clock menu more-arrow no yes mobile

Filed under:

A Look at Reliever Volatility and the Astros’ Bullpen

As the Astros try to construct their bullpen, an examination of high and low variance relievers sheds more light on the task

MLB: San Diego Padres at Houston Astros
Kendall Graveman, shown pitching against the Padres this year, will not be available in 2024 due to injury.
Thomas Shea-USA TODAY Sports

The Astros will have vacancies in four of the 2023 bullpen slots. Hector Neris, Phil Maton, and Ryne Stanek, who are free agents in 2024, pitched 184 innings out of the bullpen in 2023. The fourth vacancy arises with the news that Kendall Graveman will not be available in 2024 due to surgery. The Astros now face the task of replacing those pitchers either externally (trade or free agency) or internally (promotion of minor league pitchers).

Relief pitchers are notoriously volatile, meaning their performance can swing back and forth between seasons—which increases the difficulty of constructing a bullpen. This is partly due to the small sample size of their annual innings pitched. But it’s not like relief pitchers are all equally volatile. Some will exhibit a higher probability of volatile outcomes than others. As we try to project reliever pitchers’ runs allowed, some relievers’ will be more high variance then others.

While we normally focus on a projection system’s expected value for the most likely ERA (50 percentile), the ZIPS model also produces ERA projections for higher and lower percentiles or probabilities. The Fangraphs article on the ZIPS 2024 Astros projections shows selected results at the 80 percentile (low ERA) and 20 percentile (high ERA). Since these two ends of the spectrum probably encompass the reasonable high and low ends of the projection, I will consider the difference between these two high and low ERA projections as measures of possible variance in the players’ future performance.

An initial question for the Astros is whether to re-sign one or more of their free agents. ZIPS’ expected ERA for the three free agents is as follows: Neris, 3.72; Maton, 3.86; and Stanek, 4.13. However, examining their high and low ERA projections (80% and 20%) provides us more information on the possible variance in outcomes. It’s not really a question as to whether high or low variance pitchers are good or bad. A high variance pitcher may have significantly higher upside, significantly worse downside, or both characteristics. That could be good, if the team is enticed by upside. A low variance pitcher may not produced the best expected ERA, but the team may want the greater certainty of less downside. The table below shows the variance between high and low ERA, according to ZIPS, for various Astros’ relievers, including the three free agents.

High / Low ERA Projections
Fangraphs and ZIPS

Among the three free agents, Neris is projected with the highest variance (2.2 R/9), followed by Stanek (2.09 R/9 variance). Maton’s projection, on the other hand, is relatively low variance. Maton doesn’t offer the potential upside of Neris’ 2.83 ERA if things break right (80 percentile); however, conversely, he has a lower downside ERA (4.84) than Neris (5.03 ERA). Based on the ZIPS projection, though, Stanek’s variance is unattractive, since it is mostly due to a 5.53 ERA on the high side of his range. It’s worth noting that Ronel Blanco is the highest variance reliever, with an ERA range that varies between 2.78 and 5.10.

Since Urquidy pitched mostly out of the bullpen last year, I included him on this list—and he has the lowest projected variance. With an ERA range of 3.59 to 4.67, Urquidy could turn out to be very useful in soaking up relief pitcher innings next season. Also, the Astros have to hope that Martinez and Montero can achieve their low ERA projection—because the high end of the projection for them is not attractive.

What are the drivers of a high variance pitcher? Batted ball outcomes are a likely factor, due to their more significant random variation. BABIP and HR/Fly are important indicators of the random variation. Neris’ .219 BABIP (the lowest on the team) is a red flag for probable batted ball regression in 2024. ZIPS projects a 26% increase in his BABIP for 2024. Similarly, examining x-FIP (which normalizes HR/Fly rate) indicates potential home run regression for Neris. The difference between x-FIP and ERA for Neris is 2.74 R/9—by far the highest differential on the team in 2023, strongly suggesting probable regression.

In contrast, Martinez and Montero are projected to experience a 14% - 15% decline in BABIP in 2023. Both relievers’ ERA in 2023 was higher than their respective x-FIP, which also indicates possible improvement on the home run front. Both pitchers may be more useful in 2024 than 2023—-but it’s not clear how much they could compensate for the loss of Neris.

We don’t know how much the three free agents would cost to re-sign. Reportedly, the Yankees and Rangers are interested in signing Neris—which may indicate that he will be more costly than the Astros are wiling to pay. And the potential for regression by Neris should make the Astros hesitant to pay a high cost for his services. If Maton’s contract cost is reasonable, he could be a reasonable signing for middle inning work.

Looking for Help in the Minor Leagues

This type of analysis aimed at projected variance may offer even more utility for examining the minor league relief candidates. They have little or no experience at the major league level, resulting in ERA expected value projections which are less reliable. But the review of 80 and 20 percentile ERA may help identify those with more upside, as well as candidates likely to avoid the worst downside.

A table of variance for several of these AAA pitchers is shown below.

ZIPS High and Low ERA Projections

Mushinski, Sousa, and Gage all have some limited experience at the ML level and presumably will compete for the LH reliever role. Mushinski’s projections are attractive. Although his low ERA projection (3.49) is very similar to the low projections for Gage and Sousa, he offers a much lower ERA variance. Mushinski’s variance is less than 1 run, meaning that he avoids the 5+ high end ERAs attached to Sousa and Gage. At their best, Sousa and Gage are projected to be useful, but a lot of uncertainty surrounds their projection.

ZIPS isn’t very hopeful for Whitley—and, given his injury riddled history, that is understandable. If Whitley could achieve the low end of his ERA range, he might have some value in relief. But the potential, on the high side, for an ERA approaching six should lead to some skepticism.

Gordon, Kouba, and Arrighetti all show a reasonable variance in their projections. Gordon’s projection is a little more attractive because the low side of the ERA projection is better than the comparable projection for the other two pitchers. Arrighetti is projected to have a worse downside (5.07 ERA). Just looking at the ZIPS projections, Gordan and Kouba appear to have the best chance to make the pitching staff.

This article doesn’t address other free agent relief pitchers who may be available. However, my Nov. 28 article, “Looking for Free Agent Relievers,” provides some information on the options in the market.