This is our old blog. It hasn't been active since 2011. Please see the link above for our current blog or click the logo above to see all of the great data and content on this site.

The Value of an NBA Draft Pick: Part I

Posted by Justin Kubatko on June 22, 2009

It's draft season, when every team is a contender and every fan hopes (and prays) that their team makes the right pick. But if you're a fan of a team with a lottery pick, how optimistic should you be? How much is that fifth overall pick really worth? By looking at data from past drafts we can get an estimate of how much each pick is worth, and once we have those expected values we can answer any number of interesting questions.

I started by looking at all drafts from 1977 through 1991. Why those drafts? Because 1977 was the first post-merger draft and 1991 was the last draft for which there are no longer any active players (Shaquille O'Neal was drafted in 1992). In order to get an estimate of the value of each pick we first have to determine what value is. To me, value is winning games, and I have a statistic named Win Shares that provides a player's estimated contribution to his team in wins. My goal was to come up with an expected career value (i.e., expected career Win Shares) for each of the top sixty picks in the draft. Using statistical techniques, I came up with the following formula:

EV = 76.9 - 18.8 * log(pick)

For example, the expected value of the fifth pick in the draft is:

EV = 76.9 - 18.8 * log(5) = 46.6

On average we would expect the fifth pick in the draft to produce about 46.6 Win Shares. Let's take a closer look at all of the fifth overall selections from 1977 through 1991:

+------+------------------+-------+
| Year | Player           | WS    |
+------+------------------+-------+
| 1977 | Walter Davis     |  78.8 | 
| 1978 | Purvis Short     |  51.4 | 
| 1979 | Sidney Moncrief  |  89.0 | 
| 1980 | James Ray        |  -0.3 | 
| 1981 | Danny Vranes     |  15.0 | 
| 1982 | LaSalle Thompson |  35.3 | 
| 1983 | Sidney Green     |  17.5 | 
| 1984 | Charles Barkley  | 176.0 | 
| 1985 | Jon Koncak       |  28.7 | 
| 1986 | Kenny Walker     |  17.9 | 
| 1987 | Scottie Pippen   | 122.6 | 
| 1988 | Mitch Richmond   |  81.4 | 
| 1989 | J.R. Reid        |  22.5 | 
| 1990 | Kendall Gill     |  48.5 | 
| 1991 | Steve Smith      |  81.7 | 
+------+------------------+-------+

As you can see, eight of these players exceeded their EV while seven fell short.

Here is the value chart for all of the draft picks:

+------+------+
| Pick | EV   |
+------+------+
|    1 | 76.9 | 
|    2 | 63.9 | 
|    3 | 56.2 | 
|    4 | 50.8 | 
|    5 | 46.6 | 
|    6 | 43.2 | 
|    7 | 40.3 | 
|    8 | 37.8 | 
|    9 | 35.6 | 
|   10 | 33.6 | 
|   11 | 31.8 | 
|   12 | 30.2 | 
|   13 | 28.7 | 
|   14 | 27.3 | 
|   15 | 26.0 | 
|   16 | 24.8 | 
|   17 | 23.6 | 
|   18 | 22.6 | 
|   19 | 21.5 | 
|   20 | 20.6 | 
|   21 | 19.7 | 
|   22 | 18.8 | 
|   23 | 18.0 | 
|   24 | 17.2 | 
|   25 | 16.4 | 
|   26 | 15.6 | 
|   27 | 14.9 | 
|   28 | 14.3 | 
|   29 | 13.6 | 
|   30 | 13.0 | 
|   31 | 12.3 | 
|   32 | 11.7 | 
|   33 | 11.2 | 
|   34 | 10.6 | 
|   35 | 10.1 | 
|   36 |  9.5 | 
|   37 |  9.0 | 
|   38 |  8.5 | 
|   39 |  8.0 | 
|   40 |  7.5 | 
|   41 |  7.1 | 
|   42 |  6.6 | 
|   43 |  6.2 | 
|   44 |  5.8 | 
|   45 |  5.3 | 
|   46 |  4.9 | 
|   47 |  4.5 | 
|   48 |  4.1 | 
|   49 |  3.7 | 
|   50 |  3.4 | 
|   51 |  3.0 | 
|   52 |  2.6 | 
|   53 |  2.3 | 
|   54 |  1.9 | 
|   55 |  1.6 | 
|   56 |  1.2 | 
|   57 |  0.9 | 
|   58 |  0.6 | 
|   59 |  0.2 | 
|   60 | -0.1 | 
+------+------+

This method takes the player's entire career into account, but perhaps we should only look at the player's first four seasons, as under the current CBA a rookie's contract can last at most four years (including team options). Here, the expected value will be the player's total Win Shares in his first four years in the league. Using data from the 1977 through 2005 drafts, we get the following model:

EV = 26.5 - 6.3 * log(pick_overall)

Once again, using the fifth pick as an example:

EV = 26.5 - 6.3 * log(5) = 16.4

On average we would expect the fifth pick in the draft to produce about 16.4 Win Shares in his first four seasons. Let’s take a closer look at all of the fifth overall selections from 1977 through 2005:

+------+----------------------+------+
| year | player               | ws   |
+------+----------------------+------+
| 1977 | Walter Davis         | 36.0 | 
| 1978 | Purvis Short         | 16.5 | 
| 1979 | Sidney Moncrief      | 40.9 | 
| 1980 | James Ray            | -0.3 | 
| 1981 | Danny Vranes         | 11.7 | 
| 1982 | LaSalle Thompson     | 18.3 | 
| 1983 | Sidney Green         |  6.7 | 
| 1984 | Charles Barkley      | 45.8 | 
| 1985 | Jon Koncak           | 13.2 | 
| 1986 | Kenny Walker         | 14.6 | 
| 1987 | Scottie Pippen       | 22.6 | 
| 1988 | Mitch Richmond       | 24.2 | 
| 1989 | J.R. Reid            |  8.5 | 
| 1990 | Kendall Gill         | 16.3 | 
| 1991 | Steve Smith          | 20.5 | 
| 1992 | LaPhonso Ellis       | 17.0 | 
| 1993 | Isaiah Rider         | 12.1 | 
| 1994 | Juwan Howard         | 20.4 | 
| 1995 | Kevin Garnett        | 27.1 | 
| 1996 | Ray Allen            | 26.9 | 
| 1997 | Tony Battie          | 10.0 | 
| 1998 | Vince Carter         | 36.4 | 
| 1999 | Jonathan Bender      |  3.4 | 
| 2000 | Mike Miller          | 18.0 | 
| 2001 | Jason Richardson     | 17.8 | 
| 2002 | Nikoloz Tskitishvili | -1.6 | 
| 2003 | Dwyane Wade          | 38.5 | 
| 2004 | Devin Harris         | 18.2 | 
| 2005 | Raymond Felton       | 12.4 | 
+------+----------------------+------+

Seventeen of these players exceeded their expected value and twelve fell short.

Here is the value chart for all of the draft picks:

+------+------+
| Pick | EV   |
+------+------+
|    1 | 26.5 | 
|    2 | 22.1 | 
|    3 | 19.6 | 
|    4 | 17.8 | 
|    5 | 16.4 | 
|    6 | 15.2 | 
|    7 | 14.2 | 
|    8 | 13.4 | 
|    9 | 12.7 | 
|   10 | 12.0 | 
|   11 | 11.4 | 
|   12 | 10.8 | 
|   13 | 10.3 | 
|   14 |  9.9 | 
|   15 |  9.4 | 
|   16 |  9.0 | 
|   17 |  8.7 | 
|   18 |  8.3 | 
|   19 |  8.0 | 
|   20 |  7.6 | 
|   21 |  7.3 | 
|   22 |  7.0 | 
|   23 |  6.7 | 
|   24 |  6.5 | 
|   25 |  6.2 | 
|   26 |  6.0 | 
|   27 |  5.7 | 
|   28 |  5.5 | 
|   29 |  5.3 | 
|   30 |  5.1 | 
|   31 |  4.9 | 
|   32 |  4.7 | 
|   33 |  4.5 | 
|   34 |  4.3 | 
|   35 |  4.1 | 
|   36 |  3.9 | 
|   37 |  3.8 | 
|   38 |  3.6 | 
|   39 |  3.4 | 
|   40 |  3.3 | 
|   41 |  3.1 | 
|   42 |  3.0 | 
|   43 |  2.8 | 
|   44 |  2.7 | 
|   45 |  2.5 | 
|   46 |  2.4 | 
|   47 |  2.2 | 
|   48 |  2.1 | 
|   49 |  2.0 | 
|   50 |  1.9 | 
|   51 |  1.7 | 
|   52 |  1.6 | 
|   53 |  1.5 | 
|   54 |  1.4 | 
|   55 |  1.3 | 
|   56 |  1.1 | 
|   57 |  1.0 | 
|   58 |  0.9 | 
|   59 |  0.8 | 
|   60 |  0.7 | 
+------+------+

Armed with this information we can analyze past drafts and determine the biggest busts, the biggest surprises, and much more. Later this week, that's exactly what we'll do.

6 Responses to “The Value of an NBA Draft Pick: Part I”

  1. Johnny Twisto Says:

    Have you ever established a Win Shares replacement level? I'm not even sure how that is defined in basketball, since teams can control how many minutes and touches their worst players get (I apologize for my ignorance in cagermetrics). Anyway, I ask because based on your first chart, two #8 picks would have about the same expected value as a #1 pick. Would anyone trade the top pick for two #8s? (or say, a 7 and a 9, so it could be in the same draft.) I suppose it depends on the perceived talent in the draft and the individual team's circumstances, but I'm inclined to think teams would not give up that top pick. I'm not sure if that means the best players in the league have more value than WS assigns them, or if one should consider these WS numbers in relation to a baseline higher than 0, or something else.

  2. Justin Kubatko Says:

    Johnny, one thing to keep in mind is that the top pick will produce that value in x minutes, while a #7 and a #9 would produce that value in about 2x minutes. I'd rather have the guy that's going to give me the same value in fewer minutes.

  3. Johnny Twisto Says:

    Ah yes, overlooked that little fact. Thanks.

    How do the different picks look in terms of WS/minute?

  4. anonymous Says:

    Since lower picks end up on teams that have usually performed well in the year before, they'd get less playing time compared to those picked early, resulting in artificially low win shares. With this in mind, it seems very difficult to find any metric that would be able to compare players who play many minutes with those that barely play at all. Maybe it's best to compare the players' productions 5-6 years after being drafted when they have a chance to show their abilities?

  5. Dave Says:

    Justin the Aritmetic Mean for your #5 picks is 57.7 so we are going to expect "on average" that you number 5 pick is going to show to be better than the EV (Expected Value) that you derived. Now I know this is mostly Barkley, but then he played thousands of minutes, is it reasonable to measure his WS to James Ray's?

    Obviously the EV (regression?) equation is fitted to all picks, not just #5, but WS is correlated to MP...

  6. Gill Says:

    What are the values for undrafted players first time in league and run of the mill minimum salary free agent pick-ups?