In just under 60 days the Golden State Warriors will enter the NBA draft with the prize of the second overall pick. What they do with that pick could set the course of the franchise over the next half-decade, and possibly longer.
So how will they approach such a momentous decision? I recently spoke to Warriors Assistant General Manager Kirk Lacob to find out.
Who is the best player available?
One of the most over-used cliches about the NBA draft is that teams should just take “the best player available”. But working out who that player may be is not that simple. There are layers to this seemingly simple question. As Lacob put it “making a draft pick is not always as straightforward as who is the best player today. The question is sometimes, will this player be able to help us? When will this player be able to help us? And ultimately what will this player be? We have to be able to kind of straddle the line and figure out, of those three pieces, which is most valuable to us,” before adding “They’re all valuable to us, they all should be considered, but what is most valuable to us right now? The math is not always that simple.”
For different teams the balance between these different questions and how they assess the risks involved will be different. For the Warriors this is a particularly tricky conundrum compared to a team who might normally be picking this high. Golden State has to weigh bringing in a player who can help them win a championship in the window they have over the next few seasons with Stephen Curry, Klay Thompson, and Draymond Green versus selecting someone who might develop into their next star after that trio pass out of their primes.
What makes it even trickier is there is no sure-fire franchise cornerstone in this draft like Luka Doncic or Zion Williamson. Instead there are a handful of high-risk, high-reward prospects like Georgia’s Anthony Edwards. If he is available the Warriors could select him hoping he can become a Bradley Beal or Donovan Mitchell-style player in three-to-five years time. But he might just as easily become Andrew Wiggins 2.0 or even a Dion Waiters-clone.
How much should a contending team be willing to bet on a higher potential ceiling that may never be realized against a young player who can genuinely help them win now but still possesses All-Star upside, like Dayton’s Obi Toppin? And even if a player like Edwards reaches a higher ceiling later on, is that ceiling actually high enough to make it worth passing on someone who can help a proven championship core win a title in the interim?
Limited data increases risk
To make an informed decision on anything this complex and important, you need as much data as possible. And that is in shorter supply than usual this time around. For starters the league’s draft process is likely to be heavily altered from what teams are usually able to do. The Athletic’s Shams Chanaria reported this week that in place of the normal draft combine, and the ability for teams to host workouts, the NBA will run a pre-draft process starting tomorrow (21st September) and distribute the results to teams. This will consist of league interviews of 10 questions, while prospects will travel to the “nearest team market for medical/strength/agility testing, 8 on-court shooting drills.” Crucially no team personnel will be allowed at these tests and drills.
That adds a whole new level of complexity and risk to the process. It also fits an unfortunate pattern for all NBA teams this year. Of the many data sources teams use in making any draft pick, almost all are truncated. “There’s the quote unquote IPS [in-person scouting] going and watching players. And that’s always a huge part of it.” Lacob says. “Of course we have less of that data than most years. We were only able to watch so many games before the seasons were cut short.”
Then “there’s the interpersonal part of it, which is huge – the interviews, the workouts, getting into the mind and the psychology of a player, understanding what drives them.” Lacob explains. “That’s a huge part and that also has been somewhat truncated.” The Warriors have been speaking with prospects over video interviews, but as Lacob stresses that while it still provides valuable insight, it’s just not the same as meeting with a player in person. “Normally we interview them, maybe we take them out to dinner and then you get them in a workout where you get to see how they work and how they respond to things in real time, and how they push through different obstacles.”
That leaves the analytics as the last remaining major source of data and insight. Lacob, who oversees the Warriors’ analytics work, is in his element here, telling me that analytics will be “huge in this case, because we only have so much data on the other side.”
For starters, analytics can augment scouting or video work, providing “ways to dive into strengths and weaknesses that you may or may not necessarily notice from watching film.” Lacob explains. “Often the data will help point you in the right direction, what to watch for more closely sometimes. We think a player is good at something, but we’ve only seen him live three times, and the data will say otherwise. And so what that tells you is you’ve got to go back and you’ve got to watch a little closer.”
But the biggest element is what Lacob identifies as the Warriors’ projection models. He explains these models use “ historical numbers based on other players to create priors, and matching that with what a player is doing to create a projection.”
Projection models are probabilistic
Lacob is keen to stress though that they don’t give you a clear answer on what is going to happen. “In reality all of the projections are based on some sort of statistical likelihood of that projection happening. There’s a band of possibilities, because our projections are based on data that to some degree is going to be incomplete, and any projection model in literally anything, not just sports, would tell you there are outside factors that ultimately affect how well you continue to improve.”
Indeed the lack of workouts and in-depth one-to-one interviews feeds through into these models because “getting to know a player, what drives them, that helps us to understand which level of projection might be most applicable”. At the same time, those outside factors include unknowns like “what somebody is going to do when they have money or when they have a tonne of free time on their hands, or, frankly, how they’re going to grow up,” Lacob says.
Even with the data that are available, Lacob is quick to point out that “sometimes, you will get past production, as part of projection and it doesn’t necessarily mean what you think.” He gives the example of three-point shooting, saying “just because a player, say, shot poorly from three point range in college, does not mean that the projection is that they are a poor three-point shooter going forward. There are other pieces to that. Sometimes it’s the amount of attempts the player is willing to take, free throw percentages is a part of that, your ability to shoot off the dribble versus catch-and-shoot. There’s a lot that goes into projecting shooting.”
You also have to differentiate between an anomaly or genuine improvement in changes in a young player’s statistical output. Taking the shooting example, Lacob says “we will look at across a large number of indicators” if the sample size isn’t big enough. “Are they shooting a different variety? Are these coming in certain situations? How are they shooting in practice? Believe it or not how you shoot in practice actually matters for how you shoot the game.” Lacob continues. “Then it’s just that there is a human element to it as well. So, speaking with the coaches and the player themselves about their confidence about what they feel like they’ve worked on, what they’ve learned.”
The Warriors still have a momentous decision to take
Ultimately these projection models are all about balancing probabilities and helping inform judgements on how much risk you’re willing to take on any particular element of that all important “best player available” question. As Lacob puts it, “what it tells you is what a player could be, and the likelihood of that. At the end of the day you have to feel comfortable with taking the risk-reward profile. You have to understand that even though a projection model says one thing is more likely to happen than the other, it doesn’t mean the less likely thing won’t happen.”
The Warriors are well-known as an analytically-savvy organization and there’s no doubt this decision will rely heavily on their models. As Lacob emphasizes, “analytics is a huge part of our data tree and how we ultimately evaluate players.” Given the limitations on available data elsewhere and the importance of the decision that faces them, that’s never been truer than in their process running up to the 2020 NBA draft.