A Brief History Of Value In The NBA Draft
You may be asking yourself, “What the heck am I looking at here?”
Just before the draft I did a post for Bleacher Report looking at how well each NBA team has done, over the past few seasons, at finding value in the draft. With some feedback and the benefit of time I’ve been able to expand and refine that analysis. The visualizations above are the result.
If we are going to objectively assess the success teams have had in the draft there are two key ideas that need to be incorporated. The first is a suitable window of time. It’s difficult to assess the value of a player after just a single season. Although they are not necessarily a finished product by the end of their fourth season, that span of professional experience gives a much truer picture of the level of talent each player possesses and the level of productivity to which they are capable. For that reason, in this work I’ve used the total Win Shares (WS) produced by each player in their first four seasons in the NBA as the measure of value. Since we’re looking at four-year increments, all the data in this post comes from the 2000-2009 drafts. At the end of the 2013-2014 season the 2010 draft class will have finished their fourth season in the league and we’ll be able to include them as well.
The other key idea to incorporate is expectations. If we just look at total WS as a measure of value, the teams with high draft picks will appear to have done a much better job of finding value. But in reality they have a much larger and more talented pool of players to choose from and they should be expected to do better in a raw measure like WS. For that reason we need to establish a baseline expectation of value should be gained from each draft slot. To begin that process I calculated the average WS produced by players taken at each draft slot over their first four years in the league. The graph below shows the results.
But what this graph represents is the relative success teams have had drafting at those positions and not necessarily realistic expectations of value. It’s incorporating both the level of talent that we would expect to be available, and the relative success that teams have had. To try and isolate just the talent portion of the equation, I fit a logarithmic curve to the averages above. That curve, which fit with an r^2 of 0.7635 is shown below.
Although this curve is not a perfect fit, especially at the tail end of the second round where a handful of successful picks have skewed the results, it does a reasonably good job of expressing expectations of performance at the various draft picks. From here on out when I refer to expected value of a draft slot, this curve is what I’ll be referring to. Now that we have the time span isolated and realistic expectations set, I can explain what you’re seeing in the visualizations.
There are several different statistics used in the visualization. Most are used to refer both to teams and players, depending on the view. Here are the ones that may be new:
Total Four-Year Marginal Value – Marginal value is simply an expression of how much the production exceeded the expectations. A player example would be would be Rudy Gay, who was selected with the 8th pick in the draft and produced 15.2 WS in his first four seasons. The expected production of a player selected with the 8th pick is 12.32 WS over four seasons, so Gay’s Total Four-Year Marginal Value would be +2.88. In the team view this stat is calculated by comparing the total four-year production by all players drafted by that team, to the expected production of all their draft picks combined.
Total Four-Year Marginal Value Percentage – Marginal value compares a player’s production to expectations and thus gives us a lot more information than just looking at production alone. But it doesn’t fully capture the expectations. Take this example. Brandon Bass and Drew Gooden both produced a Total Four-Year Marginal Value of +2.99 WS. However, Bass was selected with the 33rd pick in the draft while Gooden was selected 4th. Although the produced the same number of WS above expectations, picking Bass is actually considerably more impressive since a 33rd pick is only expected to produce 5.39 WS, total, in four season. So what I did was divide each player and team’s Total Four-Year Marginal Value by their expected value, arriving at Total Four-Year Marginal Value Percentage. To me, this is the ideal measure of value, relative to expectations. A player with a percentage of 100, exactly met expectations. A player with a percentage of 200 was twice as valuable as what was expected from that draft slot. A player with a percentage of 50 was half as valuable as what was expected from that draft slot.
There are tabs across the bottom of the visualization that allow you to change the view. There are three ways of looking at this data. The first is Marginal Value By Team. This graph shows all the NBA teams listed by their Total Four-Year Marginal Value Percentage. The color of the bars indicates the Total Four-Year Marginal Value of each team.
The second view is to look at Marginal Value By Player. In this view each circle represents a draft pick, plotted by their draft slot on the y-axis and their Total Four-Year Win Shares on the x-axis. The marginal value information is indicated by the color and size of each circle. Color represents Total Four-Year Marginal Value. The size of each circle represents the Total Four-Year Marginal Value Percentage.
The last views are the team histories. There is one tab for each team showing their draft picks from 2000-2009. The picks are sorted into four groups – players who were still with the team at the end of four seasons, players who had been moved to another team by the end of four seasons, players who were out of the league at the end of four seasons and players who never appeared in an NBA game. Each player’s production is plotted by Total Four-Year Marginal Value Percentage on the x-axis. The color of each circle represents Total Four-Year Win Shares. The size of each circle represents the Total Four-Year Marginal Value.
Simply explaining the process and visualizations has eaten up quite a bit of space here, so I’m going to save my analysis of the numbers for some follow-up posts. In the meantime, dig in, ask questions, share what you find and let me know if you find any mistakes. Have fun!