A scene in the 2011 film “Moneyball” shows Oakland Athletics general manager Billy Beane and assistant general manager Paul DePodesta — Peter Brand in the film — fighting with their scouts on which players to recruit out of free agency. Beane argues that, instead of replacing three star players the team lost, they need to find three guys whose on-base percentages average out to the same number as that of the guys the team lost.
The scouts balk at his suggestions, citing a number of on-the-field and off-the-field factors that turned them away from the players. Beane, undaunted, asserts that he wants these guys because they get on base. His system is all about buying wins, not players. To buy wins, you need to buy runs. To get runs, you need guys on base. And as Beane puts it, “(He) gets on base an awful lot for a guy who only costs $285,000.”
The film frames a debate being played out across sports about whether fidelity to stats and numbers is a more effective method of coaching and recruiting than looking for intangibles and eyeballing players. Ultimately, Beane and DePodesta’s method of sabermetrics took hold of baseball, as the film’s end card says, and the Boston Red Sox used that method to win the 2004 World Series.
This debate is especially apparent in sports journalism. Any given story is going to be either a stack of numbers or a sweeping epic of imagery and imaginative language. The way people think about sports tends to split along this divide.
The effectiveness of Beane and DePodesta’s thinking is debatable outside of baseball, primarily due to baseball’s idiosyncratic gameplay and profound statistical potential. One sees its spirit in other sports. The way rankings are collated in college sports today both exploits this dichotomy and shows how it’s more of a blend of the two schools.
There tend to be two major rankings in college sports: the vote-based poll — AP polls for football and the American Volleyball Coaches Association poll for volleyball — and the rating percentage index. The RPI is the end product of a formula taking different win percentages into account, and tends to provide a single number that can be ranked.
For volleyball, these two polls go into determining the tournament seeding for the postseason. The RPI poll tends to influence the selection committee more than the coach’s poll, according to volleyball head coach Dave Shondell, but the bigger problem is a lack of familiarity with the sport.
“In the past, because our committee has not been really volleyball-inclined, they didn’t really want to try to evaluate teams on their own with their eyeball and make decisions,” Shondell said. “Now, we’re getting more volleyball coaches on that committee, senior women administrators, that used to be volleyball players.
“I think we have some people that have that ability to look at a team from one league and compare it to a team from another league, despite what their RPI might be.”
Shondell also talked about “fooling” the RPI, intentionally scheduling good teams from lower conferences to inflate your own perceived winning percentage. There are teams with high RPIs who don’t receive votes in the AVCA poll because the coaches are looking past the numbers and seeing the opponent.
More often than not, we see this mix of these two schools of thought, with people bringing up stats to back up intuitive judgements, or intuiting a conclusion out of a stack of numbers.
People bring up the idea that a team “hasn’t played anybody” when they dismiss records, and bring up Strength of Schedule rankings to prove it. At the same time, someone could sift through a basketball team’s box scores and get the idea that they rely heavily on volume shooters above position players.
Whether adherence to the box score or riding the winds of fate is a more effective method for athletic success remains to be seen. As technology rapidly supersedes any human ability to see and register the details of a game, we’re left thinking that something’s got to give.
Either humans have to stop worrying and love the numbers, or the machines have got to learn that you can’t teach hair like Matt Haarms’. Maybe the balance will hold. After all, if a machine can learn the value of a tall quarterback, who’s to say we can’t too?