Let’s Find a Multi-Inning Reliever

The height of fashion in baseball analysis three years ago was finding a reliever who could pitch multiple innings. Some people called it the Andrew Miller role, though Miller was never a perfect example of it — aside from the memorable 2016 playoffs, Miller was more of a setup man who occasionally threw the seventh in his tenure on the Indians. Chris Devenski and Chad Green were trendy examples in 2017, and Mets swingmen Seth Lugo and Robert Gsellman both performed admirably in long relief in 2018.

Whichever example you turn to, the value of having a reliever who can perform over multiple innings of work is clear to see. As starters throw fewer innings across baseball, having relievers who can handle larger workloads is increasingly important. A two-inning reliever might have been a luxury in 2009, when a seven-man bullpen would cover two or three innings a night, but 2019 bullpens go eight deep and pick up nearly four innings a game. Using relievers to cover more innings naturally results in weaker relievers getting into games, so getting extra frames out of good relievers has never been more valuable.

That’s the theory, anyway. In practice, the role isn’t easy to fill. If you’re looking for someone to throw a few innings of relief, they have to be a decent pitcher. There’s not really much point in filling bulk innings with replacement-level stuff — you could just use the back of the bullpen for that. There’s just one problem with that: a good pitcher who can throw multiple innings mostly describes a starter, and getting rid of a good starter to create a good reliever doesn’t make that much sense. Blake Snell, for example, would probably make a great reliever, but that would be a waste of his talent.

Still, not every starter is ideally cast. Take the pitchers I listed above: Andrew Miller was a failed starter. Devenski, Lugo, and Gsellman had various degrees of success as starters before their teams made them full-time bullpen members. If we take care, we can search for pitchers who struggle as starters but might excel as relievers. Will all of these be perfect fits? Perhaps not. Most likely not, even. Still, finding a few players who would be better as relievers could be quite valuable to a few teams, and if nothing else, it’s a fun intellectual exercise. Let’s try!

What’s one reason a pitcher might not be well-suited to starting? The first place my mind goes is fatigue. When you watch a great starter pitch, it often seems like they’re getting stronger as the game goes on, maintaining or even increasing velocity and mixing in new pitches. How many times have you seen Justin Verlander or Jacob deGrom spike a 98 on the radar gun in an important spot in the seventh inning? That’s a valuable skill, and one we wouldn’t want to waste on a reliever.

To look for this effect, you might think we should look at each pitcher’s effectiveness by times through the order — but you’d be wrong. That kind of split takes forever to stabilize, such that we almost certainly don’t have enough data in time for that data to matter. Instead, I want to look at pitchers whose skill level measurably changes as they throw more pitches. To test for this, I looked at every four-seam fastball thrown in the first 20 pitches of an appearance to find each pitcher’s early velocity. Then I looked at every four-seamer thrown between the 51st and 70th pitches of an appearance to look for “late velocity.” I then repeated the process with two-seamers (doing all fastballs together might yield weird results for pitchers who throw both). Take a look at the pitchers who have thrown at least 15 pitches both early and late, and who lose at least one mph off their fastballs as the game goes on.

Fastball Velocity Decliners
Pitcher Early Velo Late Velo Difference
Jon Lester 90.9 89.9 -1
Jonathan Loaisiga 96.8 95.8 -1
Brad Keller 93.1 92 -1.1
Domingo German 94.6 93.5 -1.1
Gregory Soto 94.9 93.7 -1.2
Nick Pivetta 94.8 93.6 -1.2
Tyson Ross 91 89.7 -1.3
Nathan Eovaldi 98.5 97.2 -1.3
Luis Castillo 97 95.7 -1.3
Sean Newcomb 93.5 92.1 -1.4
Pablo Lopez 94.1 92.6 -1.5
Aaron Nola 93.7 92.2 -1.5

This is a good place to start — these pitchers are bringing measurably worse stuff to the mound as the game wears on. There’s wide dispersion on this list; it’s hard to seriously argue that Jon Lester is better suited to relieving than starting, and given Luis Castillo’s phenomenal start, Reds fans would mutiny if he transitioned to the bullpen. Still, there might be something here. The Braves already made Sean Newcomb a multi-inning reliever this year. Seeing a team’s revealed preferences line up with some of our criteria is a good sign that we’re onto something.

The next cut, I’ll admit, is a bit less scientific than the first; it’s more down to personal preferences than an absolute rule. Still, I think it generally speaks to team preferences, so we’re going to run with it. I don’t want to make a potential four-pitch monster into a reliever — if a pitcher has a deep bench of usable pitches, I’m willing to believe he stands a better than average chance of being effective deep into games. Let’s cut out pitchers who throw their top two pitches less than 80% of the time — an arbitrary cutoff, to be sure, but we need some cutoff, and I think this one does a good job of weeding out starters with three good pitches. Here’s that list again, this time with our new criteria added.

Decliners With Few Pitches
Pitcher Early Velo Late Velo Difference Top 2 %
Jonathan Loaisiga 96.8 95.8 -1 84
Brad Keller 93.1 92 -1.1 97.5
Gregory Soto 94.9 93.7 -1.2 95.6
Nick Pivetta 94.8 93.6 -1.2 81.1
Luis Castillo 97 95.7 -1.3 86.1
Sean Newcomb 93.5 92.1 -1.4 81.2
Aaron Nola 93.7 92.2 -1.5 81.6

Now we’re getting down to brass tacks — seven pitchers, all of whom are heavily reliant on two pitches and whose fastballs lose steam in later innings. Newcomb is already a reliever. Keller was a swingman last year, and the Royals are running him out in the starting rotation more out of a sense of adventure and discovery than because he’s best suited for that role. Loaisiga would probably be a reliever if the Yankees didn’t have an embarrassment of riches in their ‘pen. These tests seem to be on the right track.

With the list down to seven names, though, there are still two pitchers on here who won’t be transitioning to the bullpen anytime soon — Nola and Castillo. Rather than try to make any more generalized cuts, let’s look at the remaining cases on their respective merits to decide if each pitcher is better-suited to the multi-inning reliever role than their current starting job.

The first three are easy. I think there’s enough evidence to say that Newcomb, Keller, and Loaisiga, at least in their current incarnations, would ideally be high-leverage bullpen arms. All three boast excellent stuff with little variety, and none of them are so good as starters that they’re too valuable to move over. In fact, were it not for the unique circumstances Keller and Loaisiga find themselves in, all three of these conversions might have happened already.

On the other hand, Nola and Castillo are clearly not relief candidates. He might be struggling this season, Nola put up a 5-WAR effort last year, for goodness sake. Let’s handle his case first. His 1.5 mph velocity loss as the game goes on is scary, but it’s also a career-worst. Last year and in 2017, the differential was a smidge under 1 mph, worse than the major league median (~.3 mph) but more manageable. While Nola certainly has a reliever-y profile, with one excellent breaking pitch and a tremendous fastball, he also has a killer changeup.

His changeup, which he throws 18% of the time (thus just missing our cutoff), is a perfectly effective pitch that he simply doesn’t use all that often. I’m happy to exclude him as being too effective to move to relief, though if he continues to show big velocity drops later in starts, the Phillies might want to consider giving him some rest or having him feature his secondary pitches more later in games.

Castillo is a similar case to Nola. While he throws his fastball and changeup a ton, his third pitch, a wipeout slider with a 45% whiff rate, is tremendously effective. With his awe-inspiring changeup and upper-90s fastball, though, there’s little reason to use the slider more than he currently does. It’s also reasonable to think that the velocity loss affects him less than Nola — he throws so hard in the early innings that his diminished fastball is still faster than league average. If Castillo were less effective, he might be a good candidate to reinvent himself in the bullpen, but right now he’s too good for that. His current form shows the promise he’s always had — a hard-throwing three-pitch monster whose (excellent) fastball might be his worst pitch. No bullpen for you, Luis!

Gregory Soto is on our list, so he’s worth a mention, but only barely. The Tigers have given him three starts this year, and he’s compiled a 10.8 ERA to go with poor peripherals (6.08 FIP, 2% K-BB). His stuff screams reliever — his fastball plays early in outings, up near 95 mph, but it tails off later, and his only secondary offering is a slurve that he probably shouldn’t throw too much to righties. The Tigers aren’t contending this year, but trying Soto out in relief might be worth it, because trying him as a starter sure doesn’t seem to be working.

The most interesting case, the player I’d most like to see switched to relief, is Nick Pivetta. Pivetta has two years of semi-effective starting under his belt, so he seems like an odd candidate to ship off to bullpen purgatory, but his poor performance in 2019 has him cooling his heels in the minors, and the Phillies bullpen could use another arm. His stuff is exactly the kind that plays well in relief — his fastball sits at 95 mph early in starts, and his primary breaking pitch, a slow, giant-break curveball, provides a unique look. In fact, no curveball in baseball has both more horizontal and vertical break than Pivetta’s hook. If that doesn’t sound like the kind of pitch that a reliever could use, I don’t know what to tell you.

Seth Lugo is an excellent point of comparison for Pivetta. They both have spin-monster curveballs (Lugo’s has the highest spin rate of any curve in the majors this year, and Pivetta’s isn’t far behind). They both have sliders that are best used in moderation. They both have fastballs that could badly use an extra tick or two of velocity, and they both lose velocity quickly as starts go on. Lugo was a fine fifth starter for the Mets, but he became indispensable to the team when he switched to the bullpen, where he averaged 1.2 innings per appearance and put up a 2-handle ERA. I see no reason why Pivetta couldn’t pull off a similar trick.

For the most part, looking for starters who might be better suited as relievers tells us what we already know. Sean Newcomb should relieve? Thanks, genius, the Braves already told us that. Jonathan Loaisiga is fringy as a starter? Yep, got that one. Sometimes, the search for relief-profile starters is useless — sure, Aaron Nola and Luis Castillo have a relief-looking profile, but their stuff is so good that it plays in a starting role, so no one cares if they’d be good relievers. Sometimes, though, you might find a diamond in the rough. Is your team looking for relief help? Do you wish the sixth and seventh innings were interesting again? If Brad Keller and Nick Pivetta become relievers, they’re going to be incredibly interesting players.

Ben is a writer at FanGraphs. He can be found on Twitter @_Ben_Clemens.

Newest Most Voted
Inline Feedbacks
View all comments
4 years ago

I really love how you walked through this, even if the final story is underwhelming. It seems that teams usually do this to guys who have problems with walking batters, but if you look at walk rate, Keller and Newcomb wind up at the top of the list anyway. And then if you look at the other names on the list, it’s a who’s who of guys who got moved to the bullpen. Shelby Miller? Moved to the bullpen already. Derek Holland? Bullpen. Zack Godley? Bullpen.

That said, I’m not sure that’s a great way to figure out command, since in theory a guy can be wild all game (and therefore in short stints) or fine all game. I’m wondering if there’s a way to get at this using your methodology, maybe looking at command and how it varies over the course of a start. Are there guys who have trouble maintaining their command over 100 pitches? Maybe walks over the course of a game? Or something like that?