Freelance Friday is a project that lets us share our platform with the multitude of talented writers and basketball analysts who aren’t part of our regular staff of contributors. As part of that series we’re proud to present this guest post from Mika Honkasalo, a 21-year old NBA enthusiast from Helsinki, Finland who is studying computer science and mathematics. Mike enjoys big men who can pass and players who can shoot off of screens, find him on Twitter @mhonkasaloNBA.
Not matter how much you try to avoid them, small sample sizes are everywhere.
When querying career averages from players based on their NBA Draft positions, specifically Win Shares per 48 minutes (WS/48) and using the past 20 drafts as data points, I noticed I needed to modify some entries because of players like Tyson Wheeler.
Wheeler played an entirety of three minutes in the NBA. During that incredible three minute span, Wheeler made a three-point bucket, had two assists, one foul and shot one-for-two at the free throw line, posting an incredible 1.367 (WS/48). Meaning that Wheeler, during his brief stint in the NBA, was worth about five and a half Lebron James’ on a per minute basis.
That’s obviously just a funny story and something you’d expect to run into when sorting through every single player who was drafted in the past 20 years. The original and uncreative fix for the Tyson Wheelers of the NBA was to set up a minutes limit for players eligible to be on a cool chart that would show the relative values of each pick.
If we treat players who barely played at all in the NBA as non-factors and reduce their numbers near or to 0, the graph showcasing the relative values of picks looks like you would expect1.
X = Pick # | Y=WS/48
There is a significant drops in value toward the end but when you apply a 300 minute filter2:
The original decline of expected value between the first few and some of the last picks drops from over 100% to about 30%. The equivalent of which would be the statistical difference between Kevin Garnett and Andrei Kirilenko, instead of falling from KG to Yi Jianlian.
The 300 career minute filter does disqualify about 25% of all drafted players. But basically, if your guy “makes it “ in the NBA, he’s usually going to turn into a pretty good player.
When first round picks in the 20’s or second rounders are traded we often talk about the expected value of the pick, particularly how low it is. What we don’t take into account is that teams with lower picks should be more interested in variance, rather than averages. You should count “hits,” not expected values (highlighted in the next chart).
A perfect example of this is the No.57 pick, which has produced only four players with more than 300 career minutes, but two of those players are Manu Ginobili and Marcin Gortat, and together they average out to a far better performer than the average No. 1 pick.
In other words; what is my chance of hitting an absolute home run (or at least significantly outperforming expectations)?
In the chart above3, I’ve separated players into four tiers (could have produced more, but a four tier system made for a cleaner chart and doesn’t over credit Win Shares as a statistic) of performance, to show the probabilities of obtaining targets who perform to at least a certain level of play. The lowest tier includes players from higher tiers.
The odds of getting an above 0.74 win share player in the middle of the second round is about 25%, and typical players in that group include Courtney Lee, Mo Williams Boris Diaw and Kirk Hinrich. It’s pretty easy to fleece these picks from teams, and the Sixers took that notion to the extreme when they acquired a combined nine second round picks for the 2014 and 2015 drafts.
When we get into the very best players and real difference makers, after the 15th pick in the draft there’s an approximate 7 percent chance of landing a very good to great player. Even in the second round. Typical players in this tier include: Taj Gibson, Al Jefferson and Rajon Rondo. What’s interesting about this is that after the 15th pick finding those guys is a total crapshoot.
This chart also shows that there are less “safe choices” than we would imagine already in the middle of the first round. Scouting has gotten better, but the NBA in recent years still has the same trendlines. NBA basketball is unlike most team sports because the impact of one player on the end result is proportionally larger. So if you have a chance of getting one of those players you should absolutely swing for the fences.
So go ahead and draft Giannis Antentokounmpo based on some YouTube highlights. Search for variance4
All statistics from Basketball-Reference.com
- Regression line equation: y=0.096-0.0014x, Mean: 0.052, Correlation Coefficient: -0.90 ↩
- Regression line equation: y=0.086-0.0004x, Mean: 0.071, Correlation Coefficient: -0.38. ↩
- The tail-end of each line is probably a tiny bit inflated due to small sample sizes. Insignificant amounts though. Apparently teams picking between 16-20 have done quite poorly compared to what you would expect. ↩
- Based on a quick scan of the type of players who overachieved relative to their type of players:
- Versatile or athletic big men: David Lee, Paul Millsap, DeAndre Jordan, Taj Gibson, Serge Ibaka
- Hard nosed wing defenders and hope they develop jumpers: Kawhi Leonard, Jimmy Butler, Gerald Wallace
- International (again, improvement in scouting can negate some of this. But this is by far the biggest “hit or miss” category. Still much more likely to hit.). ↩