Via BBTF, a study of the affect of exit velocity on the rate of errors by fielders turns up an interesting anomaly.

Your eyes are not deceiving you.

On the left graph, there’s a drop in error rate with increased exit velocity – on the right, there’s an increase in the rate of errors. That is, with identical data and only a slight shift in groupings, you could make either claim and feel justified. Errors appear to both increase and decrease with harder hit balls, and both contrasts are statistically significant (those are 95% intervals included with each point).

It turns out that bunts are very important in this calculation, and a generalized additive model (GAM) solves the problem nicely.