I recently read an interesting paper, entitled “Profiting from Machine Learning in the NBA Draft (paper available here).” The author of the paper is Philip Maymin, Assistant Professor of Finance and Risk Engineering at the NYU School of Engineering. Maymin has written several articles applying machine learning techniques to NBA basketball.
Here’s the study’s abstract: I project historical NCAA college basketball performance to subsequent NBA performance for prospects using modern machine learning techniques without snooping bias. I find that the projections would have helped improve the drafting decisions of virtually every team: over the past ten years, teams forfeited an average of about $90,000,000 in lost productivity that could have been theirs had they followed the recommendations of the model. I provide team-by-team breakdowns of who should have been drafted instead, as well as team summaries of lost profit, and draft order comparison. Far from being just another input in making decisions, when used properly, advanced draft analytics can effectively be an additional revenue source in a team’s business model.
Based on The Machine’s* projections, we’re going to discuss some choice decisions the Wolves made in past drafts.
Specifically, we’re going to talk about a few of the picks the Wolves made, the picks The Machine says the Wolves *should* have made, and the picks *we* would’ve made. Some of these picks aren’t of much interest because the Wolves drafted a shitty player when only shitty players were available. In that case, who cares who The Machine would’ve told us to draft? But some of the other “predictions” are more interesting, totally wrong, or at least controversial. Those are the picks we’re going to discuss. (Eds. Note: Patrick J and Andy G *get* that The Machine is spitting out the “should’ve been” picks partially based on salary. They don’t factor salary into their discussion below.)
Patrick J: Okay, leading things off. 2005. The Wolves picked Rashad McCants at #14, somewhat surprisingly. McCants was more generally projected to go somewhere between 17-20 We know in hindsight that Shaddy McCants was a terrible, terrible, turrible (Barkley voice) pick by the Wolves. Danny Granger would’ve been a better one. As horrid as he is now, Granger was an all-star in 2009 and a 20 ppg scorer for four years as one of the lone shining lights in Indiana at the time.
But now it’s 2014. Shaddy is out there somewhere, possibly seeking *more* roles in semi-pornographic films with Tracy Lords.
And for Granger, he is currently riding the bench for the Clippers.
At the same time, Green, after almost a decade mostly spent in pro basketball purgatory, is coming into his own for Suns: He was fourth in the Most-Improved Player balloting this season, and averaged 15.8 ppg en route to being the most pleasant surprise on a team that was collectively the NBA’s most pleasant surprise.
So my question to you: Knowing what you know now, do you go with the conventional wisdom (and The Machine) and take Granger? Or do you take the extremely long view, and develop Green for a decade? What’re the tradeoffs?
Andy G: Knowing what I know now, I choose option C: I STILL draft Shad.
Hear me out:
Okay, fine. I couldn’t even get started on this.
Of course I draft Granger. He wouldn’t have brought a championship to Minnesota, but the Wolves would’ve had a 1-2 punch that would’ve kept Kevin Garnett here for his entire career — his preference at that point in time. There would have been playoff appearances and who knows how everything else plays out.
With Shad, we got an enigmatic (putting it mildly), sometimes entertaining and almost always frustrating dude whose delusions about his skills rival the greatest of all time. He was born to be hated, but dying to be loved. He was a terrible defender. To be fair, he was also pretty fun to watch that first year after Garnett left. Credit where it’s due, and all that. I don’t know what happened to him after that, but he no longer looked like an NBA player. Maybe Alexey Shved came down with the same arthritic condition.
Andy G: I’d like to know The Machine Model’s rating of the Wolves’ other second rounder in 2008; the one that they sold to Miami (who then selected Mario Chalmers) when DeAndre Jordan was still on the board. I like to beat this dead horse because it’s the one time that I hated a Timberwolves draft decision and was later proven right. (In hindsight, I didn’t have nearly enough hate for the Foye, Brewer, Wes, and Derrick picks. On Flynn, I wasn’t as enraged (because we just miraculously landed Ricky a few minutes earlier) as I was confused. But yeah: The Wolves were morons for not drafting Jordan when he was still on the board in the second round. He was basically a less skilled Dwight Howard. And Dwight’s greatness doesn’t depend on skill.
Patrick J: You’re preaching to the choir on this one. I wanted the Wolves to trade up into the mid- to late first round to take Jordan when he slid further than expected. His potential was obvious, but teams were afraid to take him. I don’t think the Clippers have regretted taking him at 35. The Wolves giving away the chance to take their pick of Jordan or Mario Chalmers for pennies on the dollar (almost literally) still bugs me.
Andy G: In 2006, the model picked Paul Millsap. The Wolves picked The Rhino.
Was there a stocky power forward mandate? I guess they should’ve gone with Millsap.
Patrick J: Pretty sure the jury’s still out on this one. Be patient. Good things will come.
Patrick J: Wtf is up with The Machine telling the Wolves to pick Josh McFuckingRoberts in 2007 (when the Wolves selected one of my favorite players ever, Corey Brewer) instead of Joakim Noah? (!) The Machine just lost all credibility in my eyes! (Eds. Note: Noah might be All-NBA first team this year. McRoberts is known mostly for looking like all many indie rock bassists and clotheslining LeBron in Game 2.)
So what’s up? Does The Machine have a fetish for Cobain hair?
Andy G: Note sure about Cobain hair. Maybe. But Steph Curry over Jonny Flynn makes sense. I think The Machine beat Kahn on that one.
Patrick J: Score one for The Machine.
HOWEVER, The Machine says the Wolves should’ve taken Shane Larkin instead of Shabazz Muhammad last season. Shane Larkin?
Andy G: Umm…I’d rather have Shabazz. Wait, is Larkin actually decent?
Patrick J: It doesn’t matter. (Eds. Note: Shane Larkin’s statistics are here.) What matters is this: The Machine’s findings suggest The Machine liked Larkin more than Gorgui. Wtf?
Andy G: The Machine lost that one. Which means Flip Saunders *beat* The Machine. My whole life is a lie. Or I’m just overrating this Machine I know nothing about.
ALSO: I wish we didn’t go with The Machine on D-Thrill. It was wrong. So were we.
Julyan Stone over Tanguy Ngombo makes sense, if only because ol’ Tanguy was not eligible to be selected when David Kahn selected him. The Machine knew better. But hey, he redeemed himself in the waning moments of the very next draft, selecting Robbie Hummel — a decent role player — over The Machine’s Kevin Jones. (Eds note: Who is Kevin Jones?)
The Machine had us taking Hassan Whiteside (PDW Fave) and Jarvis Varnado with consecutive picks. Was this a Rhino-Millsap thing, where it got stuck on a specific body/player type?
Serious question: Is The Machine anti-foreigner? Compare the left and right columns and you tell me.
Patrick J: I think you’ve just put your finger on The Machine algorithm’s “special sauce” not discussed by the author anywhere in the paper – it’s the proprietary variable that’s going to net him an NBA job!
And that’s the truest form of Computer Love.