If you’ve seen Kubatko’s model before, you’ll quickly recognize that this is nothing new. I’m not inventing anything here; I’m just updating his model to reflect the 2004-2012 NBA drafts instead of the 1977-1991 drafts. I chose 2004 as the cutoff because that was the year when the NBA draft first included all 30 teams in order to include the then Charlotte Bobcats. In particular, I am only looking at the first four years of a draftee’s career, since that is the length of a first round contract. I take the total Win Shares of each drafted player over the first four years of his career and then fit a logarithmic curve to the average Win Shares of each pick. The only difference that I have here is that I attempted to establish a Win Shares replacement level. My goal was to evaluate how teams were using picks in trades, so we have to measure these players against some guy that any team could have gotten for the veteran’s minimum.
Players who don’t play
I evaluated the first four seasons after the player was drafted. If a player didn’t play due to injury, he got a 0 for that season. If a player didn’t play because he was overseas, he got a 0 for that season. My reasoning for this was that I wanted to evaluate players for the value they bring to a team while still on the rookie contract. If a player is hurt, then that is a year on a low salary that is also gone. If a player doesn’t come over from Europe for three years, then he is no longer forced to sign in accordance with the rookie wage scale. Of course, there are players that only play one or two years overseas, but I didn’t think it was fair to count the seasons when they would be a year or two older than expected when they actually started to rack up Win Shares. Thus, I left any player who didn’t play as a 0, which happens for a lot of second round picks, but very few first rounders.
According to Kevin Pelton, a replacement level player is about 83 percent of an average player. I used that number as the basis for creating the Win Shares replacement level. I figured that the average player rates around the 50th percentile, so the replacement player rates around the 41st or 42nd percentile (83 percent of 50). I simply determined the replacement level as the number of Win Shares a player in that percentile has when ranked by Win Shares. The number changes slightly each year, as it should, and hopefully gives us a pretty good idea of where to value the replacement level. For most years, the replacement level came out to about 1.2 Win Shares.
I don’t believe that it’s fair to just sum up the total number of expected Win Shares from draft picks plus the total Win Shares from NBA players traded since roster space is a scarce resource. If you have one player whose production is equal to that of two other players combined, then the one player is more valuable than the two since he can provide the same thing in half the number of total minutes and take up half as many roster spaces. Put another way, if you have an entire roster equal to Lebron James, then the Cleveland Cavaliers have a huge advantage because they have 14 other players that will surely be better than nothing.
The way I am accounting for the value of a roster space is by taking away one-fifteenth of the value for each extra player added. For example, if I trade one draft pick for two picks that I have valued at a total of ten Win Shares, then the value I get in return is actually: