Tuesday, August 9, 2016

John Stockton and other outliers

There has been a lot of discussion over the past month of my placement of John Stockton at #83 in my all-time rankings of the top players in NBA history, so I felt like I should use this post to explain a little bit about my formulas and how they work.

First of all, the results of my analysis does not always represent my personal opinion. I have several different formulas that I have created for rankings teams or players from several sports or leagues, and I have never looked at the results of any of them and felt they were perfect. I am always watching for teams or players that feel out of place, AKA outliers, and I use those to determine either errors or weaknesses in my method.

For example, in last season's college football rankings, Georgia Tech always seemed to be ranked higher than seemed correct, and losses in several games in which I picked them to win only confirmed that feeling. As a result, I have analyzed their results from last season and have pinpointed a possible cause for that aberration, and I am making an adjustment to this year's version of my rankings to try to correct it.

This has been a recurring theme throughout the time that I have done statistical rankings. Over the years I have added new components to my formula nearly every year to try to make the resulting predictions more accurate, and my correct pick percentage has risen almost every year as a result. The most recent adjustment to that formula was a method for integrating a momentum factor into the equation, and it worked pretty well.

In player rankings it is a little more complicated. There is no way to easily measure whether rankings one player over another is correct, because there are no individual matchups that show a definite winner. But just like in team rankings, when a player seems to be ranked too high or too low, I pay special attention to that player and try to find a way to correct the perceived problem.

There are a few players in the current version of my rankings that fit this criteria, with John Stockton as the biggest example and Bill Russell as another. Five years ago I released the previous version of these rankings, and there were even more examples of players I felt were not ranked correctly then. While improvements have been made in this system over the past 5 years, it is still not perfect, and that is part of the fun for me. It gives me a chance to continue thinking about ways to improve the system.

Now let me get back to John Stockton, the biggest outlier of all. He has always been one of the most difficult players to rank, no matter how many changes I have made to my formula over the years. Since finishing this most recent version of my rankings (which took dozens of hours to complete), I have spent a lot of time reviewing the data to ensure its accuracy, and I have found that he is ranked accurately according to this version of the formula.

But that leaves a problem. Does anyone really believe that John Stockton is only the 83rd greatest player of all time? I don't believe that, and you probably don't either. As I have pondered his unique career, I have tried to find an adjustment that can be made that would more place him in a ranking position that seems more accurate. I have come up with a couple of ideas, but neither would be guaranteed to fix the issue and both would take many more months of work to implement. I will continue to study these possibilities and run some tests, and maybe by the next time I am ready to release a version of these rankings they will look a little different.

In conclusion, I am aware that John Stockton's rankings is probably unfair, and I agree that he should be ranked ahead of several of the players that I have highlighted ahead of him on this list. Rather than focusing on the injustice of his ranking and comparing each player ahead to John Stockton, let's instead focus on celebrating the greatness of the players on the list and learning a little more about them.

No comments:

Post a Comment