Why you should care
Because game theory is well-suited to games.
The World Series has had many dramatic moments, but few rival the made-for-Hollywood drama of Kirk Gibson’s dramatic home run off Oakland A’s closer Dennis Eckersley in the 1988 World Series. The Los Angeles Dodgers slugger had a pulled left hamstring and a swollen right knee and could barely swing the bat, but he also possessed a critical piece of information.
Eckersley had dominated American League hitters all season, but he had fallen into a pattern — a tendency to throw a backdoor slider in 3–2 counts to left-handed batters. Gibson had learned this from a scout named Mel Didier, who’d watched the future Hall of Famer pitch more than 20 times that season. Eckersley, of course, threw the backdoor slider to Gibson, and the ball ended up in the right-field bleachers and etched into baseball history.
In baseball’s perennial chess match between pitcher and hitter, such information about an opponent’s tendencies is invaluable. And so too is the effort to avoid engaging in such pitch patterns, especially in a game now swimming in data and in-house sabermaticians who chart players’ every move on the field. Thanks to the sport’s Moneyball revolution, batters are being fed more information than ever about pitchers’ tendencies, and pitchers are getting better at subverting batters’ expectations. But are they missing the ultimate tool for stymieing hitters? Would they be able to call an even better game if they instead left the pitch selection up to … chance?
Even the act of changing one’s patterns is itself predictable.
The brains of batters, like all human brains, are primed to recognize patterns and to make predictions about what will happen next. As Ken Arneson, a blogger who writes about baseball, brains and technology, puts it, “A batter can’t consciously decide, ‘I’m not predicting anything.’ The brain does this on autopilot.” The best pitchers try to defy these inherent predictions and avoid falling into ascertainable patterns. Of course, even the act of changing one’s patterns is in itself predictable. “The problem is humans are quite bad at being random,” says Yale economist Tobias Moskowitz, an expert on the impact of sequencing on decision-making. “They overthink and develop patterns they aren’t necessarily aware of that are predictable.” Could adopting a randomization strategy in pitch sequencing help forestall such predictability? If so, what would that look like?
The best way to attempt such a strategy as a pitcher, according to Moskowitz, is to look for some cue unrelated to the decision you are making. Rumor has it, he says, that Hall of Famer Greg Maddux, known as one of the game’s best and most cerebral pitchers, would sometimes use a cue in the stadium, like the time on a clock, to help inject some element of randomness into his pitch selection. So, for example, throw a fastball if the clock time ends with an even number, and a breaking ball if it ends with an odd one.
But is such randomization effective? So far, a grand theory of pitch sequencing has eluded sabermaticians, but there is some evidence that a more randomized approach could be beneficial. As Baseball Prospectus’ Robert Arthur has shown, location entropy — the amount of unpredictability in where a pitcher locates pitches — is strongly associated with more strikeouts when a pitcher manages to remain unpredictable in the strike zone.
A purely randomized sequencing approach, however, may not be ideal. Every pitcher throws some pitches with more effectiveness and accuracy than others. Plus, batters’ brains are wired to be more vulnerable to certain sequence changes, like a change in pitch speed. “If you completely randomize the distribution,” argues Arneson, “you’re going to end up throwing exactly the thing the batter is prepared to hit for far more often than you need to.” A more optimal approach might therefore be a weighted randomization model in which pitches are selected at random with appropriate weight given to factors such as pitch quality, game situation, the count and batter ability.
Such an approach might not be far off what many pitchers already do. “To me, all pitchers randomize, trying to keep the batter off-balance by changing speeds, locations and movements,” says Sam Carpenter, a Dallas-based coach and pitching guru who has tutored dozens of big-league pitchers. “They do fall into patterns, but that is because they have a go-to pitch that is their most effective.”
Carpenter also points out that there is one major natural source of randomness in pitching that most people forget about: human imperfection. Even Maddux with his pinpoint control still missed his target about 25 percent of the time; but, one of the things that made him so special, says Carpenter, was that he embraced this “random fallibility” and would even put his mistakes to good use by using them to set up his next pitch. Of course, in some cases, as with Eckersley’s backdoor slider, you never get a chance to deliver that next pitch or to make up for a mistake.