Help with ranking the greatest players of all-time

[WillC]

I thought of an interesting way of ranking the greatest players of all-time...

I have collected the MVP voting results for every season between 1955-56 and present day. For the years before 1955, I've estimated the MVP voting based on the performance of the members of the All-League 1st and 2nd Team selections for each of those seasons (i.e. Joe Fulks was 'MVP' in 1946-47 and 1947-48, with Bob Feerick and Max Zaslofsky runners-up, respectively).

By assigning 10 points for a 1st place finish, 9 points for 2nd place, 8 points for 3rd place, etc, it allows me to give a final score for each player based on their MVP ranking each season across their entire career.

However, not surprisingly, the list is dominated by players like Kareem Abdul-Jabbar who had a long career but dominated during the relatively weak 1970s.

To get around this problem, I'd like to give a 'weighting' for each NBA season, i.e. so that the 1990s have a higher weighting than, say, the 1940s and 1950s (and to some extent the 1970s).

However, doing so is not easy. I don't want to arbitrarily assign a weighting to each season. I'd rather base it on something, like TV audience figures or someone else's rankings of the best seasons ever.

So, does anyone have sources that might help me? Or any other ideas?

Thank you.

[Nivek]

Ask MikeG over at APBRmetrics. He's done some work in measuring the relative competition levels of different seasons. He may have some ideas you can use.

[Doctor MJ]

Hehe, well this is something that can send you down an abyss.

Are we talking about strength of the league, or the strength of MVP competition?

Sounds like you're talking about the former, but if you're literally going by MVP votes as opposed to individual stats, it seems to me like it would be more precise to try to judge the actual MVP competition. After all, you could have a theoretical Wilt-Russell type situation where maybe the league is weak, but that guy keeping you from the top spot is monstrously tough.

Were I going based on that, I'd probably try to emulate something like b-r's MVP tracker:

http://www.basketball-reference.com/blog/?p=3898

And then use the combination tracker & actual MVP finish to get a score.

If you were literally going by quality of league competition, an idea that comes to mind is just to use minutes played as a proxy for quality for eras that have passed. As in: Theoretically a particularly strong era would hold on to significant minutes played longer than a weak era. So something like:

p-year = number of minutes that player played that year
p-career = number of minutes that player played in career

quality-year = sum (p-year * p-career) / total-players

And then just do a best fit function for more recent years.

One other thing: You're presumably taking the MVP voting and translating it into your 10-9-8 scale without factoring in the weighting used for the MVP. I would suggest that if your scale is important to you, you try to acquire actual voting details and apply your scale to those details. Otherwise you're letting your system be heavily influence by the NBA's 10-7-5-3-1 system while theoretically trying to apply a second standard over it.

When we did the RPOY project here btw, I cut off debate about the weighting and just said we'd stick with the NBA's weighting simply for the clarity that aligning scales provided (although I get why they want a bigger drop off than 10-9-8)..