Continuing on our recent theme, in this post I’m going to take a look at player salaries by draft position. As with the rest of our salary-related posts, these numbers have been adjusted for inflation.

First up is *average* career earnings by draft position:

Remember, this is salary, not productivity; and yet we still see a huge drop between #1 and #2, #5 and #6, and #10-#12. As a matter of fact, this graph seems to take on a familiar shape. We also can spot a few spikes that are attributable to one or two individual players: at #5, we see the **Kevin Garnett** effect. At #13, we see the influence of **Kobe Bryant**. At #29, we see the cumulative effects of **P.J. Brown**, **Toni Kukoc**, **Nazr Mohammed**, and **Josh Howard**. At #37, be reminded of the fact that **Nick Van Exel** and **Mehmet Okur** used to be relevant NBA players. And the granddaddy of outliers, #57, is skewed by **Manu Ginobili**‘s $71 million career earnings and a sample size of two (the other player, incidentally, is the very good **Marcin Gortat**, illustrating that survivour bias I mentioned earlier).

But remember, that was *average* career earnings. It would also be good to take a look at the median career earnings as well, because averages can be skewed. The median tells us the salary of the player who finds himself smack dab in the middle; 50% of players will have salaries larger than the median, and 50% of players will have salaries smaller than the median. And because I want to compare median salaries to average salaries, I’ve used the same scale and actually included a faint outline of average salaries in this next graph:

Basically, any difference between the faded and solid bars represents the skew. Some draft slots are more skewed than others, and the outlier, #57, shows no difference (because median and average are the same over a sample size of two). Other than the slots I’ve already mentioned, now we see that #9 (**Tracy McGrady** and **Dirk Nowitzki**), #17 (**Jermaine O’Neal**), and #24 (Latrell Sprewell and **Andrei Kirilenko**) have been skewed as well.

Now I’d like to look at the three types of players — players who were drafted and signed, players who were drafted and immediately waived, and undrafted players:As you can see, drafted players make much more over their careers than the other two types of players. Interestingly, it appears that undrafted players make more than waived draftees. How does all this change when we go from average to median?

Here we see quite clearly how massively skewed NBA salaries are. Waived draftees actually have higher median salaries than undrafted players, which means that you’d still rather be drafted and then waived than to remain undrafted. Final tally for the median *career* earnings of an NBA player? $4.98 million.

## Methodology

But it’s important for you to understand the methodology I used to create these graphs. Our data *only* include players for whom we have salary data. That means that players who were drafted, but did not sign an NBA contract don’t show up; this can skew salaries for second round picks by introducing a “survivour bias“. There are also a few holes in our salary data, particularly involving players who were signed to 10/15 day contracts. As a matter of fact, if our dataset was complete, the averages and median would be significantly lower, but I am forging ahead with what we have.

I’ve also recorded draftees somewhat differently than normal. For every draftee who was waived before they played an NBA game, I have removed them from the first two graphs. Why? I figure that, once they have been waived, these players are more like undrafted players. In the last two graphs, judge for yourself whether or not that move was appropriate.

– Devin

*Basketball Stories*

A.K.S.

October 24, 2011

Seems to me you’d be better off removing the survivor bias by including players who were drafted but never played. You wouldn’t have to do this on an individual basis (i.e., figuring out who those individuals actually were) of course, you’d just have to divide by the correct number. I.e., to find out the average salary of all the players drafted #57, you wouldn’t divide by 2 (as you have above) but rather by 21 (if my counting is correct, there should be 21 players in the relevant timeframe who were drafted #57).

Devin Dignam

October 24, 2011

That is a possibility A.K.S. There a a few complications that make it not so simple (for example, players who were drafted before our 1990-91 cutoff, but played their first season during our time period).

Overall, given the data we had, I decided to go with the method above. Your suggestion was considered, but not selected.