Survey Project: Wins Added by Season

[SideshowBob]

Robinson/Hakeem/Duncan/KG +4.0
B. Wallace/Mutombo +3.5
Howard/LBJ/Pippen/Shaq/Mourning/Ewing +3.0
Wade/Kidd/Payton/Kemp/Moncrief/Parish/Erving/Rodman/Jordan +2.0

Regarding Wade, does this mean that you've upped your view on him a bit? I believe you had him around +5.5 on offense, so is he now the 14th member of the "Sacred Peaks?" I ask because evaluating him myself, I feel like he could be argued in, and I see that you had him 2 spots out in your November list anyway.

[ElGee]

OK guys, getting the threads going again. All those who have been waiting to take a go at the project, feel free to start with the new threads as they open or go back and start at the top of the list. Look forward to it.

Robinson/Hakeem/Duncan/KG +4.0
B. Wallace/Mutombo +3.5
Howard/LBJ/Pippen/Shaq/Mourning/Ewing +3.0
Wade/Kidd/Payton/Kemp/Moncrief/Parish/Erving/Rodman/Jordan +2.0

Regarding Wade, does this mean that you've upped your view on him a bit? I believe you had him around +5.5 on offense, so is he now the 14th member of the "Sacred Peaks?" I ask because evaluating him myself, I feel like he could be argued in, and I see that you had him 2 spots out in your November list anyway.

No, I don't think his offensive and defensive peak were in the same season.

[Narigo]

Re-did my Wilt and Jordan numbers in their threads

[ElGee]

Re-did my Wilt and Jordan numbers in their threads

Thanks. Once we get to about 20 people filling out ballots I will start to publish some of the results.

[nikomCH]

I have a question about how you would go about comparing players to each other after coming up with a value for each season (using SRS). Do you just add up the numbers and whoever has the higher sum would, by this system, have been the better player? Or would you need to adjust that summed up number in another way before making any comparisons?

[ceiling raiser]

I have a question about how you would go about comparing players to each other after coming up with a value for each season (using SRS). Do you just add up the numbers and whoever has the higher sum would, by this system, have been the better player? Or would you need to adjust that summed up number in another way before making any comparisons?

I'm not 100% sure of the mechanics of the following since I haven't done too much research, but I think a straight comparison is problematic based on how SRS works. Both of these relationships:

impact -> SRS
SRS -> wins

are nonlinear, so summing up the SRS numbers could be trouble. I think certain levels of performance are more valuable, so you have three options if you want to go a summation route:

1) Weight higher numbers somehow (on a simple level, you could just square, cube, etc.)
2) Decide on a cutoff, and consider only seasons above that cutoff when evaluating players (maybe this is 0SRS, 2SRS, 5SRS, 8SRS, etc.).
3) Decide on a prime period, and only consider those seasons for a given player (subjective # of years or set per player).

If you go with (2) or (3) you might want to apply (1) afterwards. I think by doing so you're losing a lot of information though, so IMO it's not worth it.

[ceiling raiser]

Finally all caught up on the threads! They're still rough drafts though (and I did them over the course of a couple of months, when I had free time) so I might need to revise the first few to be in line with recent attempts. I'll probably do that after a few more threads are open.

Either way, very interesting/difficult/fun/educational project. I'm looking forward to evaluating more players when the threads are opened.

[SideshowBob]

I've been quite busy lately, but I do intend to finish this up as well.

I have a question about how you would go about comparing players to each other after coming up with a value for each season (using SRS). Do you just add up the numbers and whoever has the higher sum would, by this system, have been the better player? Or would you need to adjust that summed up number in another way before making any comparisons?

There were championship odds assigned to each SRS level (increments of 0.5, up to +10.0), divided into three separate groups for portability. All this was listed here, but Elgee mentioned that he had updated his data, but till now has only provided an update for Normal portability players.

I found the old odds here, and that's what I've personally been working with for now. Keep in mind that some sort of health adjustment must be made as well.

[SideshowBob]

fplii, just an observation , I see that you have a much larger spread for defensive ratings, even for modern guys. I've typically followed conventional wisdom regarding the cap on defensive impact in the 3 point era. I don't really have anyone from that time going past a +4 (defense only). Did your research/calculation method suggest differently?

[ceiling raiser]

fplii, just an observation , I see that you have a much larger spread for defensive ratings, even for modern guys. I've typically followed conventional wisdom regarding the cap on defensive impact in the 3 point era. I don't really have anyone from that time going past a +4 (defense only). Did your research/calculation method suggest differently?

I think my scale in general might have too wide a spread, I don't think it's realistic to have player seasons above 10 SRS as I do. Russ is the only guy who exceeds 4 defensive rating by a good amount due in part to his mobility (I think Thurmond and a guy like Reed will translate better as well), but I do have a few +5 (or slightly higher ) defensive seasons by other players who are primarily rim protectors or man post defenders. Part of the difference might be that I'm overweighting defensive rebounding (which I'm including in defensive ratings since it's too hard to parse out; same with offensive rebounding on the other side).

Even though we'll get to some more great defensive players (Hakeem, Robinson, Garnett, Pippen, Frazier, etc.) as the project continues, I think they're underrepresented compared to the great offensive players we'll cover. While it might be true that there are more great offensive players historically (or that they're at least more celebrated and discussed), it limits the pool of players up for evaluation here. I think everybody's going to be on the same page (mostly) for offensive ratings, but the other end is tougher. After I've worked on more players I'll post my splits. It actually would be nice to see defensive SRS points other posters are assigning to others.

Just wondering, into how many categories do you split bigs' defense, and what are they? For perimeter players it's simpler (man D, help D, bonus for versatility), but for bigs it's certainly more complex a task.

EDIT - Apologies, I'm uncertain as to whether you place the cap on defense in the 3pt era or before? I think there is probably a limit for both (for older players since abilities might not translate well to an era with more long range shooting, and for newer players since there are fewer outliers due to style of play).

[SideshowBob]

Updated all of my ballots with re-evaluations and filled out the one for Hakeem.

[ceiling raiser]

Updated all of my ballots with re-evaluations and filled out the one for Hakeem.

I just did the same. I tried to tighten up my values, but I probably could afford decrease the spread even further. I think some upcoming player surveys (KG, Robinson, Oscar in particular) will be useful in improving my scale, but Hakeem was a great help already so they feel much better.

[ceiling raiser]

My 2 cents? Role players matter A LOT. I've been working lately with a replacement player value of ~ -3, which is pretty standard in the industry. This basically means if we put together a team of guys who could barely make an NBA roster, we'd see something like a -15 team...which is what the worst teams in NBA history bump against.

Not to quote ElGee out of context, but I've decided that moving replacement player value from -2 -> -3 is the only way I can suitably condense my spread. It's out of control to have players at +10.0 I think. I'm not one for rigid thresholds, but +3.0 is probably a star, +6.0 a very strong MVP candidate, and +9.0 has to be a strong GOAT peak season. I don't think I should go above +9.0.

I'm also going to use ElGee's formulas for updating SRS based on portability more strictly going forward (I think someone quoted it as SRS+SIO×1.5^(1−e^|SRS/a|), where a=7 for low portability, 10 for avg portability, 15 for high portability (I do think there should maybe be some more spread, though; a=5 breaks the calculations so perhaps a=6 will suffice for low; I was playing around and got some nice results with a=27 for high portability).

Updated all of my ballots with re-evaluations and filled out the one for Hakeem.

Just wondering, in addition to the 11 we've done so far for the project, which other players have you evaluated?

[SideshowBob]

Just wondering, in addition to the 11 we've done so far for the project, which other players have you evaluated?

Oscar
West
Karl Malone
Garnett
Barkley
Robinson
Nash
Dirk
Wade
Bosh
Howard
Paul
Durant
Westbrook
Rose
Erving
McGrady


There's a handful of guys that I've evaluated in my head but never marked down (Pippen, Kidd, Walton come to immediate mind) . Still plenty to work on though. The guys that have been covered so far, I still think I can refine a bit, and the guys above I still have a lot of research to do. And even after all that it seems I'm still only scratching the surface of significant players over the last 50 years. Big work in progress.

[SideshowBob]

My 2 cents? Role players matter A LOT. I've been working lately with a replacement player value of ~ -3, which is pretty standard in the industry. This basically means if we put together a team of guys who could barely make an NBA roster, we'd see something like a -15 team...which is what the worst teams in NBA history bump against.

Not to quote ElGee out of context, but I've decided that moving replacement player value from -2 -> -3 is the only way I can suitably condense my spread. It's out of control to have players at +10.0 I think. I'm not one for rigid thresholds, but +3.0 is probably a star, +6.0 a very strong MVP candidate, and +9.0 has to be a strong GOAT peak season. I don't think I should go above +9.0.

I'm also going to use ElGee's formulas for updating SRS based on portability more strictly going forward (I think someone quoted it as SRS+SIO×1.5^(1−e^|SRS/a|), where a=7 for low portability, 10 for avg portability, 15 for high portability (I do think there should maybe be some more spread, though; a=5 breaks the calculations so perhaps a=6 will suffice for low; I was playing around and got some nice results with a=27 for high portability).

Updated all of my ballots with re-evaluations and filled out the one for Hakeem.

Just wondering, in addition to the 11 we've done so far for the project, which other players have you evaluated?

Regarding your thoughts on portability, I actually started messing around with it last year and worked with everything between a=7 and 15, but then sort of dropped it and I haven't looked too much into it since then. But I agree, there's room for larger spread and more denominations IMO.

[SideshowBob]

My general thought would be that a low impact/high portability player should be able to retain much more value on any level of team he is placed on vs a high impact/high portability player. Makes sense in practical terms, as this player is likely to be a specialist.

Example:

Player A is a +1.0 3 point specialist. He is probably going to look like a +1.0 guy anywhere he goes, as his role isn't going to be shifting at all.

On the other hand, Player B, let's call him Larry Bird (say he's a +6), isn't going to retain that +6 everywhere, even if he's classified as highly portable. On a strong team, his skillset will still show diminishing returns, but he has enough versatility that other skills will still shine through.

As opposed to Player C a highly ball dominant +6 player who cannot do much when the ball is not in his hands. Player B will not show close to 100% retention in every situation he is placed in, like we see in Player A, but he is still highly portable, especially compared to Player C, who in turn will not show anywhere near the retention that Player B does.

In that situation, Player A would have a very high a value, Player B's will be much lower but could still be classified as "High" and Player C will be middling.

[ceiling raiser]

My general thought would be that a low impact/high portability player should be able to retain much more value on any level of team he is placed on vs a high impact/high portability player. Makes sense in practical terms, as this player is likely to be a specialist.

Example:

Player A is a +1.0 3 point specialist. He is probably going to look like a +1.0 guy anywhere he goes, as his role isn't going to be shifting at all.

On the other hand, Player B, let's call him Larry Bird (say he's a +6), isn't going to retain that +6 everywhere, even if he's classified as highly portable. On a strong team, his skillset will still show diminishing returns, but he has enough versatility that other skills will still shine through.

As opposed to Player C a highly ball dominant +6 player who cannot do much when the ball is not in his hands. Player B will not show close to 100% retention in every situation he is placed in, like we see in Player A, but he is still highly portable, especially compared to Player C, who in turn will not show anywhere near the retention that Player B does.

In that situation, Player A would have a very high [b]a[b] value, Player B's will be much lower but could still be classified as "High" and Player C will be middling.

Interesting. I might be inclined to agree, I'd have to think about it more. I understand where you're coming from though and think it's a legitimate argument.

BTW what are you using for replacement level at the moment?

EDIT: Typo (we're instead of where). Need to sleep.

[SideshowBob]

I've tried to stick with Elgee's model thus far so I've gone with -3, which makes inherent sense to me as a worst-case scenario (the example you quoted).

[SideshowBob]

Unfortunately, I've lost all of my data. Most of it was just manipulations of stuff that is already online, but for this, I had actually come up with the ratings. Much of it is still listed in this project, but there were players we hadn't gotten to yet that I had evaluated and I had also made some changes to the players already covered. Recovering all of that is going to be my project for now.

[ceiling raiser]

I've tried to stick with Elgee's model thus far so I've gone with -3, which makes inherent sense to me as a worst-case scenario (the example you quoted).

Question - How many regular season games are you making playoff games worth (obviously there's no hard rule, but just for my files I'm trying to settle on a number)?

I feel they have to be worth a minimum of 2x, and a maximum of totality of regular season = entire playoffs. I think the former underweights them, but the latter is troublesome for short playoff runs.