Review of “The MVP Machine: How Baseball’s New Nonconformists Are Using Data to Build Better Players,” by Ben Lindbergh and Travis Sawchik

Growth mindset meets baseball analytics. Ben Lindbergh and Travis Sawchik describe the (far more advanced) successor to Moneyball, in which players and teams are not just better measuring value but also creating it. By using modern imaging technology (especially a global shutter camera called the edgertronics), for example, pitchers can precisely see how a ball rolls off their hand with various grips, enabling the learning of new pitches in a summer. They can also intelligently change pitch compositions (e.g., throw more curveballs and fewer sinkers), and batters can adjust the launch angles at which balls leave their bat. Instead of trial and error, i.e., players can practice deliberately and implement specific changes.

Smart teams have overhauled their entire player development and acquisition program around such techniques. For example, the Astros infamously acquire pitchers and help them dramatically increase their spin rate, to devastating success. They have outfitted all their minor league parks with the edgertronics cameras and similar tools, and they deliver personalized insights to the most lowly of their prospects. It’s unlikely they would have won the World Series without these developments.

The book itself traces the origin of this revolution in baseball, in an anecdote driven manner. It especially focuses on the pitcher Trevor Bauer, who is obsessed with data-backed training techniques and first used the edgertronics to learn new pitches. Unlike most anecdote driven books, however, (perhaps unsurprising given the subject matter), the book is honest about the limitations in being able to attribute individual successes to changes influenced by the technology. It is also full of non-obvious tidbits, e.g.:

1. One of the most important growing roles on teams are “conduits,” former players who are analytics-minded and can translate the insights/lessons/data coming from the analysts into a language players want to hear and can immediately apply.
2. The Astros have basically killed scouting in their organization. Instead of sending scouts to watch players and write reports, they send interns and camera-people to measure potential players far more precisely, with “scouting analysts” studying the data offline.
3. While at an individual level improvements enabled by technology have enriched some players, the equilibrium effect on the league has been far more subtle. Baseball players don’t reach free agency until much later when compared to players in other leagues (due to “arbitration” years); by the time that they become free agents, they are often past their peak, and teams are realizing that they can develop younger, cheaper players to get close to their performance. There have been many calls to fix these issues.

The book delivers lessons beyond baseball. Something I’ve been curious about recently is how our education system and labor market can use such deliberate practice to achieve similarly fantastic outcomes. For example, online labor markets such as Upwork should be able to use the market-level data they have on client jobs, in combination insights on individual performance, to construct personalized education plans for workers. Of course, precise measurement of a person’s “mechanics” is a challenge, but one that can be overcome soon if not now.

It would not be the first time that innovations from baseball were transferred elsewhere. Sports have long been at the forefront of data decision-making, for good reasons: tens of teams — each worth hundreds of millions of dollars — are in ferocious competition, in a setting where outcomes (wins, points/runs) are easy to measure.

Baseball in particular has been the playing ground for statisticians for over a century, as it has several additional attractive features: (a) each team plays many (~162) games a year, (b) each game is a sequence of discrete events (at-bats), that each involve a small number of players, and (c) Every aspect about a single at-bat or pitch can be (and more recently, actually is) measured and stored, from pitch speed and ball launch angle to the at-bat’s outcome. These features make easy the gathering of relevant data — the most important challenge in most data science projects.

In one sense, the Moneyball era was the precursor to the modern tech company mantra of measuring and optimizing the right metrics; it’s no surprise that many prominent data scientists, including Nate Silver, cut their teeth in baseball before moving onto other domains. If this history is any guide, and with some luck, measurement-enabled deliberate practice will also grow in other domains.

In the meantime, I’ll continue enjoying the Astros’s success while their development edge lasts.