SSAC 2015 Takeaway – Accepting Yes For An Answer


Mike D’Antoni gets it (photo credit: USA Today)

As the analytics “movement” gains more converts, declaring victory and moving on from the Cold War of “jocks vs. geeks” are becoming simultaneously more important. The dichotomy was on full display in Boston, as categorizing teams, coaches and executives into camps of “believers” and “non-believers” shared time with anecdotes of successful adaptation and improvement among teams.

Even casual observation of the direction of the Sloan Sports Analytics Conference over the last few years would dampen expectations of any cutting edge revelations coming from the panel discussions. The two kickoff basketball sessions were certainly entertaining. But the educational value was far more about the basketball and professional lives of players and teams, rather than new methodologies or stories of integrated organizational structures.

Shane Battier talked about how “rest” days can get taken over by mandated public appearances and how the schedule all but guarantees a few no-show efforts a year. Darryl Morey and Jeff Van Gundy did some light-hearted sparring about things that can and can’t be counted. Khris Middleton was brought up as a modern day comparable for Battier[1. As the Milwaukee Bucks’ front office quietly cursed as the impending restricted free agents price probably just went up.].

Developments in analytics were barely even discussed. PER was compared to Real Plus-Minus, there was some discussion of biometrics[3. Battier was not surprisingly very wary about their use while Mike Zarren of the Celtics attempted to reassure that the interests of the team and player were in alignment.]. A joke or two was made at the expense of the Kings. But nothing earth-shattering.

Cutting past the storytelling and general quant boosterism[2. It is a trade show, and affirmations that the trade involved is super wicked awesome comes as part of the starter kit, across industries.], there was a discovery, in plain sight. If the question is how do “we” learn to take yes as an answer, the solution is to realize when we’ve met an ally. Mike D’Antoni is an ally. On one level, of course he is. The SSOL Suns and his general pace and space philosophy are well-regarded by new metrics. At the same time, he’s also been seen publicly as crusty, cantankerous, stubborn and so on — old school on everything but X’s and O’s.

But crucially, he’s willing to experiment. He likes things that work. The verbiage and thinking process might be radically different, but the underlying curiosity and ambitions are the same.

A (hypothetical) conversation with D’Antoni about trying a new numbers-based approach to an aspect of the game would probably start with him saying “I’m not totally sure what you’re doing, but show my why it works, how it can make us better?” That is what yes looks like. At this point, convincing stops and assisting starts.

An understandable and natural inclination is to realize someone is interested and to continue to explain, and over-explain and then over-over-explain. In effect, to kill the sale.

Less can be more. As Battier noted, players and coaches have very different thresholds for the amount of new information they can take on, process and implement in a short time. For Battier, a detailed scouting report on the offensive tendencies of Carmelo Anthony was useful and actionable. For others, “he likes to go left and pull up” might be as much as they can handle.

In the imagined conversation above, D’Antoni’s interest is piqued, his attention focused. There is a narrow window of opportunity to answer that question “how can it make us better?” A discourse on methodology is not appropriate. A league wide set of rankings isn’t useful. A way to score an extra point or save a basket every game, however? That’s the money.

The best practical illustration of this came from Seattle Sounders’ performance analyst Ravi Ramineni. After distilling a mountain of athletic performance data from training and matches, Ramineni and his team had developed what they thought was a pretty accurate injury prediction model, which resulted in a warning to the coaching staff that a star player had a roughly 20% chance of getting injured in an upcoming high-profile match. “Is that a lot?” was the response.

That number wasn’t actionable. No degree of explanation for the analysis and validation which went into that estimate would do much to make it so. To address this, Ramineni changed the output from a percentage to defining a set of six risk factors — a player subject to four of these factors would be coded as “red” or high risk of injury. The coach wasn’t told the player was unavailable to play, but he was aware of the high risk, which could be weighed in some ways against the importance of the next match. Moreover, the coach now knew that if the Sounders were up 3-0 at half-time, substituting the player would be wise.

Assuming the accuracy of the underlying analysis, six factors and color codes is less precise and exact than a percentage. But it is also substantively similar, and presented in such a way as the intended audience can use.

To bring it back to D’Antoni, he wants to win. He needs any analysis to support that goal. If abstracting findings to a higher and higher level is what is needed, the quant side needs to stop letting perfect be the enemy of the good and give him just that. To broaden back out myself, if acceptance is the goal, the next grand “theory of NBA everything” might be perfect in capturing the influence players have on the outcome of a game. But if that perfection is confined to a tiny corner of the basketball-obsessed internet, what was the point in the first place?[4. If you’re saying to yourself “Seth, I feel like you’ve said this before” you would be right, and I’ll probably be saying it again and again,]

If all of this seems like it could have been written as a dispatch from the Sloan ’14 or ’13 conference, that’s probably because it easily could have been. It will probably be true next year as well, because recognizing the “yes” for what it is will remain difficult as long as the question and answer are spoken in the disparate languages of “believers” and “non-believers[5. Progress is being made on this front. The co-winners of the research paper competition for their work on matchups and improving defensive metrics) openly admitted the need to incorporate more “basketball-specific knowledge and expertise” into the defensive model to transform a useful theoretical tool into something with practical applications.].”