Scott’s Miscellany – More on High-Stress Pitches

The title of the article is an allusion to Schott’s Miscellany, which you should definitely check out if you never have and feel compelled to know that a group of larks is called an exaltation or that a member of the 32nd degree of Freemasonry is known as a Sublime Prince of the Royal Secret.

— More on High-Stress Pitches —

A couple of weeks ago, I wrote an article about Matt Harvey for which I did some research into how frequently he threw pitches this season in high-stress situations, which I defined as pitches during plate appearances with a leverage index of 2.0 or more. The inspiration for the article came from concerns over potential injury for Harvey, but as I reflected on the approach I took, I started to wonder whether high-stress innings might have implications on pitcher performance separate from injuries.

Here is the logic as it played out in my head. While good relievers will likely be called upon more frequently in high-stress situations than bad relievers, starting pitchers should not see much consistency in the stress context of their pitches from season to season. In my mind, it was similar to run support. High-stress plate appearances happen to pitchers; they do not create them with their performance.

The harder part of the equation to wrap my mind around was the effect high-stress pitches should have on a pitcher’s performance. This dances dangerously close to clutchness. In an individual plate appearance, high stress may impact the performance of the pitcher, but it would also have an impact on the batter, and I did not have any reason to suspect that the pitcher would suffer more or less from the stress than his opposition.

The reason I could convince myself that this might matter is the notion of a cumulative effect of consistent stress. Sure, in that one plate appearance, a pitcher may see a small decline in his effectiveness that is offset by what the batter sees, but if that pitcher has to deal with those stressful situations start after start during a particularly unlucky season, might that tip the scale to his detriment?

To investigate, I ran a series of correlations. For qualified starters in consecutive seasons between 2002 and 2015, I tested the correlation strength of their rate of high-stress pitches—pitches within plate appearances with a leverage index of 2.0 or more—from one season to the next. There was next to no correlation (0.02). That supported my hypothesis that starters do not see consistency in the context of their pitches from season to season.

Next, I tested the effect of a change in high-stress pitch rate from season to season on ERA. The resulting correlation of 0.07 is very small. Correlation coefficients range from 0.0, which represent no correlation, to either -1.0 or 1.0, which represent perfect inverse and parallel correlation. As an example, ERA for the same sample of pitchers had a correlation strength of 0.37 from year to year. Still, the correlation of high-stress pitch rate to ERA did exist, and its direction matched intuition. Starters who saw an increase in their high-stress pitch rate from one year to the next also saw an increase in their ERAs and vice versa.

Looking at the biggest risers and fallers in high-stress rate from 2014 to 2015, you can pick out examples of pitchers where changes in stress context could help explain changes in their performance.

Among the risers, Jon Lester, Julio Teheran, Garrett Richards, and Jordan Zimmerman all saw their ERAs jump by close to or more than a run this season compared to last.

Among the fallers, Jake Odorizzi, John Lackey, and Shelby Miller all saw their ERAs drop significantly compared to last season. However, you can also see that pretty much as many pitchers showed the opposite trend, and other changes in context—like team and/or league changes for Lester, Lackey, and Miller—and randomness no doubt have played a part in these changes in effectiveness.

Based on the findings from this initial testing, stress pitches still interest me enough to continue my research of them this offseason. I think the next step will be to see if adding high-stress rate as a variable to ERA estimators like FIP improves their predictiveness.