Published in the San Diego Union-Tribune, November 14, 2016
The Cubs have won the World Series. What role (if any) did the umpires play in the outcome — in other words, is there bias in baseball? To explore this, I consulted with Professor Etan Green (Wharton) and his new paper, “What Does It Take to Call a Strike?” The subtext of his work is that even with decision making by experts (the umpire behind the plate) “there are opportunities for statistical discrimination everywhere.”
So we all have seen the imaginary strike zone that appears on the replay to see if the call by the umpire is correct according to the stereoscopic camera algorithms that show the ball’s location in the strike zone.
Green has looked at a zillion pitches and counts and he finds that the umpire often trades off accuracy for bias. When the count starts at zero / zero — the umpire calls the pitch right on the boundary 50-50 ball and strike, but when the count favors the batter–when it is three balls and zero strikes, and the pitch is right on the margin–it is almost always called a strike. In other words, even with experts behind the plate, the bias tilts on the boundary pitch depending on the count.
What is happening according to Green is the “umpire is trading off accuracy for bias or bias for accuracy. Here is the kicker: Even though he is not making his decisions based solely on the location of the pitch, “this behavior actually helps him make more accurate calls.” This is because the umpire has “expectations about where the pitcher is going to throw at a certain count and whether the batter is going to swing.”
In this column we have often talked about pattern recognition and how that favors entrepreneurial decision making — the “I have seen this movie before” moment. This has huge relevance when it comes to negotiation and to financing. It is the expectation factor that sets up the negotiation, because you are applying everything you have learned from the previous 100 negotiations. It is the “how it is likely to resolve itself” based on not just factual financial data, but also bias — “they want to make the deal, so hang in there.”
In the baseball discussion, when it is likely that the pitcher is going to throw a strike (e.g. at three balls and zero strikes), then the umpire calls the close one based on the pitch he would expect to be thrown — namely a fastball strike down the middle.
The same is true when the count favors the pitcher and the boundary pitch comes in — and it is to a known good hitter. If the hitter does not swing, that tilts the umpire to calling it a ball.
Green calls this “Bayesian updating” which is a fancy way of saying that the umpire is “statistically discriminating” — meaning they have other prior information to put into the mix beyond the exact location of the pitch (which in fairness arrives in about four tenths of a second) – inferential statistics.
Now how does this relate to the entrepreneurial adventure? Take hiring. We do not just look at the resume, we look at fit, previous history, as well as race, sex and age (which are illegal). We discriminate — on the boundary pitches. We apply probability inference. And the dark sentence is that it actually helps us get it right more often (the right person in the right seat on the right bus going in the right direction), even at the cost of our own bias.
This is both problematic and fascinating. I want to have a system for hiring that is honorable and true — but I am human and I have bias, and if you believe the baseball story, my bias (my propositional and probabilistic logic) actually works in my favor — in the favor of the company.
But the final question is this. Does it also work in the favor of the applicant? Maybe the seats on the bus are not all the same size. Has the hiring bias served a mutual benefit?
I am not willing to go hard on that one. But it is certainly worth a short walk on the beach to consider.
Rule No 486: Regardless, three strikes and you’re out.