Sign in / Join

Officiating Errors & The Role of "Models"

Recent poor calls in collegiate football and MLB (Yankees-Angels video) have generated a lot of internet traffic. While all "human error" in a generic sense, specific factors probably make such errors more or less likely such as weighting officials down with too many rules to enforce (TSE-Why Good Refs Make Bad Calls) along with variation in evaluation and incentive systems across leagues (TSE-Tail Wags Dog in Some Leagues).

The episode in the Yanks-Angels game brought to mind the role of "anticipation" in these errors -- making a call based on a "model" of the situation that the mind builds based up on the most easily observed or common "inputs" rather than the final result itself. In the Yanks-Angels scenario, the usual inputs are two runners tagged while one or more occupy the same base. Instead of the usual, both runners stood off the base (by an easily observable fairly large amount) but it's likely the ump's mind had already plugged in the usual inputs.

Official instruction, training, and discussion warns against such "anticipation." The difficulty for officials is that anticipation of unfolding events is not all negative. Instructional guides for officials encourage it when discussing getting in a better position to make a call. (Here's a soccer example). In John Feinstein's book, A Season Inside, collegiate ref Joe Forte and others emphasize moving so as not to get "straight-lined." Feinstein, in fact, thinks one of Forte's gifts as a referee is his implicit awareness of where to be.

Whether readily admitted by officials or not, I would conjecture that they use anticipation (easily observed inputs) to cut down on uncertainty of highly dynamic and hard-to-observe outcomes. RFor example, the seemingly widespread use of the "ball-beat-the-runner" in MLB tag plays. Tag plays are very hard to see in real time. Observing when the ball arrives relative to the runner's position supplies a predictable model that may improve the percent correct over the long haul versus trying to plug in all of the noisy observable data.

A second conjecture: officials employ such "modeling" of plays less often when subject to greater scrutiny such as plays at home, or possibly playoff games. Doing so does not increase the average accuracy of calls -- after all, that's the point of using the model in the first place. However, it may reduce the variance of calls, or, at least, reduce the number of egregiously bad calls.