Feb 19 2013

Doug Collins vs. Science

Thoughts on the Sixers’ WARPed, PERplexing resistance to analytics.


That Osama bin Laden was shot and killed in his Abbottabad compound during a crisply, if imperfectly, executed 38-minute raid by a team of Navy SEALS in May of 2011 was a triumph of the peerless tactical capability of the US special forces, the commitment to OBL’s apprehension Barack Obama reinitiated after his election, and the collaboration of the nation’s sundry and scattered intelligence agencies. Chiefly though, it was a triumph of data.


-Photo by Keith Allison "Men who are born tall and handsome have little use for mathematics."

-Photo by Keith Allison
“Men who are born tall and handsome have little use for mathematics.”

Faced with a unique problem in the aftermath of 9/11—how do you wage war against an enemy you can’t see?—the US, after a few false starts, deemphasized conventional military strategy and embraced a more data-centered approach to dismantling al-Qaeda. It built a supercomputer. Each pertinent bit of intel collected in the field—a list of names found in a hard drive during a raid, call logs, the daily travel habits of mujahideen recorded from high above by Drone—was fed into the system, analyzed, parsed, cross-referenced, and fit into what grew to be a sprawling jigsaw puzzle; the aggregate effect of which slowly exposed the hidden networks within the terror organization. Signal from noise. It was in this way that a long-shot tip about a courier named Abu Ahmed al-Kuwaiti led the SEALs to the Khyber Pakhtunkhwa province of Pakistan that night.

“Information and intelligence is the fire and maneuver of the 21st century,” Lt. General Michael Flynn, now the head of the United States Defense Intelligence Agency, explained to Mark Bowden while the reporter was working on The Finish, hammering home a point that would become central to Bowden’s excellent account of the efforts that led to the killing of Bin Laden.

So math, it turns out, is pretty useful if you’re tracking a mass-murderer hidden away in Northern Pakistan. It’s also helpful if you’re trying to figure out whether to, say, re-sign Spencer Hawes.




In recent months, the Sixers have made encouraging noises about the value they assign to analytics. The ownership group, after all, made its money in private equity, where men don’t pass gas without first consulting a spreadsheet. GM Tony DiLeo, though he was hired over two candidates whose thinking about the sport was more deeply rooted in numbers than his own (Tom Penn of ESPN and Mike Zarren of the Boston Celtics), told reporters after he got the job that he wanted to usher in a “Moneyball”-like movement within the organization. “We’re going to try and bring someone in that’s an expert in statistics and analytics, just to give us a competitive edge over these other organizations that don’t do it or are not at that level,” DiLeo told ESPN in September. He made good on the promise, hiring the highly regarded Aaron Barzilai of BasketballValue as the team’s first Director of Basketball Analytics. The Sixers are also one of 15 teams that have invested in the StatsLLC system; a network of cameras situated throughout the arena that capture player movement in three dimensions, enabling management to quantify aspects of the game the box score misses. But despite these gestures, there’s a powerful bulwark at the organization’s front and center, standing athwart this coming tide and yelling stop.

“If I did that, I’d blow my brains out,” Doug Collins infamously told a Philadelphia Inquirer reporter after an October practice, explaining what he’d do to his brain if it started thinking about numbers. “I would kill myself.”

To clarify the point, Collins went on. “My analytics are here,” the coach said, pointing to his head. “And here,” he added, pointing to his trim gut.

It’s not obvious how much of this Collins actually believes and how much is bluster. And while it’s widely thought that the coach has final say on all personnel decisions, it’s not clear what the Sixers’ decision-making flow chart really looks like. What is abundantly clear though is its effect. If Collins and the Sixers aren’t an organization that eschews the new science of basketball, they’ve been doing a flawless impression of one.


Shots fired

The numbers have told us, in increasingly unequivocal terms, that 3-pointers and shots at the rim are the most efficient in the sport while 16-23 footers are the least. The league has evolved accordingly. According to Hoopdata, in the each of the last three seasons, 3-pointers and point-plank attempts have increased, while the “long-two” has fallen out of fashion. The Sixers, in the most generous interpretation, have been slow to respond to this.

This season, they rank 23rd in the NBA in attempts at the rim, 24th in attempted 3-pointers, and, despite playing at a relatively slow pace, lead the league in shots from 16-23 feet—where they take 29.2 percent of their attempts. (Bafflingly, they shoot 33.9 percent from this range; four percentage points below the league average and good for 27th place. The Sixers, in other words, shoot more of the worst shots in the game than any team, while converting these already low-percentage opportunities at a lower rate than all but three of them. Insert your expletive of choice here.)

As Kirk Goldsberry has explained, Kevin Durant’s rise to superstardom is attributable not just to his long hours in the gym and consequent improvement in field goal percentage from every area of the floor, but also to smarter decisions about where he shoots from. Durant, in Goldsberry’s telling, gives up good shots in favor of great ones. The Sixers just take long twos.


It’s not personnel, it’s just business

In the area of player evaluation, the various popular metrics (PER, WARP, Win Shares, Wins Produced) are often in disagreement, which makes the Sixers’ recent roster moves all the more incredible. They’ve managed to sign players who don’t just grade out poorly in one or two of these systems, but in all of them.

Take this past offseason. The players the Sixers added over the summer who have logged minutes—a group that includes Nick Young, Dorell Wright, Kwame Brown, Royal Ivey, Jason Richardson, and Damien Wilkins—had, in the three years preceding their arrival in Philadelphia, combined for one season of a PER over 15 (15 is the league average), four seasons of a WS48 greater than 0.1 (0.1 is league average), and seven seasons of a WP48 greater than 0.1 (0.1 is, again, league average). Four of the six—Nick Young, Kwame Brown, Royal Ivey, and Damien Wilkins—posted aggregate WARPs of below replacement value in the three years before the Sixers signed them. In other words, the team brought in a group of players who were, depending on your metric of choice, considerably below league average in somewhere between 61 and 95 percent of recent seasons.

Andrew Bynum complicates the picture somewhat, but hardly exonerates the Sixers’ decision makers. The team gave up what looks, in retrospect, to have been significant value—Andre Iguodala, Nikola Vucevic, Maurice Harkless, and a lottery-protected first round pick—to get the center. Furthermore, the popular argument that the Sixers are struggling because they assembled a team that was designed to function around a player who, through no fault of their own, hasn’t been able to get on the floor, doesn’t stand scrutiny either, as only Damien Wilkins was acquired after the August 10 deal. Though it was the right move, given the information that was available at the time, the old saying “even a broken clock is right twice a day” comes to mind.

So, minus Bynum, the two banner acquisitions were Jason Richardson and Dorell Wright, who combined for the only positive PER performance, three of the four positive WS48 seasons, five of the seven above-average WP48s, and the lone above-replacement WARPs. Richardson is 32, and now out for the year with a knee injury, while Wright only plays 21 minutes a game. Also: the two banner acquisitions were Jason Richardson and Dorell Wright.

The Sixers, at the All-Star break, have the seventh-worst scoring differential in the NBA.




The fashionable line to trot out among humble analytic thinkers is that the stats vs. scouting debate is a false choice. The two schools of thought aren’t mutually exclusive: they’re complementary. There’s wisdom in this. Numbers can tell you what’s happening, but they can’t always tell you why. Is Player X struggling because he tweaked his ankle three weeks ago and it hasn’t fully healed, or has he simply lost a step and begun the long, slow decline all athletes eventually face? Will College Player Y be able to finish at the rim like that against NBA competition? Is Player W’s performance a function of skill, or are his weaknesses being camouflaged by a savvy coach with a well-designed scheme? These are consequential questions, and ones that scouts, with a deep study of the tape, are uniquely well qualified to answer. But while it’s certainly true that the smartest organizations incorporate ideas from both camps, the split-the-difference line obscures something important: analytics don’t merely have a place in the conversation, nor are they just as valuable as conventional scouting—they’re much more so.

In 1954, Paul Meehl, a professor at the University of Minnesota published a small book called Clinical vs. Statistical Prediction: A Theoretical Analysis and a Review of the Evidence. In what became a foundational work in the field of behavioral economics, the professor did a meta-analysis of 20 studies that tested the relative effectiveness of clinical predictions against simple statistical forecasts. Basically, he set out to learn who was smarter, the experts or the numbers. What he found was that across a broad swath of domains—criminal recidivism, heart disease mortality, collegiate academic performance—the numbers didn’t just predict outcomes more accurately than the experts, they won in a rout. Predictable outrage followed, but in the roughly 200 studies on the subject conducted since, Meehl’s findings have only been reinforced. (Meehl’s story was outlined in Daniel Kahneman’s Thinking, Fast and Slow.)

“There is no controversy in social science which shows such a large body of qualitatively diverse studies coming out so uniformly in the same direction as this one,” Meehl said 30 years after his initial paper.

Expert judgment is shaky foundation upon which to build an organization. It’s especially shaky when the expert in question is a 61-year-old with a history of concussions and health problems. In this way, the fate of the 76ers doesn’t merely hinge on Andrew Bynum’s wobbly knees, Jrue Holiday’s development, or the city of Philadelphia’s appeal to an elite cadre of young black men. It rests on an old, proud man making a choice about how much he trusts himself. And, maybe, deciding to let go.

  • Anthony
  • tsunnergren

    I saw that Anthony. Interesting contrast.

  • There are plenty of coaches that don’t use analytics at all. Im not making excuses for Doug just stating a fact. Doug may not be into things like WARP or PER but I’ve heard him mention things like +/- and percentages ie what shots are the best to take. To me if we could start by using analytics to figure out which players are actually valuable that’s a good start. Even if Doug doesn’t use them DiLeo should he hired that analytics guy. The main problem with Doug imo is the ownership gave him way too much juice. He is running the show. The only thing he should be doing is coaching.

  • Asher Steinberg

    The bit about the history of concussions and health problems is a low blow. Otherwise, nice post.

    • tsunnergren

      I think you might have point there Asher. What I wanted to express was that Collins, it’s possible, is not the same basketball thinker at 61 that he was at 51. Probably could have communicated that with less snark. Thanks for reading.

  • Pingback: 3sob.com | Eastern Connections()

  • Pingback: Hoop76 at the Sloan Sports Analytics Conference - Hoop76()