With more sophisticated data available than in decades past, teams should be getting better at drafting. Basic evolution, like how a PlayStation 4 puts a Sega Genesis to shame. That’s not the case.
Ahead of Thursday’s NBA draft, The Washington Post studied data from every draft in a 25-year span and found no indication that teams are beginning to decode this high-stakes puzzle.
“Some processes, some futures, are just impossible to predict,” said Cade Massey, a professor at Penn researching how well people predict uncertain futures, which he applies in consultations with teams around the draft. “. . . The thing we’re forecasting is human beings. And not only that, they’re, like, 18 years old.”
Drafts are especially tricky, Massey said, because “feedback is delayed and ambiguous.” Unlike a weather forecaster or a sports bettor who quickly knows whether they made a correct prediction, a general manager drafting a basketball player must battle a lot of noise and time to determine whether they made the right pick.
And if their pick is wrong, they might not be employed long enough to make many more.
A fine line between right and wrong
Let’s use the iconic 2003 draft, spearheaded by LeBron James, to explain how The Post analyzed each class for accuracy.
James was the No. 1 pick and also has the most career win shares of any player from that draft. (Win shares is an all-encompassing stat used to measure how a player affects winning. Though imprecise, it is a good baseline for judging players’ performance.) Because James was taken No. 1 and turned out to be the No. 1 player, he is a perfectly accurate pick. Therefore, that pick is docked zero points for accuracy.
The No. 2 pick, Darko Milicic, bounced around during a largely disappointing career that ranks 32nd in win shares among players in that draft. That is a 30-spot difference between where he was picked and how he produced, so that pick equals a 30-point miss.
A pick can have a high variance in the opposite direction, too. Kyle Korver was picked 51st in that draft but became an all-star and has submitted the sixth most win shares. The Korver pick receives a damning score of 45 — the widest gap between a player’s draft number and his win share ranking that year.
The average variance for each player in 2003 is 13.5, which makes it the 22nd most accurate year of the 25 in our data set, which runs from 1990-2014.
For each year, we also looked at how many picks in the top 14 (now the lottery) had careers that ranked as good as or better than the spot in which they were picked. Again, no upward trend.
Busts at the top of the draft are just as common these days. So, too, are hidden gems that fall. There are misses all over the draft, occurring at similar rates and magnitudes.
The most accurate draft in our data set is 1992, when each of the top seven picks, led by Shaquille O’Neal, ranked in the top 13 of the class in career performance.
Ten years later, 2002 provided the least accurate, as teams collectively undervalued longtime pros Tayshaun Prince (No. 23 pick), Carlos Boozer (No. 35) and Matt Barnes (No. 46) while whiffing on flameouts Jay Williams (No. 2), Nikoloz Tskitishvili (No. 5) and Dajuan Wagner (No. 6). In fairness, injuries are often unpredictable and can damage a player’s career before it really gets going, as was the case with Williams and Wagner.
The system is inexact but gives a big-picture look at drafts over the years and reinforces observations made by people who have played a role in the selections.
Though eligibility rules have changed — most notably in 2005 when the NBA began requiring players to be at least 19 and a year removed from high school — the basic procedure of scouting has remained fairly static over the past few decades.
After watching players in college or overseas, teams use the NBA combine and private workouts for physical and psychological tests. They hire private investigators to perform thorough background checks.
But it’s rarely so easy as taking the best available player. Teams must consider how well he’ll fit into a system, whether he fills a need and if he can help immediately.
And because mediocrity might be the least desirable state for NBA teams, they often gamble on high-risk players rather than the more certain bet who might be less sexy.
“What’s stayed the same [over the years], we have fallen in love with upside,” said Bobby Marks, a front-office executive with the then-New Jersey Nets from 1995-2010 and now an analyst for ESPN. “We tend to shy away from something that we know will be a consistent product for five to 10 years, where we’re looking for that home run pick that can keep you employed for a long time.”
Franchises often have competing interests. A coach, fighting for his job, might want to win right away; a GM could take a longer view; an owner might seek a big-name player who excites fans.
“We always tell our owners, don’t watch the NCAA tournament,” Marks said, since it’s easy for casual basketball observers to be swayed by performances on a big stage.
Perhaps the most infamous instance of an owner exerting his input — because it was caught on camera — came in 2014.
As their No. 8 pick approached, the Sacramento Kings considered guard Nik Stauskas, the leading scorer on a Michigan team that had recently reached the Elite Eight.
“For me, Stauskas,” owner Vivek Ranadive said to his war room in a moment captured by the now-defunct website Grantland, which the Kings allowed to document their draft night.
Then-general manager Pete D’Alessandro parroted him. And Ranadive, smitten with his decision, said Stauskas’s name two more times to lock it in.
Stauskas didn’t pan out, and Ranadive’s persistence has become something of a joke to basketball fans, a classic case of the groupthink teams strive to avoid.
Groupthink comes in various shapes, especially in the NBA, where teams’ top decision-makers often have similar backgrounds.
“They’re talking to each other throughout the year,” said Massey, the Penn professor. “The opinions are so highly correlated that you’re not getting eight or 10 opinions; you’re getting one or two opinions.”
Laugh at the Kings, but they knew this in 2014, and with that draft approaching, they solicited outside help. Introducing a contest on a Reddit chat, they invited “the best analytical minds” to devise a methodology for evaluating the upcoming draft class — wisdom of the crowd.
Clearly, it didn’t work. And it’s not just a Kings problem.
Despite the Celtics’ success in recent years, Austin Ainge, Boston’s director of player personnel, admitted on a panel at the MIT Sloan Sports Analytics Conference in February that he’s often stumped by the draft.
In 2012, the Celtics were trying to scout Austin Rivers, whose father, Doc Rivers, was the team’s coach at the time. “We knew him as well as you can know a prospect in this draft,” Ainge said on the panel, appropriately entitled Draft Day Analytics. “We still had no idea how good he was going to be.
“We’re not good at projecting anything — in economics, in sociology, in psychology. I think it’s really hard to predict human behavior in any field, no matter how much data you have.”
Perhaps Ainge is playing coy: The Celtics have hit on their past four drafts, picking players who turned out to be key performers to the team’s surprising run to this season’s Eastern Conference finals.
The Kings, meanwhile, choose No. 2 this week, a spot that has produced an all-star in 10 of the 25 drafts we studied. Maybe they’ll find an elusive superstar. Maybe they’ll add to the unflattering list of Thabeet, Mayo, Flynn, Oden and Bennett.
If a quarter century of drafts is any indication, the odds are not in their favor.