What Jump Can Tell About an Outfielder’s Defense

By Rameen Forghani

What defines a great catch by an outfielder? If you use SportsCenter’s Top 10 as a barometer, an outfielder is at the pinnacle of the position when he makes a diving catch. Flashy plays, like laying out headfirst or charging towards the infield to make a sliding snag, are easy enough to call great. Maybe you rely on the trained eyes of the announcer perched inside the broadcast booth, using his decades of experience to enthuse, “He got a great jump on that ball!” to appreciate unassumingly difficult plays.

Major League Baseball and Statcast have developed a metric to quantify this facet of an outfielder’s defense. Jump is a tool that reports the distance covered in the correct direction in the first three seconds of a ball’s flight, relative to league average. Jump is a sum of its three components: reaction, burst, and route. Reaction and burst roughly measure acceleration and speed, respectively, while route evaluates how direct of a line a fielder takes to the ball.

Jump is the next iteration of baseball’s desire to better understand how agility contributes to the game. From the 40-yard dash and the first-to-third time to the modern sprint speed, jump significantly increases analysts’ understanding of how speed affects outfield defense in live ball situations. In 2019, outfielders Ronald Acuña Jr. and Kevin Kiermaier both had the same maximum sprint speed (29.4 feet per second). Kiermaier has a jump of 3.8 feet, whereas Acuña Jr.’s is -0.4 feet. On any given fly ball, Kiermaier averages about a four-foot head start over Acuña Jr. As they have the same sprint speed, Kiermaier and Acuña Jr. will roughly move at the same speed towards the ball after the initial jump period; Kiermaier will remain four feet closer to the ball. For a flyball Acuña Jr. must dive to catch, Kiermaier makes the same play standing up.

Jump assigns a numerical value to most plays an outfielder makes. Intuitively, a higher jump (especially when normalized to league average) is better; any outfielder with a large, positive jump is faster than his MLB peers. Baseball games are decided by recording outs and saving runs, not by raw speed. Whether jump is correlated with recording outs, preventing runs from scoring, or winning ballgames is the more significant question at stake.

“We’re still trying to figure out what it [jump] means,” senior writer for FanGraphs Dan Szymborski said. “Does it tell us something in a predictive sense that the number of plays does not? You can have as much burst as you want, but that doesn’t mean you catch the ball.”

Jump is often compared to Outs Above Average (OAA). Outs Above Average is a range-based metric that quantifies how many outs an outfielder makes over his peers. OAA is a cumulative statistic based on catch probability: what is the likelihood that a fielder catches this ball given the amount of distance he must traverse and the time he has to do it? Catching a ball with a low catch probability adds a comparatively large number to OAA, whereas not catching a ball with a high catch probability subtracts an equally large amount from the running Outs Above Average tally.

Jump has a moderately strong positive correlation with Outs Above Average (Figure 2a). A fairly robust relationship is seen between jump and OAA — as jump increases, OAA usually also increases. A broad generalization is, therefore, that players with large jumps are typically better defenders who record more outs than players with smaller jumps.

The correlation is not perfect. Only 59.5% of the variation seen in Outs Above Average can be explained by variation in jump. Some players with poor jumps still have high OAA (Byron Buxton had a jump of -0.5 feet but 10 Outs Above Average in 2019). OAA is not the only metric available to examine defensive performance.

“Outs Above Average is fairly correlated with the range component of DRS [Defensive Runs Saved] and UZR [Ultimate Zone Rating],” Szymborski said. “Defensive statistics are still difficult and have a lot of things to take into consideration.”

OAA is fundamentally based on speed, and it does not take into account arm strength or accuracy, holding baserunners, home run robberies, or errors. Every out is not made equal; a home run robbery directly prevents at least one run from scoring, and arm strength can alter a baserunner’s decision from attempting to advance.

Despite its shortcomings, MLB.com reporter and Statcast researcher David Adler said, “I think Outs Above Average is the best evaluator of outfield defense that baseball has ever had.”

Reaction and burst are both also positively correlated with OAA, although the response to a faster reaction is more varied (Figure 2b & 2c). Figure 2d suggests that route has a negative association with OAA. That is, taking a more direct route to the ball results in recording less outs. The slope of the trendline is shallow (r = -0.1451), indicating a minor effect (if any) on overall defensive performance, and the coefficient of determination is similarly profoundly weak (R2 = 0.0211).

Although the sample size is too small and variation too large to conclude that any association exists between route and Outs Above Average, route does have a strong negative relationship with the amount of distance covered during the first 1.5 second reaction period (Figure 3c).

“It makes intuitive sense,” Szymborski said. “There’s a small risk when you break faster, you go in the wrong direction, which [you] hope is generally cancelled out when you run faster.”

The panels in figure 3 also demonstrate that the three components of jump are not colinear. Reaction, burst, and route are all independently valid measures of a fielder’s skill; a high value in any one component does not necessarily dictate the value of another component. The lack of multicollinearity affords increased statistical confidence in the regression analyses.

Reaction, burst, and route are able to be analyzed independently. In a multiple linear regression model, burst had a significantly larger coefficient than reaction or route (Figure 4). Changes in burst cause three-fold increases in OAA and overall defensive production compared to similar improvements in reaction and route. For every one foot increase in burst above league average, a player is expected to make an additional 4.16 outs above average. Jump must still be considered as increases in burst could be negated by decreases in reaction or route.

Reaction, burst, and route all contribute novel information to the composite jump metric. The highly significant p-values in the model depicted in figure 4 mean that removing any one of the components of jump from the model would result in different, less accurate predictions of defender performance.

Jump is a newer advanced statistic (publicly released data dates back to 2016). Players, coaches, and front offices are still finding ways to effectively synthesize data with personal evaluations. Human-evaluated top talent is often incongruous with statistical leaderboards.

The Gold Glove award provides one such example. Manager votes are weighted 75%, and a SABR-derived composite defensive statistic, roughly comparable to OAA, comprises the final 25% of the voting. Jump is again seen to be correlated with OAA among outfield gold glove winners in the past four seasons, as rows have similar colors (Figure 5). An outfielder’s jump rank (or OAA rank) is not strongly correlated with Gold Glove winners. Six of the 24 winners had a jump in the bottom half of all league outfielders.

The Gold Glove awards have other methodological issues hindering fair voting. Player popularity, reputation, incumbency, and minimum inning requirements all combine to create odd ballots, assuming coaches take voting seriously in the first place. Jump is not predictive of future success in any case.

Jump is a relatively stable metric. The overwhelming number of players either marginally improved or regressed from one season to the next (Figure 6a). Seven-of-ten players had an absolute change in jump of less than one foot in consecutive seasons. The distribution of changes in jump, plotted on a histogram (graph not shown), is unimodal, centered around 0, and has a tight spread with small tails.

Jump is not predictive of future change in OAA. Knowing a player’s jump in the current season has little predictive power for anticipating how many Outs Above Average he will record next season. The change in OAA from one season to the next is essentially uncorrelated to current jump (Figure 6b).

Small sample sizes diminish the predictive value of jump. Jump is only calculated on plays with a catch probability of 90% or less. Such plays make up a fraction of all outfield plays. According to data from Inside Edge, 266 balls were hit to Marcell Ozuna in the outfield in 2019. 195 were routine flyballs with a catch probability greater than 90% (73%), and 48 were balls nearly uncatchable with a catch probability less than 10% (18%). An outfielder’s ability boils down to 23 or so plays a season, a mere fraction of the total events in a baseball season.

Defensive statistics have problems. Batters are measured by outcomes that happen, whereas defenders are measured by outcomes they prevent from happening. Catch probabilities are just that: estimates that require some level of guess work and extrapolation. Jump is volatile due to its small sample size. Only additional seasons of data can help quell such variability. Coaches and players are already using jump to better identify strengths and weaknesses, personally tailoring offseason workouts and pre-game drills.