Rafael Devers, Inefficient Thief

Rafael Devers was an absolute stud last year. He amassed more than 700 plate appearances, the first full season of his career, and put up career highs in pretty much everything. Each of the three slash stats, ISO, wRC+, WAR, defensive value, baserunning runs — seriously, pretty much everything. But I’m not here to talk about that today; we get it, Rafael Devers is great. Instead, let’s talk about another career high: eight times caught stealing.

That sounds bad, right off the jump. Eight times? The rule of thumb with stolen bases is a 75% success rate; succeed any less often, and you’re costing your team value. Take a look at the caught stealing leaderboard, and you can see that most baserunners implicitly get this tradeoff:

Caught Stealing Leaders, 2019
Player Stolen Bases Caught Stealing Success Rate
Whit Merrifield 20 10 66.7%
Amed Rosario 19 10 65.5%
Ronald Acuña Jr. 37 9 80.4%
Jonathan Villar 40 9 81.6%
Victor Robles 28 9 75.7%
Mallex Smith 46 9 83.6%
Rougned Odor 11 9 55.0%
Rafael Devers 8 8 50.0%

Going 50% on your attempts clearly isn’t that. Take a look at this one, from a May 8 game against the Orioles:

That’s at least a good spot to be stealing — one where the baserunner can break even with a success rate below 75%. To be precise, it’s around 65% in this case, depending on what you think of Núñez’s expected production at the plate after an 0-2 count. But still, you have to succeed! Givens isn’t particularly quick to the plate, the ball’s in the dirt; if you’re going in this circumstance, things broke about as well as could be expected.

In fact, if you look at all of Devers’ eight times caught stealing, an interesting pattern emerges. In every single one, his breakeven success rate was below that naive 75%. He stole in the spots where the extra base mattered most. Sometimes that’s score-related; when the game is tied late, the first run is incredibly valuable. Sometimes the base/out state matters more: getting from second to third with one out is a good idea, as is going from first to second with two outs.

So was Devers simply a victim of finding spots that were really good and trying steals in those? Was he upping his own degree of difficulty by going for steals in spots where the other team knew he should try to steal? Yes and no.

Devers was caught stealing eight times last year. He also stole eight bases. And those eight bases he stole were, on average, easier bases to take than the ones he failed at. Take this situation:

Man on first, no one out, tied in the second — this is a time you want baserunners more than you want scoring position. The breakeven here is up around 73%; the upside is lower, and the downside relatively higher, than that spot against the Orioles. And he barely beat the throw! Was this actually a better process than that steal that failed?

In fact, Devers is just a convenient patsy for the point of this article: stolen base success rate is a really weird metric. There are times in a game where the defense almost doesn’t care about a stolen base. I’m not talking about defensive indifference, as that’s scored differently, but imagine a runner on first with nobody out, with his team trailing by five, in the bottom of the fourth.

It’s not completely irrelevant whether the runner is on first or second, but it’s pretty close. And making an out is terrible when you’re trailing by five; in fact, you would need to be safe nearly 80% of the time just to break even on this steal. It only makes sense to steal here if you’re very likely to be safe.

So don’t steal there, right? There’s just one problem: stealing isn’t a one-sided activity. The defense has something to say about it too. And if they know that it doesn’t make sense for a runner to steal, they’ll guard against it less. After all, there’s a real cost to holding a runner on. All you have to do is listen to a game of baseball, any game of baseball, and the announcers will let you know.

Thinking about the runner on first exacts a toll on the pitcher. Pinching the defense to account for a player in motion affects BABIP. When a catcher has to constantly think about popping out of his crouch to throw, he generally does a worse job framing. Wouldn’t it be nice to just forget about all of that and play? When a stolen base doesn’t cost the defending team much win probability, they can!

Imagine a player who only stole in situations where it made no sense. Stealing third with two outs, stealing when down 10 runs; just, every goofy spot you can think of. There would be some real benefits! Our hypothetical bizarre runner would never face a defense set up to stop him. He’d stand a good chance of racking up juicy steals totals — a better chance than someone who attempted to steal when it mattered.

One downside — they’d be empty steals. You don’t play baseball to accumulate stats; you play to win. And while in general context-neutral statistics work quite well, our bad-faith bandit’s steals would be less valuable than our unbiased estimate of the value of a steal.

More than any other offensive stat, steals lie. There’s no other spot where an offensive player can choose with such strong control whether or not to make an attempt. RBIs don’t really work as a statistic because they apply contextual value to something that isn’t contextual; the hitter doesn’t think “Oh, I’ll hit a double because there’s a runner on second,” he thinks “Oh, I’ll hit a double because doubles are good.” Crediting a bases-empty double and a bases-loaded double differently, as RBIs do, doesn’t make sense because the hitter doesn’t get to pick whether or not the bases are loaded.

But the opposite is actually true for steals. We mostly treat all steals the same — all we want are the raw number of steals and the success rate. But here, the runner does get to pick! We should be doing the exact opposite.

Players understand this implicitly. They may not know the exact math — the exact math is hard to tease out, in fact, and depends on a multitude of factors. I’m only going for broad approximations in this article. But watch a good base stealer operate, and they know what’s up. When it’s a relatively beneficial time to steal — when the reward is high relative to the risk — they’re likely to test any edge, push the boundaries of their primary lead, and risk a steal even if the situation isn’t perfect. If the situation doesn’t call for a steal, it’s never a close play — they steal the base on the pitcher, and are fiddling with their body armor and sliding glove before the throw even arrives at second.

That’s good base stealers. Rafael Devers, notably, isn’t one of those. He knows what he’s doing, conceptually — steal when it’s a good situation for him, hit the brakes when it isn’t. And yet, he’s a little too confident. Sometimes he’ll find a good situation — and be out by a mile:

Sometimes he’ll do almost everything right, and then leave his heel up for Elvis Andrus to make a highlight:

The question of when to steal, and how to determine who is and isn’t a good baserunner, isn’t a new sabermetric concept. Joe Sheehan addressed it back in 2004, and even before that Michael Wolverton talked about it from the catcher’s standpoint. Russell Carleton has done more recent work on what stolen base attempts tell us about baserunners — and his book The Shift covered it in even greater depth.

But that’s more detail than we really need to go into here. The fact of the matter is that Rafael Devers probably shouldn’t be attempting 16 steals in a season. Before 2019, his only taste of successful thievery had been in Hi-A in 2016, when he stole 18 bases and was caught six times. He’s slower now, and stealing is harder against major league catchers. It’s probably not worth the risk unless the situation makes for an easy swipe.

But part of the fun of baseball is that players don’t always make optimal decisions. Sometimes bad runners take off even when the defense isn’t giving them the base. Sometimes your heel sticks up and Elvis Andrus goes full Neo on you. That’s why we follow — not because there’s some formula to tell runners when they should steal, but because there isn’t one of those, and yet people try to figure out when they should do it anyway. It’s enough to make you feel like Pedro Severino:

We hoped you liked reading Rafael Devers, Inefficient Thief by Ben Clemens!

Please support FanGraphs by becoming a member. We publish thousands of articles a year, host multiple podcasts, and have an ever growing database of baseball stats.

FanGraphs does not have a paywall. With your membership, we can continue to offer the content you've come to rely on and add to our unique baseball coverage.

Support FanGraphs

Ben is a contributor to FanGraphs. A lifelong Cardinals fan, he got his start writing for Viva El Birdos. He can be found on Twitter @_Ben_Clemens.

newest oldest most voted

This is a really interesting article, Ben. Is there any way to know how much a fast baserunner on first (or any base) affects a pitcher? It seems like an unanswerable question, given all the other variables (pitching from the stretch, defenders holding him on).

What I’m wondering is, how much value does the baserunner add if he’s perceived as an actual threat to go? Is it possible to isolate those other variables and compare outcomes with a baserunner like Jose Ramirez on first vs. Carlos Santana?

Odor is a terrible basestealer. But if he added even a little positive outcome percentage to the batters by being a threat to go, that would bring down the required break even percentage for his steal to be worth it.


Conceivably one could create a + stat for each player that would attempt to normalize steals similar to what wrc+ does
for hitting based on various factors. You could also create a similar stat for pitchers or catchers, both would probably be necessary, that attempted to measure their performance at holding runners on and catching them stealing against a hypothetical league average base stealer. You could then compare a pitcher’s fip or xfip with the bases empty vs with base runners of a certain minimum quality on first. There would be several potential issues with doing this namely small sample sizes, and likely confounding effects due to the necessarily inter connected nature of the runners stolen base + stat, and the pitcher’s and catcher’s stolen base prevention + stat. Another potential is that there are some catchers, Yadier Molina comes to mind, where opposing teams attempt to steal so rarely that it may be impossible to get a sufficient sample size to normalize the catchers ability or that the stat might drastically underestimate the catcher’s stolen base prevention ability because opposing teams are extremely selective about when they steal against those catchers.