Last year, Brewers’ baseball stumbled to a 14-24 record, their worst since 2012 (Vassar College Athletics, “Baseball Cumulative Statistics,” 02.09.2020). Their ERA of 6.68 was also the highest in eight years. Was this just bad luck, or are the Brewers in need of a serious pick-me-up with less than two weeks remaining until the start of the 2020 season?
An oft-used indicator of luck in baseball is a team’s run differential: the difference between the amount of runs they scored and the amount of runs allowed. Run differential is often more indicative of a team’s overall skill than their win-loss record ]. You can get lucky and win a lot of close games, get blown out in your losses, and emerge with a better record than run differential. The Brewers’ run differential last year was -65, their worst since 2012. No bad luck there.
To pinpoint the source of the Brewers’ poor play, I used a statistical technique called linear regression. This type of analysis examines a linear relationship between one or more explanatory (independent) variables and a response (dependent) variable.
Just how important is ERA, though? Since it proved to be the only significant predictor in the multiple linear regression model, I ran a simple linear regression modelling winning percentage based on ERA alone, and it produced an R-squared value of 0.42. This value means that 42-percent of the variance in the Brewer’s winning percentage is due to the variance in their ERA. That is a very high number.
In a project I worked on last year, I attempted to predict ERA from skill-based pitching statistics. There are a lot of things a pitcher has no control over: depending on who’s fielding behind you, a double can become an out, and the ballpark you’re pitching in can determine whether a 300-ft fly is a home run or a lazy pop-up. I defined skillbased statistics as those which did not vary significantly from a pitcher’s performance one year to their performance the next. This lack of variance seemed to me a good indicator of a statistic not reliant on luck. The statistic that proved most significant in this analysis was the difference between a pitcher’s strikeout rate and their walk rate. I gathered the data to approximate this statistic, using strikeouts per nine innings minus walks per nine innings. While this statistic seemed insignificant in the smaller dataset, my larger analysis last year showed it was undoubtedly significant, as did other analyses done by sports writers more professional than myself (The Hardball Times, “Should we be using ERA estimators during the season?”, 02.09.2020).
With this in mind, let’s take one more look at the strikeout and walk rates for Brewers’ pitchers. The Brewers last year struck out more batters per nine innings than any other season in the dataset. However, they also walked their most batters per nine innings. As a result of the high strikeout rate, their strikeout minus walk rate last year was not their worst, and since this statistic is the most significant predictor of ERA, it means their sky-high ERA last year should not have been their highest. However, if the Brewers want to focus on something that will assuredly lower it, it is cutting the walks while maintaining their gains in strikeouts.
Stat of the Week: 42 percent