While I was applying for a job with an NBA team, they asked me to project the three-point shooting percentage for each player in the NBA for the upcoming season.  This was in October of 2014, so the projections were for the 2014-2015 season.  I used an interesting method to make my projections, that was based off of work that I did for my PhD in Nuclear/Particle Physics.  It’s a pretty versatile method!  While I may convert this to a full blog post later, for the time being you can find the full write-up here in its original format.

While I was applying for a job in the NBA, my research kept leading me to people who said they were making their predictions using regression-to-the-mean.  This didn’t make sense to me, because they were not dealing with purely random statistics.  It is true that there is a degree of randomness to basketball statistics, most of which can be determined using binomial statistics, but there is also a significant amount of spread in the true abilities of the players.  So I wanted to look into exactly what people were doing, and how I would do it myself.  It turns out that there are a ton of different methods and so I didn’t cover them all, but looked at one of the main ones.  And I derived my own way of doing things.

All of my work is summarized in this paper.  I may convert the whole thing into a blog post later, but for now you can just check out the pdf.

This post is an adaptation of a paper that I wrote to answer the question of what is the best way to value draft picks in trades.  The original paper can be found here in its intended format.  Converting this to a blog post led to some difficulties in formatting, especially when it comes to captions on multi-figures, but I hope it is as clear as possible.

I.   Introduction

Draft picks are an important commodity in the NBA, able to be used to directly add players to a team, or they can be traded as assets for current players, other draft picks, or other assets. Therefore it is important to be able to place a value on those draft picks in an abstract sense. As with everything in the NBA, how well one is able to do that depends entirely on the situation, however it is possible to make assessments based on generic situations, and then to allow the actual circumstances to dictate the final decision.

This paper will be concerned with assessing the value of draft picks in prospective trades; it will consider the generic case of a team trading the rights to a future draft pick given certain protections, finding the probability of where the pick should be expected to fall based on the team’s performance and projections for the future, and giving a relative value to each drafting position. There is a bit of feedback in this situation, because accurately projecting how well a team will do depends on being able to predict how much the rookies on the team will help them, but that will be accounted for.

In the end, the goal is to have a function that will be able to provide the expected value of a team’s future draft pick, taking into account any protections placed on the pick, as well as the likelihood that it will be transferred (i.e. not protected) in the nearest draft.

In Section II, I will cover some of the basic rules for the draft, focusing mainly on the lottery system.  In Section III, I will describe how to find a mapping of a team’s winning percentage onto their likely finishing position in the league that season. From there, I will describe how to determine the drafting order based on a team’s likely finishing position in Section IV. Then in Section V, I will describe how to come up with a way to give a value to each pick in the draft, from 1 to 60, based on past performance of players picked in each position. In Section VI, a number of methods will be presented to predict a team’s winning percentage at the end of the season that leads into the draft. Finally, in Section VII, I will put all of the parts together to form a cohesive way of evaluating a team’s draft pick for the purposes of a trade.

 

Continue reading →