Retro Player of the Year Project

[sp6r=underrated]

I have another suggestion. I'm not sure what year you are starting with, but I would suggest alternating from beginning to end. If 1960 was your starting year, I would suggest this type of order

1960
2010
1961
2009
1962
2008

This will help keep interest in the project going by alternating between the present and past and force voters to vote and learn about both eras if they want to participate.

[Silver Bullet]

No reason why we can't have two threads going at the same time. We could start from both ends 1960 and 2009 and end up somewhere in the middle.

[bastillon]

scottiepippen6=underrated with a great idea, as always. mad props
one topic should be enough though. I'd suggest starting from shotclock era. the year Neil Jonhston led the Warriors to the title (or smth).

[Doctor MJ]

I really like this project.

I do think we need to include some more parameters than just top 5 if we are going to use the results in an all time ranking type setting though. For example; one player could potentially have had the 6th best season for 15 years straight and end up with 0 points in this project, while someone who only had one good season where he was 5th will be ranked higher.

However, for purposes of measuring year by year who had the best seasons, the format works just fine IMO.

Ah, you are right, I just don't see anyway to do a project like this to get everything perfect for an all-time ranking. I'll listen to ideas on the matter, but I worry about making this too complex, and still not accomplishing the more ambitious goal.

[Doctor MJ]

I have another suggestion. I'm not sure what year you are starting with, but I would suggest alternating from beginning to end. If 1960 was your starting year, I would suggest this type of order

1960
2010
1961
2009
1962
2008

This will help keep interest in the project going by alternating between the present and past and force voters to vote and learn about both eras if they want to participate.

Interesting. I've got no objection to this. If others don't object, we'll do it this way - though I'd probably start with the first year of MVP voting (1956) - and go back and do the earlier years afterward if interest was still keen.

[TrueLAfan]

tkb--that's actually pretty much never going to happen. As long as you've got at least 15-20 voters, you're going to find players broken into distinct groupings. There will 2-4 real MVP candidates (or one clear winner, and 3-4 players in the next group. Then there will be 3-5 players in the next group. After that, you've got stragglers...players that get an occasional vote.

Basically it works like this. About 75-80% of voters will agree on 4 of the top 5 players. That's the top 4. A couple of other players will be considered with that top 4, and be ranked with some arrange of them. About 5-8 other players will get votes at number and an occasional #4 or #3. Those top 5-8 players will get 99% of the points. That's because there will largely be consensus about who is elite and near elite in a season. It's why MVP voting is largely terrific. People complain that a player "didn't win" or "should be higher." What MVP voting shows is who is elite and who is near elite and who is just a bit short of those levels but still terrific. It's almost impossible for the 6th best player not to get significant recognition--and if you do it for a few years, you're going to work out as a pretty elite player. If you'd been 6th in MVP voting for the last 5 years, you'd have .269 of an MVP win share. That's doesn't sound like much, but that' would be 65th of all time. That's pretty huge.

If we have 30-40 voters for this project, I think a lot of people will be surprised at both how clear cut the voting divisions are and how few players actually get votes. I'll be surprised if 40 voters have 20 players getting even one fifth place vote; I think we'll have something like that half that number most years

About changes in the original premise. I think we're carried away from what was a good, simple, fun idea. There's often complaints about the MVP voting because it takes place before the playoffs, and because non-voters think the voters got it wildly wrong. This idea solves both issues in a neat, uncomplicated way. I don't mind jumping between years in the selected era range...but I think people don't realize how complicated these things are to do (they are) and how long it's going to take. If we get three selections a week--72 hours for discussion and voting, with a day of overlap so we begin voting on Monday, Wednesday, and Saturday--covering 50 years would take 4 months. We're a lot better off getting interest and committed voters ASAP. I personally think it's much easier and probably better in the long run to start in 2010 and go back from there, but I'll go with whatever Doctor MJ says.

[bastillon]

I don't mind jumping between years in the selected era range...but I think people don't realize how complicated these things are to do (they are) and how long it's going to take. If we get three selections a week--72 hours for discussion and voting, with a day of overlap so we begin voting on Monday, Wednesday, and Saturday--covering 50 years would take 4 months. We're a lot better off getting interest and committed voters ASAP. I personally think it's much easier and probably better in the long run to start in 2010 and go back from there, but I'll go with whatever Doctor MJ says.

that's a great point. I don't know how many posters would be commited to 1950s votings.

[tkb]

I didn't say that we won't find the correct players for each vote. I do think we will, and that's why I stated that this format works just fine for the purposes of ranking POY (top 5). It will also work well to rank a few players in an all time setting, but when we get a bit down the list this system won't really work for that IMO. You even stated the reason for why it won't work for those players in your own post TLA when you said people will be surprised at how few players will receive votes each year.

There could be 1 million voters who all agreed on all votes for all I care, and we would still not take into account what I was trying to say, which is that a player could have had a great, great career but never have been a top 5 player and be rated lower than a 1-year wonder. If everyone agrees that he was a great player, but wasn't top 5 any season for whatever reason, he will end up with 0 points in this format.

John Stockton is an example of a great player I don't think will get a lot of points in a format like this if we use the results to rank the players in an all time setting, largely because I don't think he had a lot of top 5 seasons (if any). Doesn't mean he wasn't a great player though, and without question top 50 IMO.

[Doctor MJ]

I didn't say that we won't find the correct players for each vote. I do think we will, and that's why I stated that this format works just fine for the purposes of ranking POY (top 5). It will also work well to rank a few players in an all time setting, but when we get a bit down the list this system won't really work for that IMO. You even stated the reason for why it won't work for those players in your own post TLA when you said people will be surprised at how few players will receive votes each year.

There could be 1 million voters who all agreed on all votes for all I care, and we would still not take into account what I was trying to say, which is that a player could have had a great, great career but never have been a top 5 player and be rated lower than a 1-year wonder. If everyone agrees that he was a great player, but wasn't top 5 any season for whatever reason, he will end up with 0 points in this format.

John Stockton is an example of a great player I don't think will get a lot of points in a format like this if we use the results to rank the players in an all time setting, largely because I don't think he had a lot of top 5 seasons (if any). Doesn't mean he wasn't a great player though, and without question top 50 IMO.

I don't think there's any way a project like this can address the Sidekick problem, which I'll say is: How much credit do you give the guy who works can adapt his game around another star? There's also no way it can deal with the fact that it's not equally difficult in all years to get MVP/POY shares. I see this as something that will inform a Top 100 project (which penbeast has indicated he plans to run after this is done), rather than replacing it.

The specific "Stockton problem" (great longevity, but not a top 5 guy) is something that we could try to deal with, but given the 2 limitations above, I don't think it will really give people satisfaction. To logic it out a little bit: We're using a 10-7-5-3-1 system like the MVP voting uses. To keep with the spirit of that kind of weighting, placing any lower down has to count on the order of a point or less (so 0.1 shares or less). When Nash got his MVPs he averaged about 0.8 shares each year. You can see that there's basically nothing we can do while keeping in that spirit that would make Stockton's career the equal of Nash's - which many of us think is totally wrong. (The fact that this might not end up being the case in this project I don't think should be used in rebuttal - the scenario where people rate Player A over a Player B with high peak but shorter longevity is legit)

Now you may say: So change the weighting! I'm afraid I'm going to veto any severe change to the weighting. I've got a lot of experience designing algorithms, we're not only not going to come up with anything that satisfies everyone...we're not going to come up with anything that totally satisfies ANYONE. If we were to adjust the weights to make the Nash-Stockton comparison "feel" right, then we'd find that some other comparison starts to feel totally wrong.

I've mentioned before that I have a metric that's like what we're going to do here, but more elaborate. Every time I mention it I try to make clear that I'd never give my Top 100 votes just by going down the results of the metric. I think a really solid understanding of the great players of all time requires you to go through year-by-year analysis, ideally along with some other knowledgeable people, but in the end you have to get a bit more holistic to maximize the common sense.

[penbeast0]

Since when is this about career value or even how good a player actually is? It's a year by year snapshot -- making it something else would be a different project.

[TrueLAfan]

^^ Yup.

If you want to find career player value, MVP shares aren't the end-all, be-all. Some sort of sliding adjustment (with current voting, something five hundredths of a win share for every year from 5-9 you get an MVP vote; a tenth of a win share for every year from 10-14; .15 for every year after 14 seems like it does it) can be used. When determining player greatness, peak trumps career...but there's got to be some mix of career valuation.*

But, unless I'm mistaken, that's not the point of this and shouldn't be. This is to provide a post-playoff assessment of player value in past seasons...a year-by-year snapshot, as penbeast says. That's a nice, doable project that people will (hopefully) be into.

*Side note. tkb is right in saying that if a player "wasn't top 5 any season for whatever reason, he will end up with 0 points in this format." But that's if everyone agrees the person was never top 5...and that just rarely happens. When I say "few players will receive votes," I mean around 12-16 players (most years) will get votes, and about a third of those will receive a solitary fifth place vote...which means a player is more in the top 8-16 players in the league than the top 5. So it's more accurate to say if a player isn't top 5, but is somewhere between 10 and 20, he's about 50-50 to receive a single MVP point.

[Doctor MJ]

I will say something I am very curious about is how similar the standard deviation among this group will be to MVP voters. It's possible we'll have significantly more agreement than MVP voters do - no way to know until we do it.

[ElGee]

I didn't say that we won't find the correct players for each vote. I do think we will, and that's why I stated that this format works just fine for the purposes of ranking POY (top 5). It will also work well to rank a few players in an all time setting, but when we get a bit down the list this system won't really work for that IMO. You even stated the reason for why it won't work for those players in your own post TLA when you said people will be surprised at how few players will receive votes each year.

There could be 1 million voters who all agreed on all votes for all I care, and we would still not take into account what I was trying to say, which is that a player could have had a great, great career but never have been a top 5 player and be rated lower than a 1-year wonder. If everyone agrees that he was a great player, but wasn't top 5 any season for whatever reason, he will end up with 0 points in this format.

John Stockton is an example of a great player I don't think will get a lot of points in a format like this if we use the results to rank the players in an all time setting, largely because I don't think he had a lot of top 5 seasons (if any). Doesn't mean he wasn't a great player though, and without question top 50 IMO.

I don't think there's any way a project like this can address the Sidekick problem, which I'll say is: How much credit do you give the guy who works can adapt his game around another star? There's also no way it can deal with the fact that it's not equally difficult in all years to get MVP/POY shares. I see this as something that will inform a Top 100 project (which penbeast has indicated he plans to run after this is done), rather than replacing it.

I have a different project that would handle the Sidekick problem (in a way it's designed to eliminate the sidekick problem). I'm not overly keen on top-x ranking in a vacuum, because that's not entirely representative of how basketball dynamics, or building a team, works.

I'll summarize the idea here if anyone is interested, but obviously it has nothing to do with year-by-year POY:

Instead of a top-100, I've wondered how the results would turn out if you designated a certain number of teams (2 -- teams build just for head-to-head -- or 4+ -- versatility required) and staged a draft with the panelists acting as GM's. The goal, of course, being to build the best team. In fact, running multiple drafts and altering the order, number of teams, or panelists would then provide a fairly interesting player ranking from the average draft positions. This allows for the conditional nature of basketball (wouldn't want 8 centers on the same team), maximizes role player value, eliminates the Sidekick problem, etc. And the discussion/arguments about what direction to draft (size, speed, defense, offense, etc.) would be great.

[Doctor MJ]

I don't think there's any way a project like this can address the Sidekick problem, which I'll say is: How much credit do you give the guy who works can adapt his game around another star? There's also no way it can deal with the fact that it's not equally difficult in all years to get MVP/POY shares. I see this as something that will inform a Top 100 project (which penbeast has indicated he plans to run after this is done), rather than replacing it.

I have a different project that would handle the Sidekick problem (in a way it's designed to eliminate the sidekick problem). I'm not overly keen on top-x ranking in a vacuum, because that's not entirely representative of how basketball dynamics, or building a team, works.

I'll summarize the idea here if anyone is interested, but obviously it has nothing to do with year-by-year POY:

Instead of a top-100, I've wondered how the results would turn out if you designated a certain number of teams (2 -- teams build just for head-to-head -- or 4+ -- versatility required) and staged a draft with the panelists acting as GM's. The goal, of course, being to build the best team. In fact, running multiple drafts and altering the order, number of teams, or panelists would then provide a fairly interesting player ranking from the average draft positions. This allows for the conditional nature of basketball (wouldn't want 8 centers on the same team), maximizes role player value, eliminates the Sidekick problem, etc. And the discussion/arguments about what direction to draft (size, speed, defense, offense, etc.) would be great.

You're right that that is something that informs the Sidekick problem. There have certainly been some All-Time Leagues done along those lines. TrueLA and I ran the first one, tkb ran the second - and then I lost track of where that was (anyone else know?). It would be interesting to see the stats on that.

There are some limitations though. The first being that you're not going to be able to do enough of them to really average the different permutations out. Probably the biggest though is how to extrapolate the information. The second ATL my top 3 draft picks were 1) Pippen, 2) Gilmore, 3) Barry, despite the fact that I rank Barry highest of the 3, because I knew I could get someone of Barry's quality in the 3rd round, and Pippen would be gone in the second. Does that mean Pippen should actually be ranked ahead of Barry? Well, in the real NBA, the level of quality available is never that high so Pippen won't get used as a Sidekick in many circumstances. Seems kinda crazy to boost him too much for a quality that won't always be able to use, y'know?

[ElGee]

I don't think there's any way a project like this can address the Sidekick problem, which I'll say is: How much credit do you give the guy who works can adapt his game around another star? There's also no way it can deal with the fact that it's not equally difficult in all years to get MVP/POY shares. I see this as something that will inform a Top 100 project (which penbeast has indicated he plans to run after this is done), rather than replacing it.

I have a different project that would handle the Sidekick problem (in a way it's designed to eliminate the sidekick problem). I'm not overly keen on top-x ranking in a vacuum, because that's not entirely representative of how basketball dynamics, or building a team, works.

I'll summarize the idea here if anyone is interested, but obviously it has nothing to do with year-by-year POY:

Instead of a top-100, I've wondered how the results would turn out if you designated a certain number of teams (2 -- teams build just for head-to-head -- or 4+ -- versatility required) and staged a draft with the panelists acting as GM's. The goal, of course, being to build the best team. In fact, running multiple drafts and altering the order, number of teams, or panelists would then provide a fairly interesting player ranking from the average draft positions. This allows for the conditional nature of basketball (wouldn't want 8 centers on the same team), maximizes role player value, eliminates the Sidekick problem, etc. And the discussion/arguments about what direction to draft (size, speed, defense, offense, etc.) would be great.

You're right that that is something that informs the Sidekick problem. There have certainly been some All-Time Leagues done along those lines. TrueLA and I ran the first one, tkb ran the second - and then I lost track of where that was (anyone else know?). It would be interesting to see the stats on that.

There are some limitations though. The first being that you're not going to be able to do enough of them to really average the different permutations out. Probably the biggest though is how to extrapolate the information. The second ATL my top 3 draft picks were 1) Pippen, 2) Gilmore, 3) Barry, despite the fact that I rank Barry highest of the 3, because I knew I could get someone of Barry's quality in the 3rd round, and Pippen would be gone in the second. Does that mean Pippen should actually be ranked ahead of Barry? Well, in the real NBA, the level of quality available is never that high so Pippen won't get used as a Sidekick in many circumstances. Seems kinda crazy to boost him too much for a quality that won't always be able to use, y'know?

Well, that's why # of teams comes into play (essentially, expansion changing the roles and player values on teams in real life). It's funny you mentioned Pippen, because everyone of those I've played with he goes very high. I like the idea because it showcases versatility and uniqueness. You can get a player comparable to Barry, but there are very few in NBA history like Pippen. To me, this doesn't show up in the rankings in a vacuum (maybe because whether it's #1 or #97, voters still ask themselves the same thought questions about "who to build around," "leading a team" etc.).

Do you have a link to the one you did?

[Optimism Prime]

As far as the alternating years goes, here's my thoughts:

I'm not sure what the average age of the participating posters is, but I'm 26. Obviously, I'm more familiar with the current era... I'd like some time to "warm up" as it were, learn the basis of what we're doing here, with players and seasons I know about. As we work our way back, I'll get more insight into the research aspects of it.

Again, I'm on board for whatever we end up doing, but I'd prefer the comfort of working into this.

[Doctor MJ]

Do you have a link to the one you did?

I believe they were lost with a software upgrade a while back. A google search turns up more recent stuff. Things I remember from them in case anyone's interested:

First one had 30 teams, the second had 16.
Wilt-MJ went 1-2 in the first one, and 2-1 in the second.
tkb won the first one, penbeast won the second - both had Shaq as the best player, which made me think that maybe subconsciously we're all a little more impressed with Shaq's overwhelmingness than we even realize.
In the 30 team won I had the #2 pick (Jordan), by the time the #59 came around, Parish was the best big left - still hurts.

[Doctor MJ]

As far as the alternating years goes, here's my thoughts:

I'm not sure what the average age of the participating posters is, but I'm 26. Obviously, I'm more familiar with the current era... I'd like some time to "warm up" as it were, learn the basis of what we're doing here, with players and seasons I know about. As we work our way back, I'll get more insight into the research aspects of it.

Again, I'm on board for whatever we end up doing, but I'd prefer the comfort of working into this.

Yeah, that was my thought as well.

Kinda thinking that's how we should start, and then we can decide to alternate later. I dunno, I don't feel that strongly about it as long as we definitely get some recent years in their first (sure as hell don't want the project to die having done years 56 to 60 only ) If others come back and fiercely dislike this plan, I'll go with the flow.

[bastillon]

Do you have a link to the one you did?

I believe they were lost with a software upgrade a while back. A google search turns up more recent stuff. Things I remember from them in case anyone's interested:

First one had 30 teams, the second had 16.
Wilt-MJ went 1-2 in the first one, and 2-1 in the second.
tkb won the first one, penbeast won the second - both had Shaq as the best player, which made me think that maybe subconsciously we're all a little more impressed with Shaq's overwhelmingness than we even realize.
In the 30 team won I had the #2 pick (Jordan), by the time the #59 came around, Parish was the best big left - still hurts.

what I noticed playing in the ATL is that people barely pay attention to spacing and efficiency. Barry and Havlicek for example are considered as great scoring threats, despite the fact that their efficiency was sub-par during their era and their game wouldn't translate very well today (low FTA, lots of midrange jumpers). spacing is entirely different problem, I've seen knowledgeable posters pick a team of KAJ and Thurmond for example, nevermind the fact that Thurmond failed alongside Wilt, despite the gigantic gap in talent vs their contemporaries.

Russell is disrespected in the ATL, he doesn't go very high.

Olajuwon goes way higher than #7-10, more like ~#4. what I noticed about Hakeem is that you can't really defend him with anyone, because he has proven he can dominate great defenses ('86 C's, '94 Knicks, '95 WCFs).

Shaq is thought of as the dominant force too, so nobody tries to imply that he'll stop Shaq. I actually made a research and Shaq wasn't very succesful h2h vs great centers. Ewing, Hakeem, Yao, Mourning, Robinson, Mutombo... they all limited him when they were still young. I'll post those numbers sometime in the future. maybe that'll change someone's opinion just a little bit. I know it changed mine.

also, role players are terribly disrespected. recently I've learnt that Shane Battier was the very worst starter in a league that had Spree (inefficient chucker without 3P range) and many other not-really-good-for-ATL players. Battier is great IMO because he can space the floor and play amazing defense, but most people will look at his stats and think he's below average NBA player, let alone ATL player. I remember when SabasRevenge assembled a great team in the ATL and someone actually criticized him for playing Cooper in the starting line-up, which I thought was ridiculous.

I could go on and on, but I think guys like penbeast0 or Warspite could give you better feedback on ATLs.

[Sedale Threatt]

Since when is this about career value or even how good a player actually is? It's a year by year snapshot -- making it something else would be a different project.

This is what I signed up for -- doing a year-by-year vote on who the five best players were. Not interested at all with anything else -- career value/rankings, any sort of All-Time League or whatever. Just who the five best guys were in any given season. That's all.