With the start of the season about three weeks away and every player in the best shape of their lives after a summer of five a day workouts filled with rising and grinding, irrational optimism regarding team success is at its zenith. Especially given that it seems so many of the of the second year and incoming rookie players have such obvious potential. Therefore, I decided to take a stab at quantifying that potential and then determining what players reached that threshold and those that did not. There will be several chapters in this series, so stay tuned. The articles will contain varying levels of statistical complexity, and this one is pretty low.
Introduction to Absolute and Relative Quality
It’s fairly well established that in general, lottery picks are likely to be better than non-lottery, and that high lottery are better than low lottery. But we should take a quick look at the data nonetheless. One of the difficult parts of determining potential is that you first have to establish a metric for measuring potential/success. Luckily, this has been done by others in the form of Win Shares. Like most methods for measurement, there are legitimate criticisms in Win Shares (WS), but for our purposes here, WS will work nicely as a catch all for quality. Take a look at the career leaders in WS and you can see that there is definitely a trend for better players near the top. Again, I would not use WS to say that Reggie Miller was a better player than Kobe Bryant, but we mostly are concerned with the general trend that more WS means a better player, and more importantly a better career, as WS is a cumulative statistic.
One other point needs to be clarified prior to beginning our analysis. When we judge the quality of a draft pick, we should use both absolute and relative measures. The absolute quality of a pick is just simply the sum total of that player’s career and ability with no context applied. The relative quality takes into account the players that were also available to draft. For example, if you are a team picking fifth in the lottery and pick a clone of Chris Bosh, you did a good job from absolute quality standpoint. However, if the draft class had ten clones of Michael Jordan that you passed up, you did a pretty terrible job in terms of relative quality. Conversely, if you picked a Nerlens Noel clone fifth, and the rest of the draft class was composed of people like me who had not played an organized game of basketball since sixth grade, the absolute quality of the pick was not great, whereas the relative quality was tremendous. The take home point here is that a draft pick can be judged both on how good they turned out to be and also how good the rest of their class turned out to be. Figure 1 shows the career win shares for the 2002 and 2003 draft picks for an example.
While Darko Milicic and Jay Williams are both absolute quality busts, Darko is a legendary relative quality bust based on the gentlemen picked at three, four, and five.
Determination of Relative Quality
As indicated previously, total WS can be used as a pretty decent approximation of a player’s career quality. However, if we compare the 2010 draft class to the 2003 draft class, the total WS will be biased towards to the 2003 class as they simply have had more time to accumulate wins. Therefore, we need to figure out a reasonable metric to take time out of the factor. This can be accomplished by turning each player’s total WS into a percentage of all WS accumulated by that draft class. This takes into account both time and the relative quality of a draft class to get a metric of Win Share percentage (WSP) that can be compared across years. For example, if a player has the same WSP as LeBron, that does not mean that they are absolutely as good as LeBron, but that they are approximately as good in comparison to their own class. As always, these are not big bright line demarcations, but a general trend when observing the NBA as a whole. Using that metric of WSP, Figure 2 shows every pick from 2000-2015, colored by their WSP. Players with WSP greater than 10% (arbitrary cutoff) were labeled.
The black line is a simple linear regression of WSP and pick position, indicating that as you might expect, WSP decreases (gets worse) with a worse draft pick. Also, you will note that there are many more players who did not make it to the court in the second round, producing NA WSP values.
For a year by year look at best and worse WSP within the lottery, take a look at Figure 3.
Keep in mind that LeBron has greater than 15% of all WS in his draft class, which also contains Carmelo Anthony, Dwyane Wade, and Chris Bosh — that is pretty incredible.
Expected WSP by Draft Pick
Moving forwards, we’ll now look at what what sort of WSP has been the historical average for each lottery pick (mostly chosen as a cutoff to make graphs readable). Figure 4 is a boxplot based on the WSP of each player taken at each pick. For a review on boxplots, click here.
The boxplot uses median as its measure of central tendency, which for this specific application is consistently less than the mean. This should make sense as extreme upper values (LeBron, etc.) can pull the distribution in that direction, and there are no corresponding extreme low values — WSP of zero is about the lower bound. If you’re so inclined, you can see the graph of the differences here. For consistency moving forwards, we will be using the average, as that is more familiar with most readers than then median.
If we calculate an average WSP for each lottery slot, we can then assign that average as the expected WSP for that pick. If you had to guess what the WSP of that pick was likely to be, you would choose that value. Now, with an expectation for what the WSP “should be” for a pick, we can look at at the residual, or the difference between the the observed (the actual WSP for a player) and the expected (the average WSP for that pick). Then we can rank by size of that residual with a positive residual indicative of relative overperformance for that draft pick, and a negative residual of relative underperformance for that draft pick. For each team’s lottery picks from 2000-2015, Figure 4 shows which player had the best total WS, worst total WS, best WSP, worst WSP, and then as just discussed the best overperformance and worst underperformance. Keep in mind, this is a) lottery picks only and b) purely based on the team that drafted them, and not where they ended up or were traded, etc.
Keep in mind that Total WS is a cumulative statistic, so if you look at the Pacers, I am relatively comfortable saying that Myles Turner will have a better career than Jerryd Bayless, but Bayless has had time to accumulate those WS. The underperformance column is a pretty depressing read across the board through.
Coming soon...On the Clock Part 2: Does Your Team Draft Well?