When does something stop being a curiosity and start becoming a trend?

Now that All-Star reliever Clay Holmes has agreed to a deal with the Mets, reportedly to be a starter ¨C a role he filled just four times in his career, all back in his rookie season of 2018 ¨C we might just be on the verge of finding out. After decades of seeing pitchers who couldn¡¯t hack it in the rotation find new life in the bullpen, suddenly it seems like we¡¯ve seen a number of pitchers going the other direction in the last few years, like Seth Lugo, who finished second in the 2024 AL Cy Young race.

Of course, it doesn¡¯t always work, like when Miami tried to put A.J. Puk into the rotation early in 2024 and removed him after four disastrous starts, or when the Tigers did the same to Phil Coke in June of 2011. Sometimes, it never even gets out of the starting gate, as when the Reds tried to make Aroldis Chapman a starter during the springs of 2012 and ¡®13 and he ended up in the bullpen each time ¨C all these years later, he¡¯s yet to make his first Major League start. He probably never will.

We tend to remember the success more than the failures, and the aborted attempts are all but impossible to track down reliably. But we know it¡¯s happened a few times recently, and with rumors surrounding Holmes, Jeff Hoffman and Griffin Jax, we might end up seeing more this year. Can we look back over the last few years and see if we can learn anything from the ones that pulled it off? You wouldn¡¯t ask Kenley Jansen to do this, for example. What makes Holmes a (potentially) better candidate?

Let¡¯s start by figuring out if this is even happening more often right now, or if it just feels that way. The answer? Actually, it seems like this has been happening a few times a year, every year, for decades.

Since 1973, we found 104 reliever-to-starter conversions. So this isn't a new phenomenon. But over these past five decades, it's also never been a particularly common occurrence.

This, of course, comes with a lot of caveats. To begin with: How do you define a reliever-to-starter conversion? It¡¯s surprisingly tricky. There¡¯s not a rulebook definition. Your own version might vary.

For example: back in the early 1990s, a young Pedro Martinez started only three of the 67 games he appeared in for the Dodgers before being traded to the Expos, who then immediately inserted him into the rotation. But Martinez had been a starter in the Minors, and teams breaking in a young arm in relief before moving him to the rotation is hardly unheard of. A few years before, this is how David Cone had broken in. A few decades before, it was Sandy Koufax. Elite starters don¡¯t always arrive fully-formed on day one.

None of this is unusual, even now. It¡¯s what the 2021 Cubs did with Justin Steele, moving him into the rotation just 11 relief appearances into his first season. It¡¯s what the 2023 Blue Jays did with 2024 breakout starter Bowden Francis, working him into the bigs. It¡¯s what the Astros did with Ronel Blanco, giving him 16 relief games before moving him up. It¡¯s common, and not really what we¡¯re looking for ¨C what we want are established relievers moving to the rotation.

Like, say, Clay Holmes.

So, let¡¯s try to set some guidelines. Since 1969, there have been more than 4,600 pitchers with a season of 20 or more innings. Many of those arms were almost-always starters or almost-always relievers. Not all of them, though, and that¡¯s what we¡¯re after. Yet we don¡¯t want starters who went to the bullpen, like Luke Weaver in 2024, and we don¡¯t want obvious starters who had brief trips to the bullpen before returning, like Yusei Kikuchi in 2022 or Cole Ragans in 2023.

We landed on a simple set of rules:

• At least two consecutive seasons (min. 20 innings pitched) of at least 90% relief usage;
• Followed by a season of at least 50% starting usage.

That¡¯s it. Simple as that. Holmes, Jax and Hoffman would all qualify, if they start half their games in 2025. Now: This isn¡¯t the full picture, as failed experiments like Puk and Chapman won¡¯t get included, and rarities like Garrett Crochet making his first start after missing most of the previous seasons as an injured reliever fall outside the filter too.

But within this group, we get a lot of signal. We have some notable conversions, from quite recent (Reynaldo L¨®pez and Michael King in 2024; Lugo in 2023; Jeffrey Springs in 2022), memorable (Chris Sale in 2012; C.J. Wilson in 2010; John Smoltz in 2005; Derek Lowe in 2002), disastrous (Daniel Bard in 2012; Rick Aguilera in 1996), surprising (Dave Righetti in 1995; Goose Gossage in 1976) and absolutely hilarious (Doug Bird did it twice; both Greg Harrises did it within a year of each other, in 1990 and ¡®91).

So it happens, and it¡¯s always happened, and it might still happen. What have we learned about the recent conversions ¨C and Puk¡¯s failure ¨C and why it might apply to Holmes? One study from back in 2009, by Tom Tango (now at MLB), suggested that pitchers were a full run per nine innings worse as starters than relievers. That probably still holds true, but the game has changed since then. Let¡¯s set aside the long-ago legends of Ted Power (1986), David Wells (1990) and Brian Tallet (2009), and focus on the more recent success stories.

What can we learn from this group?

It probably helps to have been a starter in the past.

It¡¯d be one thing if these were all long-time relievers who suddenly showed up in the rotation one day ¨C like, say, if Craig Kimbrel or Jansen did it. But a lot of our recent conversions had previously had starting experience. L¨®pez, for example, made 96 starts in his first six years in the Majors, and Lugo was the first man on the rubber 26 times in his first two years with the Mets. Michael Lorenzen, as a rookie, started 21 of 27 games with the 2015 Reds. King made 10 starts for the Yankees in 2020-21 before moving into the bullpen. Part of the reason Jordan Hicks didn¡¯t officially qualify for our list was because of the eight starts he¡¯d already made for St. Louis in 2022.

This would apply to Hoffman, who started 38 of his 52 games in his first four seasons in the bigs, then 11 more with the Reds in 2021. It would apply to Jax, who started 14 games as a Twins rookie in 2021. It would apply to Holmes too, who started in the Minors before not really getting a fair shake with the Pirates, as he made only four starts ¨C one of which, six innings of scoreless ball against Milwaukee, is primarily interesting now because of who relieved him: Tyler Glasnow.

Many of them acted like starters in the bullpen anyway.

¡°For me with how healthy I am and what I've done the last few years with my arsenal, it's an interesting thought,¡± Hoffman told ESPN¡¯s Kiley McDaniel this offseason. ¡°It makes sense that guys with deeper arsenals than most relievers have found success."

Lugo throws nine pitches ¨C and he¡¯d desired a move back from the bullpen for years ¨C and Lorenzen throws seven, so that tracks. Jax already throws five as a reliever, King had four, and Hicks upped his splitter usage upon returning to the rotation. It¡¯s hardly as simple as ¡°more pitches makes you better,¡± but if you only have two pitches, then either one or both are elite (like Spencer Strider, or Cristian Javier's fastball) or you¡¯re probably in the bullpen. This is part of why Puk tried to learn three new pitches last spring, which didn¡¯t work, and why Crochet¡¯s new cutter was a big part of his unexpected 2024 success.

¡°I also think as teams have gotten better at optimizing pitch shapes,¡± an anonymous executive told MLB.com¡¯s Mark Feinsand in May, ¡°more teams will try and tweak or add a third pitch that previously prevented a reliever from starting.¡±

Holmes¡¯ repertoire is very much east/west ¨C that is, sinkers going one way, sweepers the other, and a slider in the middle ¨C and he¡¯d probably need to add at least one kind of vertical pitch. It¡¯s worth noting he brought back an old four-seamer in the playoffs, which showed promise.

You probably want to make sure you can pitch to both batter sides.

When The Athletic¡¯s Eno Sarris looked into this recently, he found that ¡°almost all of the conversions could command at least one pitch well, and most could command two or more well,¡± which is also part of what doomed Puk. Holmes¡¯ slider rates as excellent in FanGraphs¡¯s Location+ metric; his sinker is about average.

More importantly, if a pitcher is going to turn a lineup over multiple times, you can¡¯t really do that with a massive platoon split. Over the last two years, Holmes has a .615 OPS against lefties, and a .631 OPS against righties. To be clear, those would not simply stay the same as a starter, and he probably wouldn't be able to throw the sinker two-thirds of the time to lefties as he did in 2024. It is, at least, not starting from a problem spot against one side or the other, which would be enough to prevent many relievers from even trying this.

What does it even mean to be a starter?

This is a key question. If Holmes is going back to the rotation, it probably just means he¡¯s throwing four or five innings, not seven or eight. He's not going to be Nolan Ryan. It's just about getting a few more outs from him each week, compounding over a whole season.

¡°Especially with the expectation that starters may not have to go three times through the order, this may become more common,¡± the anonymous executive told Feinsand.

There¡¯s the risk, too, of what happens to your pitch shapes and effectiveness if and when you don¡¯t throw as hard as a starter. For Hicks, who was touching 105 mph as a reliever, he had mph to spare. For L¨®pez, that meant dropping from 98 as a reliever to 96 as a starter, which is still excellent. It might not work as well for someone like Michael Soroka, who was dropped from the White Sox rotation last summer only to go to the bullpen, add velocity, and start dominating ¨C yet is reportedly being sought after as a starter again.

It might not work for Holmes, either. It might be that come Spring Training, the Mets' pitching staff requires him more in the ¡®pen than the rotation. It might be that he tries it in spring, and the results just aren¡¯t there. It¡¯s not for everyone. It might not even be a trend. But it is another data point that increasingly, there¡¯s not really starters or relievers. There¡¯s just pitchers, and when you happen to see them.