically the maximum total of any 5 years in a row. It doesn't necessarily have to be ages 25-30 but I would imagine most of the time, those seasons fall in that age area.

This gave me the 3 regular season components.

After that, I looked at their career playoff numbers (Simply looking at any 1 season is going to skew the results because of the limited sample size of playoff games). I took their career WS in the playoffs and divided that by games played. I could've used WS/48 but I wanted minutes played to be factored in. So anyways, with career WS / games played, you have Win Shares per game. Then I multiplied that by 82 to get essentially a season worth of playoff games.

The following is the results of all these calculations:

career best 3 5 consc playoff WS

Hakeem 162.8 42.5 60.3 12.78

Barkley 177.2 50.1 75.8 13.00

Duncan 170.0 47.5 73.6 13.42

Kobe 156.3 44.0 64.2 10.57

KG 174.7 50.0 77.7 10.00

However, these are all uneven totals and their needs to be some form of weights applied.

So I divided the career total by 10 (yes, its an arbitrary # but it produces a comparable result to the other numbers in the table). I divided the best 3 seasons by 3 and the 5 consecutive best seasons by 5.

This gets you comparable #'s, although slightly different. I then added those 3 numbers up (career/10, best 3/3 and 5 consc/5) and divided by 3 to find the average. This number is the regular season total.

Then it's simply about deciding how much weight you want to place on the playoffs vs. the regular season. Obviously, you have to keep in mind that in the regular season, you have a much larger sample size. At the same time, the playoffs are the ultimate test and show how you do under pressure and against good teams.

I weighted it 40% regular season and 60% playoffs. This was done by

reg season*.4 + playoff WS*.6 = Overall

Here were the results:

career best 3 5 consc reg season Overall

Hakeem 16.28 14.17 12.06 14.17 13.34

Barkley 17.72 16.70 15.16 16.53 14.41

Duncan 17.00 15.83 14.72 15.85 14.39

Kobe 15.63 14.67 12.84 14.38 12.09

KG 17.47 16.67 15.54 16.56 12.62

It's interesting to see Barkley on top, although barely above Duncan. But if you fudge around with how much weight you give to the playoffs- maybe instead of 60-40, you do 70-30- than Duncan will come out on top. It's easy enough to do if you want to mess around with the weights.

There could be some issues with my whole method here. For example, it'd probably make it more accurate if I factored in the % of performance loss from the regular season. KG would be the biggest loser in that regard. And then as Chronz said, if I included 7 seasons and 9 seasons, it could change the entire results around. Another issue is that I've taken career playoff #'s and extrapolated them out, instead of taking some individual playoff runs and giving them any weight. Personally, I don't like the idea of doing that since its such a small sample size.

Anyways, enjoy. Let me know your thoughts. For me personally, I enjoyed doing this and it certainly makes me feel better about picking Duncan here.

I do want to do this out for EWA (PER's cousin) and WARP to see if anything is different.

And Chronz, I'd love it if you weighed in on what I did here.

PS- This only furthers my opinion that Barkley is one of the most underrated players ever. If he had a championship, people's opinion of him would be drastically changed. And thats a shame that a team accomplishment is holding back Chuck in most people's eyes.

I'll continue to do this out for some of the all-time greats. I've already done it for MJ, Shaq and Magic.

In fact, it was when I was doing this out for Shaq that I noticed some few things I'd like to change. Shaq's playoff numbers get watered down from his older days and unfortunately, that gets reflected in his Win Share numbers.

So I plan on rectifying that my talking each players 3 best playoff runs with 500+ minutes and weighting it at 10% each. So the playoff numbers will end up being 70% career playoff numbers and 30% 3 best playoff runs.