This weekend, I had the distinct pleasure of attending, along with much of my TrueHoop blogger brethren, the Sloan Sports Analytics Conference in Boston. I had an incredible time learning, chirping and even hooping with awesome people. Really my only complaint was that each presentation I attended came at the expense of hearing another interest piquing talk.
Luckily, other TrueHoop bloggers were on the scene, and dependably recorded the proceedings. As I wandered from room to room, and in reading the forensic accounts of my colleagues, I realized that the conference format naturally compartmentalized the mountain of presented information in such a way that ideas that will or do interact with one another were presented with little reference to each other.
But when attendees poured from the rooms into the hallways to recap and discuss what they saw, the ideas were put in conversation with each other. For example, Graydon Gordian did some great reporting from a panel on sports labor, while Rob Mahoney was covering the anticlimactic unveiling of proof that NBA players perform better in the year before a contract.
[The President of the Houston Rockets] pointed out that issues like revenue sharing, which is hotly debated among the owners, can handicap the owners’ bargaining ability. Focusing on the disagreements they have with one another can distract from the disagreements they have with the players. Although, as Zimbalist pointed out, their disagreements with the players can also be what enables them to put issues like revenue sharing temporarily aside.
“When the owners can’t agree on things like revenue sharing, they can agree they want the players to have less,” he said.
It seems like Arup Sen’s presentation entitled “Moral Hazard in Long-Term Guaranteed Contracts – Theory and Evidence from the NBA” would be more than a little relevant to the owners if one of their objectives is less guaranteed money in player contracts.
Sen convincingly argued that NBA players, especially younger ones, tend to play much better in the year before their contract expires than in the previous year. While not exactly revelatory, this suggests that players near the beginning of their primes play better going into their first major contract opportunity than older players looking to make one last push at a big contract. Sen did not postulate too much further on the implications for the upcoming CBA, but the benefit of shorter contracts was an obvious conclusion. Will owners take this hard data to the negotiating table?
Elsewhere, I found out that coaches use statistics more than one might assume. Sebastian Pruiti shared his surprise in his report on the “Coaching: Guts vs. Data” panel filled by, among others, longtime NBA coach Del Harris. Pruiti explained that the type of data that interests coaches caught his eye:
[Del Harris] also mentioned how he likes to receive situational data in game so he can use that information when making decisions on what defense or offense to run. In addition to this data, Harris mentioned that when he coached in college he used to have a student manager chart possessions for him. He used this data as a tool to measure momentum as he would have his manager let him know if his team would put together a string of stops or makes and vice versa and make decisions based on that. For example, he mentioned that if his team was on a run of makes, he wouldn’t call timeouts or make a substitution until the opponent took back momentum (the only exception to this rule was in the final two minutes when he would play his best players). This is interesting because momentum is usually considered this unexplained thing that coaches use to make gut decisions, but Del Harris tried to quantify it and use that data to help make coaching decisions.
This seems like a great point to insert Tarek Kamil of WhatIfSports.com into the discussion. On Saturday morning, I caught Kamil’s talk in which he presented his theories about the way artificial intelligence would influence the coaching profession. According to Kamil, we’re not only moving towards the kind of highly developed in game data desired by Del Harris, but programs that can make decisions for coaches using such data:
Kamil’s thesis is based on his belief that human coaches are “not optimally designed” to handle the tremendous amount of data they must process to make the correct decisions. He expects (as do many other very smart people) that computers will surpass human intelligence’s capacity for complex and abstract problem solving by 2040. Today, NBA team coaching staffs have become almost as large as the player roster in an effort to corral this data and feed decision making information to the head coach. No one person can completely keep track of the duties ascribed to NBA coaches. Managing substitutions, match-up strengths and weaknesses, player fatigue, opposition tendencies, play calling, etc. takes a village, and even then coaches make decisions that are easy to second guess in retrospect.
However Kamil predicts that in twenty five years or so, a properly programmed computer could process all the thousands of variables impacting an NBA game and spit out the optimal strategy or next maneuver. In this future, Kamil imagines coaches taking on more of a teacher role, rather than an in-game general.
Although the manufactured dialogues above weren’t formally held by the presenters themselves, the conference hallways became a forum for excited talk regarding the overlapping significance of their ideas. It’s likely I learned just as much gabbing with the assembled writers and NBA statheads as I did from the presenters themselves.
If anything, the weekend proved that analytics and data can’t be fully appreciated in isolation. It’s the discourse between other studies, variables, ideas, and people that ultimately offers the most compelling education.