There is no question that umpires have biases. The question is whether we want them or not.

A couple of years ago I reported on a book that I was reading titled Scorecasting by Tobias J Moskowitz and L. Jon Weerthem. This writeup is going to copy a lot of that story. The book is similar to the book Freakonomics, except that it focuses exclusively on sports. If you have an iPad or iPhone, you can download the first chapter for free – and just that much changed forever how I watch baseball.

The chapter is about a bias that umpires have because they are, with the possible exception of Joe West, human. And humans are far more willing to forgive an error of omission over an error of commission. That is, we are more willing to forgive an error caused by doing nothing over an error caused by doing something. And thus humans are for more willing to commit an error of omission over an error of commission, because it gets us into less trouble. I’ll give an example from the book:

“In a well-known psychological experiment, the subjects were posed the following question: Imagine there have been several epidemics of a certain kind of flu that everyone contracts and that can be fatal to children under three years of age. About 10 out of every 10,000 children with this flu will die from it. A vaccine for the flu, which eliminates the change of getting it, causes death in 5 or every 10,000 children. Would you vaccinate your child?”

Most parents opted to NOT to vaccinate their child, despite it halving the chances of their child dying. The thought of doing something to the child which would cause his or her death was worse than the though of doing nothing and doubling the chances of death. The same bias is statistically apparent in umpires when it comes to calling balls and strikes and now I can’t help but notice it.

In 2007, mlb.com installed the pitch f/x equipment in all the ballparks, providing data on 2 million pitches, including 1.15 million called pitches. Suddenly we could see from data how accurate umpires were in calling balls and strikes, and whether there are any circumstances that made them less accurate. It turns out there are.

A ball that is in the strike zone is called accurately by an umpire 80.2% of the time. But that number dives if there are two strikes on the batter (and it isn’t a full count). Then, a ball in the strike zone is called a strike just 61.3% of the time. He’s almost twice as likely to mistakenly count a strike as a ball. Again, don’t forget – we KNOW that these are really strikes from the f/x data.

The same thing happens the other way on pitches outside the strike zone on three-ball counts, though it’s not quite so drastic. A pitch outside the strike zone is called a ball 87.8% of the time, but if there are three balls (and it’s not a full count) it’s only called a ball 84% of the time.

The reason? Because calling strike three or ball four ends the at-bat. It’s active – it affects the game far more than giving the batter and pitcher another pitch to resolve the at-bat themselves. The incentive is toward the error of omission rather than that of commission.

Incidentally, this is most apparent on borderline pitches. Over all counts, a borderline is called a strike 49.9% of the time – almost literally a coin flip. But with a 2-strike count (again not a full count) it’s called a strike just 38.2% of the time. And with a three ball count, it’s called a strike 60% of the time. The percentages become even more extreme on 3-0 and 0-2 counts.

This may be a bias that we, as fans, want to reward. For the first time, I thought about whether or not I really want to take that kind of call out of an umpire’s hands. Don’t we want someone who prompts the batter and pitcher to resolve their conflicts themselves? Even if it might not be a perfectly accurate call.