Review of “Astroball” by Ben Reiter
I’m a lifelong Houston Astros fan (spending more than I care to admit on standing room ALCS game 7 tickets last year) and especially love that Houston sports teams are leading analytic-based revolutions in how their games are played. So I had to read this fun book about the their multi-year path to the 2017 World Series, written by the journalist responsible for my favorite magazine cover and story of all time, a (ultimately correct, of course) prediction in 2014 that the Astros — then one of the worst teams in the league — would win it all in 2017.
Reiter (briefly) tells the life stories of the most prominent Astros responsible for the win — Carlos Correa, Jose Altuve, Alex Bregman, Carlos Beltran, and Justin Verlander.
He also, in more detail, tells the stories of the General Manager Jeff Lunhow, his assistants, and the decision-making culture they built. This culture blended data science with more qualitative assessments from scouts. I think the focus on “decision-making” vs “analytics” — the former indicating room for qualitative methods — is the correct one, and I liked that the book emphasized it.
However, the book focused more on the story and approach than on any specific techniques; this isn’t a book that will teach you about baseball statistics. Several interesting tidbits:
- Lunhow and his assistants had unusual backgrounds (consulting, engineering) but they weren’t just analytic nerds. They loved baseball and dominated their fantasy teams before getting their big breaks into baseball front offices.
- The astros have a system that rates scouts, supposedly containing what types of skills they are good at identifying and which skills/types of players they tend to over or under-estimate. I’m a bit skeptical that such a system can overcome the data challenges (high dimensionality of relevant player features, censored data in terms of which players ever get a shot, the delayed feedback of observing player outcomes, and the raw number of players a given scout may observe or rate), but if it works that’s a great example of “rating the rater” techniques.
- The Astros had a huge miss on J.D. Martinez, releasing him weeks before he became a star for Detroit. Partially in response, they invested heavily in a system that detects and projects player improvements and potential improvements, in surprising detail. For example, for pitchers, they look at the spin rate of their various pitches. Then, the Astros can figure out which new pitches the pitcher can learn, or otherwise shift how often the pitcher throws each type of pitch. This system has helped them improve their own minor leaguers and identify trade targets — such as Gerrit Cole most recently — who would improve under better coaching/management. I think there are lessons here for freelancing platforms such as Upwork, who care a lot about their freelancers potential for higher-paying gigs, but the book doesn’t go into enough detail for me to identify specifics.
- Carlos Beltran — team elder — united the clubhouse (historically split culturally between Spanish-speaking and English-speaking players) and mentored younger players. The book gave some lip-service to studies trying to quantify chemistry and the role of glue players, but I’m skeptical in general of the ability to measure such things and, more importantly, to identify in real-time through data which players would most help clubhouse chemistry.