Pitchers are evaluated in countless ways. Beyond measuring all the ways their actual pitches perform, like spin rate and velocity, loads of statistics are used to determine how good or bad a pitcher is. For decades, ERA has reigned supreme as the premier pitching statistic. Arguably the most important question is answered with ERA. How many earned runs does a pitcher give up per nine innings? Other supporting statistics like walks-plus-hits per inning pitched (WHIP) and strikeout rate (K%) elaborate on how those runs may have been allowed to score.
These statistics come to life on the mound. Joe Ryan’s freaky fastball and Pablo López’s nasty changeup pile up strikeouts and help limit hits. Bailey Ober’s precision is reflected by his low walk rates.
Poor statistics also materialize in real life. Sometimes things don’t go according to plan, as was the case for Ryan this summer. His ERA inflated, his WHIP ballooned; there were a number of statistics that correlated with his actual play. We now know that a groin injury played a significant role in the struggles he faced. Still, Ryan’s summer is an example of ERA volatility.
In nine of Ryan’s starts, he’s allowed only one or zero earned runs. However, he’s also allowed five or more earned runs five times. He is a high-quality pitcher, but he encountered a rough patch.
Ryan isn’t alone. Pitchers will undoubtedly encounter slumps, given that an MLB season contains 162 games plus playoffs spanned across seven months. Through sheer length of time, players are bound to have poor stretches of starts. However, some pitchers may be more susceptible to turbulent outings. A couple of mistake pitches could quickly alter a pitcher’s outing from fine to horrible. Are there any statistics that can predict who those pitchers are most likely to be? Who would experience more volatility across their starts?
Pitchers who give up plenty of hard contact could fit such a description. Making the assumption that extra-base hits could equate to volatile starts, statistics like average exit velocity and hard-hit rate could be valuable predictors. Other statistics could provide this answer, so the focus cannot be entirely on quality of contact.
The answer to these questions can be found through the following analysis:
Using Baseball Reference’s Stathead, you can find a list of every start made so far this season. After grouping the data by each pitcher, you can perform the standard deviation of their earned runs in those starts. Standard deviation is a mathematical term that essentially describes consistency. It is a measure of how dispersed data is from its mean. In other words, how steady is a pitcher over the course of their starts?
There is at least one issue with doing the analysis this way; it assumes every start is equal. Sometimes a pitcher gives up three runs in nine innings. Other times, they give up the same three innings but in 1.2 innings before being forced to exit. Of course, this is undesirable. But I figured with enough pitchers in the dataset, I could arrive at a general consensus. Giving up a wide variety of earned runs while disregarding the innings they were scored in would still show start-to-start volatility.
Immediately, we can answer who the volatile pitchers are. Graham Ashcraft, Lucas Giolito, and Lance Lynn emerge as names that have wide results from start to start. Dane Dunning, Kodai Senga, and Tanner Bibee are the opposite. Remember, the standard deviation of earned runs does not answer who is the best pitcher; it simply looks at how consistent each pitcher is. Someone who gives up six earned runs every time they start would not be a very good pitcher, but they would be a laser-consistent one.
After looking at everyone’s standard deviation, the next step is pulling Statcast data from BaseballSavant and plotting them against each other. Specifically, average exit velocity (average EV), barrel rate, hard-hit rate, chase rate, and ground ball rate. Personally, I thought each of these statistics could play a role in earned run volatility.
It turns out that neither average EV nor hard-hit rate has much value at all.
As pitchers give up more hard contact or higher exit velocities, the variance in their starts does not increase or decrease. A pitcher with an average EV of 87 is just as likely to have sporadic outings as one with an average EV of 92.
Instead, barrel rate is better at explaining earned run variance.
We can measure the correlation between these two statistics with Pearson’s Correlation Coefficient. The closer a value is to 1 (or -1), the more strongly correlated they are. The correlation between barrel rate and earned run variance is 0.1479. Nothing crazy, but it suggests some sort of relationship between the two.
Since barrels often result in home runs, we can almost say pitchers who give up more home runs than their peers will also be less consistent in their outings.
Ground ball rate and chase rate are even better indicators of earned run variance.
The correlation between ground ball rate and earned run variance is -0.1794; stronger than barrel rate. Pitchers who create more grounders than their peers will see less volatility in their starts. This plays off what we know about barrel rate.
Chase rate has the strongest correlation to earned run variance with a number of -0.1937. Pitchers who are more adept at getting batters to chase have seen a tighter distribution in the number of earned runs they give up. It’s tough to say exactly why, but we can speculate that pitchers who generate the poorest quality of swings (chases) are in the least amount of danger of giving up extra-base hits. It is far more difficult for a batter that chases at a pitch to do any better than a whiff, weak groundout, or a foul.