NBA’s “Greatest of All Time” According to the Data
Using 50+ features (statistics), regression, and many NBA analyst lists to rank 100+ of the greatest NBA players to ever do it.
--
Introduction
The NBA (National Basketball Association) founded in 1946 has contained the best current and past players that have ever played the game of basketball. Everyone who is ever played basketball or knows even something about this sport always debates, “Who is the Goat”… an age-old debate that will never seem to fade out of the conversation of many sports fans. Everyone has an opinion that may be biased or unbiased on why this certain player is the greatest to ever do it.
In this article, I use regression to assess hundreds of the greatest players to ever play in the NBA. Adding data into the conversation uncovers and settles many debates on players based on their statistical impact and tangible accomplishments in the game of basketball.
My Video On this Subject, But… Greater:
The Results as of 6/20/2023 (Top 10):
Data
All of the statistics used are from basketball reference and NBA stat muse. For more information on the different statistics used in this project click here.
Raw Data File: (Statistics used for this project) NBA Goat RawData.xlsx
The articles/NBA analyst’s lists of NBA’s top 10 of all time ranking that are used as a prediction in regression are: (All in 2023)
*Other lists not linked include lists such ChatGPT, ESPN, Jalen Rose, Nick Wright, etc.*
Linear Regression Model -> Goat Share
The output works based on regression coefficients computed from several NBA sites’ lists of the top 10 greatest NBA players of all time. The output itself is mainly focused on the top 10 spots.
- I simply take several NBA analysts’/fan’s posted top 10 players of all time and average them out into 1 ranking (average rank).
2. I then use correlation to test all 50+ different player statistics on the average rank and find what are the correlation values for the average list of a common NBA analyst/fan.
3. I use regression and let the coefficients decide the output multiplier for each statistic. Simply put, regression predicts the ranking of the players based on the average list (average rank).
4. The predicted ranking by regression is re-calculated as Goat Share (my custom statistic name for player score) as (minimum predicted ranking(out of all predicted rankings of players)/player predicted rank)*100 = Goat Share for that player.
It simply recalculates the predicted ranking by regression into a cleaner number for graphing appeal.
This way the entire list is unbiased simply based on average ranking and many different player statistics. For information on the lists I used and statistics used for all players check the raw data file listed in the data section of this article.
What is Goat (Greatest of All Time) Share?
This is simply the regression predicted ranking re-calculated as Goat Share or a player score (custom statistic name). The reason for the re-calculation (Goat Share) is that simply makes the player’s score look cleaner for a descending graphing.
Statistics/Features Information
As you will see I have gathered many regular and advanced stats that show the player’s impact on the regular season, playoffs, and awards.
Something I would like to mention is how the MVP award or DPOY award is measured by most people just by how many of that award, that individual has obtained throughout his career. I take into consideration the “shares” aspect of the award giving a more accurate depiction of how that player has performed in those contests over the years, which assesses the true value of the player over their entire career.
For example, Chris Paul, a player who has never won an MVP award, has obtained 2.0+ MVP shares throughout his career, meaning he has enough shares on average that actually qualify him for earning an MVP over his career.
On another note, I came across an interesting article on the medium by Eric Hofmann that used logistic regression to assess how many finals MVPs Bill Russell would’ve won.
Bill Russell has won 11 titles, but the amount of FMVPs is unknown as this wasn’t a factor back in the 1950s. Hofmann’s research has found the number of FMVPs he should have won would have been only 5 surprisingly. The other players who should’ve won it were Sam Jones, Bob Cousy, John Havelock (Russell’s other Celtic teammates), and Cliff Hagan.
I even did some side research into updating the past players’ un-tracked awards/statistics to make this list as accurate as possible.
Model Output [Correlation Values]
What is the Secret NBA GOAT Formula?
With these depicted charts the GOAT formula can simply be summarized in these 4 steps (Most Important-Least Important):
- The player must perform exceptionally well over his career in the MVP contest (Get many MVPs and Votes Every Year): Awards_MVP_Shares, Awards_MVP, etc.
- The player must be clutch in the NBA finals and must have a high offensive impact on the court, specifically in the playoffs (Win finals MVPs and be Statistically Dominate on Offense): Awards_FMVP, Playoffs_WS, Playoffs_OWS, etc.
- The player must be valuable to the team in terms of point differential and must have a defensive presence, specifically in the playoffs (Impactful on both ends of the Court): Playoffs_VORP, Playoffs_DWS, etc.
- The player must make many deep post-season runs and must be an overall impactful player during the regular season (Dominates the Playoffs and Plays Well in the Regular Season): Awards_Conference_Finals_Appearances, Season_WS, Awards_Championship_Appearances, etc.
Other Statistical Graphics
The graph above shows which players were ranked 1st in each statistical category (feature). Michael Jordan absolutely crushed this assessment, even though LeBron James ranked closely behind Jordan in the model output no one even touches Jordan in terms of statistical dominance in terms of ranking. The fact that Michael Jordan and LeBron James are on this list is remarkable, simply because big men (historically) are the most efficient making them dominant in many advanced stats and closing rim shooting stats. Michael Jordan truly is a master of many.
Who is the GOAT?
With this unbiased regression model, you can see that Michael Jordan and LeBron James are undoubtedly the two best to ever do it. Jordan may have an edge in terms of peak, feature (statistical) leadings, and have a higher share in the model, but I want to give my respect to LeBron James as his longevity is unmatched in the game of basketball and is arguably the most complete basketball player ever.
Feedback
This has been an overview of my statistical process that assesses hundred of the best NBA players ever to do it. If you wanted to contribute to my future updated list drop your top 10 greatest NBA players list below so it can be added to the overall average.
