Thread: The Ultimate NBA Greatest Player List and Neutralized Stat Reference!

[Lionel20]

Link @ the Bottom of the page!

Each player is ranked according to individually estimated contribution, or Value Share.

My concept of Value Shares is based on linear weights including Rosenbaum's Alternate Win Score, and the Value Over Replacement Player (VORP) metric. Because each of these metrics seem to lack especially in measuring individual defensive effectiveness, through my research I've implemented my own tweeks, most importantly to factor in defensive value not captured in the box score.

The Value Shares formula is a counting stat with a high correlation with wins (I can bore you with the details in a later post), it comprises of other details including my own recalculated Usg%, but significant adjustments are the reassessed box stats.

The first reassessment is the time period regulation, which covers among others things the change in pace, generally from decade to decade although I made some adjustments to mirror major NBA rule changes. It's fundamentally impossible to even start with comparing players across from different time periods without first forming a neutral period to serve as the common denominator. Generally, RPG, FTA in the 60s is higher than it is today, possibly due to pace. I didn't get into the "why" much, although there are very good explanations there. I simply took account of the league averages. The most important feature of the time period regulations, is of course which time period do I denominate all the box stats? In concluding my research, I found the early 90s to be the most neutral time period in NBA history to this point. By neutral I mean, the league averages and pace were fairly moderate, and "top-to-bottom" there was generally high level play from all positions.

That brings me to my next reassessment, the league quality adjustment. There are many factors that have affected the NBA quality of competition in my estimation: Expansion, Demographics, Rule Changes, etc. In my selection of the 90s as the most neutral period, the rules in particular allowed for a more balanced game. The lane dimensions compared to the early 60s, physicality compared to today's NBA (hand-checking, defending around screens, etc.), Rules governing offensive/defensive play in the paint, all allowed for a more quality game. If you were a poor defensive player, you were more likely to get exposed and exploited, if you were a great offensive talent, you were more likely to showcase your skill. In measuring League Quality, I used ADJP48 for average players, generally per decade for the seasons available, and average player WS/48 for the seasons where ADJP48 is not available.

Lastly,

*I prorated the adjusted box stats for strike-shortened seasons
*All Season stats totals include post season stats.
*I made slight adjustments in 2pt FG, FGA for pre-3pt NBA periods
*I made slight adjustments for Assists totals for pre-NBA/ABA merger seasons, it's my belief that scorekeepers were less generous during these periods
*70s ABA league quality = 60s NBA league quality
*Used multiple statistical models to estimate unavailable Box Stats
*GP are calculated as Total Minutes/36 = GP
etc.


https://docs.google.com/spreadsheets/d/1XUjK5NG4FTPoSQtGg_wJzpPuNOFA6SpirolxUVp-aoE/pubhtml

[yellowknifer]

Took a little time to browse through some of the rankings... they feel problematic. I think adjusting your data introduced a lot of bias. I liked seeing certain names repped more than they have been historically, McAdoo in particular, but I think a lot of your work was for naught as you clearly influenced the data in a negative way. You have Curry's 2015 season as superior to his 2016 season. Not sure how that math works.

Still a nice/fun effort.

I agree that eras changed and that the talent level is higher now, but I still think adjusting things only causes problems. It's simply not realistic to statistically compare games where the rules have changed dramatically, which I think bugs some historians but for me it is what it is.