Early Insights From Statcast’s Outfield Catch Probability Metrics

Hunter Renfroe Brett Phillips
Darren Yamashita-USA TODAY Sports

Amazingly, in an Opening Day game where Shohei Ohtani struck out double-digit batters in six shutout innings, the most memorable highlight of the night didn’t come from him, or even Mike Trout. In the bottom of the fifth, Oakland third baseman Jace Peterson sent a fly ball to right field. Hunter Renfroe gave chase, but it appeared to be going over his head — until he leapt up, stuck out his glove to the left while facing right, and somehow made an incredible no-look catch to the delight of Ohtani and the Angels. Even Peterson had to smile.

Baseball Savant has recently released outfield catch probabilities for individual plays, and we can learn a lot from analyzing the differences between the perceived difficulty of a play from watching it on a broadcast compared to its actual catch probability. Renfroe’s circus catch in Oakland offers a perfect example: While his acrobatics were necessary to make the catch, that was only because of a poor jump. He backpedaled for the first few steps, then ran at less-than-full speed while having to crane his head around to keep track of the ball. Renfroe ended up making the catch 39 feet from his initial position in an opportunity time of 4.2 seconds — a play that has a catch probability of 99%, and that’s even when accounting for the difficulty of running backwards (which is included in calculating the odds).

For comparison, here’s a play with a near-identical distance and opportunity time made by Renfroe’s backup, Brett Phillips.

Phillips didn’t need luck or heroics to make the out here; in fact, he was able to camp out for a bit before the ball fell into his glove. A good chunk of his route was completed before the broadcast had switched to the outfield camera.

In other words, Renfroe’s play is made with little fanfare almost every time. That includes him: He was perfect on fly balls with 99% catch probability in 2022, though he did let a few in the 90–95% range drop for hits. Much of the focus that observers put on the quality of a outfielder’s defense naturally comes from what can be seen on TV – but the data indicates that what we can’t see is what truly separates the great fielders from the poor ones.

Statcast’s jump metric breaks down the first three seconds after a ball is struck into three parts: the reaction, burst, and route. We’ll work backwards from the time the ball is caught, starting with route efficiency, which measures how far a player runs compared to the straight-line distance between his initial position and the ball’s landing spot. A bad route can often be seen on a broadcast camera: players struggling to change direction while trying to make up lost ground. In this example, Jarren Duran (who finished second-to-last in route efficiency last year) fails to make a play with an 80% catch probability, and it’s clear that he thought the ball was going deeper than it actually did.

But many players can make up for poor route efficiency with elite reactions and speed. Trent Grisham, tied with Duran in route futility, had 14 OAA in 2022 thanks to his top-tier instincts and sprint speed. In fact, there appears to be a mildly negative correlation between route efficiency and overall distance covered for outfielders. That isn’t to say that taking good routes makes you worse at hunting down fly balls. Having a good feel for the end location of a fly ball is a byproduct of experience, and with experience comes age that can slow a player down. For example, 2021’s best route runner was Andrew McCutchen, whose 15,000 innings of experience in the grass certainly aided him. But because of poor reactions and slower legs, he was worth -7 OAA that season. And only one of last season’s top 14 players by route distance above average finished with a positive OAA. So while routes might be the easiest aspect of outfield range to evaluate outside of dropped catches and errors, route efficiency tells us surprisingly little about a fielder’s overall skill.

While many players are adept at taking direct paths to the ball, top speed is often the biggest differentiator on difficult plays. The burst component of Statcast’s jump tracking tests a player’s quickness closing in on a ball, measuring feet covered 1.5–3 seconds after a ball has been struck. This component most directly correlates with sprint speed; seeing Phillips, Jose Siri, and Michael Harris II near the top of the leaderboard confirms this. On the other end of the spectrum is Andrew Vaughn, who covered three feet less than the average outfielder. He had 25th-percentile sprint speed on the basepaths but often didn’t reach his top speed in the outfield; in 27 play opportunities with a catch probability below 75% last season, he converted just one of those balls into outs.

But while watching video, I found a play that I thought looked rather impressive. On a ball slicing toward the corner, Vaughn got a quick beat on the ball, tracking it into his glove on the run. Try to guess the catch probability as you watch:

I polled a few people and got estimates in the 50–80% range, which is reasonable for a seemingly difficult play that required a good jump and route but not necessarily elite athleticism (like this play from Luis Robert Jr.). So I was surprised to find the actual probability was 99% — a routine play. Vaughn’s clean route and running catch masked the fact that he never reached full speed. And while a quick burst isn’t strictly necessary for a routine play, he was significantly slower than other outfielders when he had to run a longer distance. His routes were actually a bit above average last year, but his subpar speed puts a hard ceiling on the radius of his range.

But while the lightning-quick speed of Robert Jr. and sluggishness of Vaughn is easily discernible to viewers, more subtle differences in burst speed can have substantial effects on OAA and be harder to compare. For example, a one-foot gap in ground covered in the burst stage was the difference between Max Kepler’s 8 OAA and Aaron Hicks‘ -1, despite exactly equal marks in the other jump components and the same number of fielding opportunities.

Finally, there’s the reaction phase — the initial read that a fielder gets on a batted ball. Specifically, Statcast measures the distance covered in the correct direction in the first 1.5 seconds after contact. Unfortunately, broadcast cameras usually don’t switch to the outfield until after this window has elapsed, making it often impossible to evaluate reaction time on the screen. But jumps are extremely important when it comes to making catches; even a player with top-tier sprint speed and a perfect route will find it difficult to make up for a poor read, like Bryan Reynolds on this play. Again, make a prediction on the catch probability here:

Reynolds comes in charging hard, getting near his top speed (he’s a 75th-percentile runner) and taking a clean path to the ball. But his sliding catch fails, and the ball drops right in front of him. It seems like there’s not much more he could have done, right? But his catch probability on this play was 90%. Consider what he’s doing right as the broadcast switches to the outfield camera: He’s coming forward but hasn’t turned on the jets and is looking up as if he hasn’t found the ball yet. His deep positioning as the camera switches also suggests that he took a step back at impact rather than immediately crashing in.

Contrast Reynolds’ attempt with this catch by Enrique Hernández on a ball with near-identical opportunity time, distance covered, and direction:

In the world of reaction time on fly balls, Hernández is a god amongst men: His 4.3 feet above average in the reaction stage was a full foot ahead of anyone else, and in 2021, his 4.7 feet cleared the field by a full two feet. That’s why despite being in the bottom ten in route efficiency among 104 qualified defenders, he still covered more ground in the first three seconds of a play than any of his competitors. In that clip above, he slows down mid-route from full speed, knowing that he’ll get to the ball easily. Reynolds, on the other hand, had far more ground to make up after the first second or two of the play, and that was a constant issue for him: Last year, he finished tied for fifth-worst in reaction, losing 1.4 feet relative to average.

Let’s wrap up by computing the importance of each individual jump component on overall defensive efficacy as measured by OAA. As expected, reaction and burst have very strong correlations, though it’s a little surprising to see that the latter has a slight edge. On the jump leaderboard, the spread of measurements is a bit wider for reaction; Hernández’ 4.3 feet above average beats out Kyle Isbel’s league-leading 2.9 feet above average on bursts. But it makes sense that great acceleration in the burst phase lets players reach a higher top speed, which is often a difference maker on long runs. Route efficiency does end up with a slight negative correlation, both due to older players being better route runners and because players with elite reactions might take suboptimal routes but still come out ahead in the end.

Correlation Between Jump Components and Outs Above Average
Component Correlation to OAA
Reaction .740
Burst .916
Route -.319
SOURCE: Baseball Savant

Statcast’s catch probability data gives us insights on which fielders give themselves a huge advantage before the TV cameras are on them. Catches like Renfroe’s are immediately recognizable as excellent, but some of the best fielders in the game can cover absurd distances without the need to slide or stick the glove out without looking. These metrics teach us a lot about what we can and can’t evaluate from traditional broadcast angles, and allow us to put shine on players who make their incredibly difficult jobs look like just another day in the office.

Kyle is a FanGraphs contributor who likes to write about unique players who aren't superstars. He likes multipositional catchers, dislikes fastballs, and wants to see the return of the 100-inning reliever. He's currently a college student studying math education, and wants to apply that experience to his writing by making sabermetrics more accessible to learn about. Previously, he's written for PitcherList using pitch data to bring analytical insight to pitcher GIFs and on his personal blog about the Angels.

Newest Most Voted
Inline Feedbacks
View all comments
1 year ago

Where can we find this on Savant?

1 year ago
Reply to  slogger

I mean where to search by individual plays.

1 year ago
Reply to  lavarnway

Yeah curious on the individual play data too

1 year ago
Reply to  Shauncore

Any help @KyleKishimoto?

1 year ago
Reply to  Kyle Kishimoto

Thank you! I didn’t realize that was a feature. Cool stuff.