There are times when online editions of newspapers deny readers the curious juxtapositions of stories found in print, unintended though they may be, that occasionally identify fascinating ironies in politics, current events or the social world. For example, an article about on sex education in schools may inadvertently be placed side-by-side with one about the struggles of some single mother somewhere. Or, a photograph of a public servant charged with taking bribes may be coincidentally situated above a story about a newly-discover Ponzi scheme.
In yesterday’s Sunday Globe, the print edition that is, a skillful and lengthy examination of efforts over the years to improve parole prediction tools culminates on page K4 of the Ideas section, just above a small snippet about the limitations on using minor league pitching performance to predict major league success on the mound. In both settings—predicting future criminality and future earned-run-averages—the past is a useful, but rather flawed, indicator of the future.
The most challenging problem in either case is the inherent difficulty in predicting rare events. It is a statistical fact of life that unusual or extreme outcomes can never be reliably anticipated, no matter how sophisticated the prediction device or how voluminous the data used to craft it.
Consider, for example, any effort to predict airplane crashes. While there are undoubtedly certain characteristics fairly common in airline calamities -- such as poor weather, small aircraft and an inexperienced pilot -- the vast majority of newly-licensed flyers taking off in a single-engine Cessna during a storm will land safely at their destination.
Of course, recidivism among parolees, like washout performances by rookie hurlers, is not a rare phenomenon. However, prisoners on conditional release who commit murder and top prospects who are a complete bust in the big leagues are what drive public and fan opinion. Both these extreme failures are largely unpredictable.
The other major problem to predicting parole success or pitching success is that much depends on external factors. A parolee who has the support of family and gainful employment will overcome a troubled past, just like a mediocre pitcher will be victorious if given substantial run support.
I do not mean to trivialize murder with the comparison to bad pitching. They are hardly comparable in severity or significance. Yet, the practical issues in foretelling the future in these two contexts are similar.
So before expecting too much of emerging and sophisticated tools for predicting criminality or baseball performance, we should keep in mind limitations that will can never overcome in trying to anticipate the future. Meanwhile, the public and fan base need to be more understanding when things just don’t turn out as the experts predicted.
The author is solely responsible for the content.