A new way to measure MLB's best infield defenders

Infield Outs Above Average now available at Baseball Savant

February 4th, 2020

Here's a list of baseball's best infield defenders in 2019:

+19 Javier Báez, Cubs
+17 Nolan Arenado, Rockies

+16 Andrelton Simmons, Angels
+16 Nick Ahmed, D-backs

+15 Trevor Story, Rockies
+14 Matt Chapman, A's

If you were to come up with a list of guys you'd think would be atop such a list, this is a pretty good start. Three of them (Arenado, Chapman, Ahmed) were Gold Glove winners this year. The others routinely find themselves in defensive highlight reels.

But what exactly is that list showing? It's the 2019 leaders for Outs Above Average, the Statcast defensive metric that's been limited to outfielders only for the last two seasons. As of today, it's been expanded to include infielders as well. You can find the 2019 leaderboard here, and there's a team version, with the Cardinals atop it at +42, here. That the individual list starts with "those defensive stars" and ends with "Vladimir Guerrero Jr., -16 OAA" gives us a nice amount of confidence that this is measuring something that we think it ought to be.

This metric starts with something you're familiar with -- the top-level number shown in the list above -- but there's also a lot of new information here you've never had a chance to investigate before. Let's dig in.

What is this, and how does it differ from other available metrics?

MLB.com's Tom Tango has delivered a very deep dive into all of the math and science that goes into this. You should go read it for a deeper understanding of how it works.

The short version is that there are four primary items that affect the chance of a play being converted into an out:

• How far the fielder has to go to reach the ball ("the intercept point")
• How much time he has to get there
• How far he then is from the base the runner is heading to
• On force plays, how fast the batter is, on average

The first two items form the basis (though not the entirety) of the outfield model, though the batter's speed isn't accounted for in outfield OAA. In the infield, it of course has to be; a ground ball with Billy Hamilton running is surely a different play than one with Albert Pujols running. The direction of the play from the fielder is accounted for, as well.

For example, this play below was considered to have a mere 10% chance of being converted into an out, since Carlos Correa had just 1.5 seconds to move 12 feet, then had to get the ball 147 feet to first base before Aaron Judge, who possesses above-average sprint speed (28.2 ft/sec), could get there. He did, so he adds +.90 towards his seasonal total of +9; if he'd missed it, it would have cost him just -.10.

Not every impressive infield play requires a throw, obviously. When Matt Chapman made this great dive to grab Michael Brantley's pop-up in April, he had 3.7 seconds to move 61 feet. This is also a play that had a 10% chance of being converted into an out.

That middle part -- "he had to move 61 feet" -- is what's important here. The ball wasn't foul by that much, and certainly not by 61 feet. But Chapman wasn't standing at his traditional third base position, not with the lefty Brantley up. He was shifted over, basically playing a shallow shortstop position more than third base. (Hang on to that idea -- we'll get back to it with Gleyber Torres in a second.) You can see in the image below where he'd actually been standing to begin the play.

That's a big deal, and that allows us to get into whats's different about this from other infield metrics. As useful and effective as they've been over the years, DRS and UZR use either the eye test via video scouting or zone-based systems. The Statcast technology allows us to know exactly where each fielder stands, which is helpful in a baseball world where shifting and out-of-position defenders are commonplace.

What that means is that every tracked play is accounted for, regardless of if the third baseman is standing in his regular spot or at shortstop or in short right field. It allows you to know exactly "how far" and "how much time," regardless of shifts. (In 2019, about 35% of pitches, or over 255,000 pitches, came with fielders either in the shift or otherwise not in their standard positions. That's a lot.) UZR has generally ignored shifted plays entirely; DRS hasn't included them on a player basis, though a forthcoming update may change that.

Worth noting, #1: A runner's average sprint speed is used in the calculation, rather than "on that play," because a fielder has to plan for a runner's best, even if on some plays a runner jogs, trips, etc. For new players with no data, a league-average (27 ft/sec) score is used; once the player qualifies for the leaderboard, all of his previous plays are re-run.

Worth noting, #2: Only plays tracked by the system are included. Fortunately, most of the untracked plays are high popups or weak grounders, which are low-probability plays, and there is little to no bias between home ballparks. This is about getting the model into place, with the intention that future hardware updates will collect more data. Numbers may be expected to increase or decrease by approximately 20% for full-time players if all plays are tracked.

What's new? Fielder roles.

Nolan Arenado, as we showed above, was +17 OAA, all coming as a third baseman, because he played every single inning of his season at third base. But that doesn't mean he was always standing at third base, does it? Due to shifting, Arenado's +17 actually breaks down into this:

"+14 where third basemen play, and +3 where shortstops play."

For example, here he is robbing Corey Seager, making a great play from shortstop -- even if he was "a third baseman" in the lineup card.

To explain this further, let's show you screenshots of two different defensive plays for Gleyber Torres in 2019.

They look identical, because they are identical. In each image, Torres is playing 158 feet from home, and he's 32 feet behind and to the right of second base. (Or, 4 degrees away, with 0 being a line from home plate to center field.) They're both at Yankee Stadium. Torres is standing on the same patch of Bronx dirt for both plays, just a few weeks apart.

You can probably see what's happening here. While he's standing in the same spot for both plays, in one, he's listed as the second baseman, shaded over towards the bag, while in the other, he's listed as the shortstop, shifted over to the right side of the bag. He is standing in the same spot. If Torres is doing functionally the same thing on these plays, which is "playing second base," essentially, why are we crediting him as though he's doing something different because of the spot he's occupying in the lineup? We've been talking about the idea of "position-less baseball" for years. This is it.

You can see how that impacts him on a seasonal basis. If we just look at performance while in the lineup as a second baseman or a shortstop, we can see that Torres performed at roughly the same slightly below-average level at both spots.

Torres while listed as a 2B: -4 OAA
Torres while listed as a SS: -3 OAA

But if we look at him where he's standing, you can see the story is different. (The numbers have slightly different sums due to rounding.)

Torres while standing at 2B: -5 OAA
Torres while standing at SS: 0 OAA
Torres while standing at 3B: -1 OAA

By this view, Torres was an essentially average shortstop, not a below-average one. To make this a little more visual, here's a play he failed to make in late May. (It had a 90% likelihood of being an out, because he was only 5 feet away from the intercept point, though it was scored a hit.) Because he was playing shortstop at the time, with DJ LeMahieu at second base, it hurts his rating as a shortstop. But he wasn't really being a shortstop, was he?

It's sort of the same for this Marwin Gonzalez single, or, especially, this Chance Sisco "single." Those are all plays that should have been made while he was a shortstop, but not while he was at shortstop.

You can slice-and-dice this for any infielder yourself. For example, here's Chapman in his various roles, where you can see that despite playing every single inning as a third baseman, he also had 55 tracked opportunities hit to him in the shortstop spot. He was +2 OAA in those chances, making him one of the Top 25 players in the shortstop spot, despite his limited chances there. In addition, breakdowns have been added to player pages, too.

Here you can see that shortstop Elvis Andrus was +4 playing on the second base side of the bag, but Brandon Crawford was minus-4 there. Here, you can see that Christian Walker (+8) and Matt Olson (+7) were the best first basemen playing close to the line, but that Olson (+4), Paul Goldschmidt (+3), and Joey Votto (+3) were the best playing towards second base. The two best second basemen playing on the left side of the bag were Jonathan Schoop (+3 in that area) and Max Muncy (+2), while Ozzie Albies (-3) struggled there.

What's new? Field visuals.

Numbers are fun, and useful. But what about visuals? On each Baseball Savant player page, you can find an infield grid chart that shows a player's performance, selectable by either start point or end point, with red being good, and blue being poor. For example, let's look at Boston's Rafael Devers, who had one of the largest year-over-year jumps from -7 OAA in 2018 to +7 in 2019. That aligns with the hard work that he put in on his fielding.

“It just all comes down to the experience [on] defense,” Devers said in October. “I know there were some errors I was making last year where I was just like, ‘Man, how did I make that?’ This season, I know I’ve improved. Obviously I’ve made some errors here and there. But I’ve just continued to learn on how to improve.”

You can see that on his yearly chart, which shows not only how he's gone from below average to above, but where. The squares represent his various starting points, and the size of the box reflects the number of chances he received in each one (larger box equals more chances).

What's new? Monthly breakdowns.

Remember when Mets shortstop Amed Rosario's defense was considered to be so poor that in June, the team admitted that they were thinking of trying him out in center field? A mere six weeks later, then-manager Mickey Callaway said that suggestion was no longer in the picture, due to Rosario's in-season improvement.

“The one thing we did focus on ... was to continue to try and help him develop at short,” Callaway said on Aug. 2, “and it seems like that’s really paying off. So I think for now that we’ll probably continue just to focus at short and go from there.”

The eye test pretty clearly showed that to be the case, but it's generally been difficult to prove out in-season trends with defensive data. What did the Statcast numbers say? Over the course of the full season, only four qualified shortstops rated weaker than Rosario's -6. But if you look at it on a monthly basis ...

April: -6
May: 1
June: -2
July: -1
August: +1
September: +2

... you can see what Callaway was talking about. In April, Rosario was nearly unplayable at short -- this is why you start hearing stories about a position move. From May on, he was considered a league-average defender -- which aligns well with the eye test.

You can play around with that, too. Marcus Semien was +4 in May (tied for the best shortstop that month) but -1 in September. The best player standing at second base, regardless of position, in June, was Adam Frazier (+4); in August, it was Cesar Hernandez (+4).

Who are the biggest differences from DRS and UZR?

There's a lot of similarities between the rankings, of course. Every defensive metric thinks that Simmons and Story, for example, are good. We're not here to tell you otherwise, nor really to get into "right" and "wrong," so much as: What's different? There's a few notable cases. (Do note here that this is not apples to apples, as "outs" and "runs" are not the same thing, though the scales are similar and people tend to perceive them similarly.)

Examples of infielders OAA grades out more highly

Freddy Galvis // +12 OAA, vs -2 DRS and -1.3 UZR
Justin Turner
// +4 OAA, vs -7 DRS and -6.8 UZR
Carlos Correa
// +9 OAA, vs +1 DRS and +0.1 UZR
Hunter Dozier
// 0 OAA, vs -15 DRS and -5.3 UZR

Galvis is an "outstanding defensive shortstop," Reds manager David Bell said in August, so his position here is nice to see.

What about in the other direction?

Examples of infielders OAA grades out more poorly

Didi Gregorius // -13 OAA, vs -5 DRS and +0.2 UZR
Marcus Semien
// -4 OAA, vs +4 DRS and +6.4 UZR

Gregorius also hit just .238/.276/.441 in his return from Tommy John surgery, so below-average defense along with that isn't terribly surprising.

Finally, how about notable players where current metrics simply don't agree?

Examples of infielders OAA where UZR and DRS disagree

Xander Bogaerts // -3 OAA, vs -20 DRS and +1.2 UZR
Rafael Devers
// +7 OAA, vs -6 DRS and +3.2 UZR
Nick Ahmed
// +16 OAA, vs +18 DRS and +2.1 UZR
Jorge Polanco
// -16 OAA, vs +1 DRS and -8.8 UZR
Amed Rosario
// -6 OAA, vs -16 DRS and -1.1 UZR

Who are the biggest surprises?

We'll admit that when we fired up the leaderboard and saw Fernando Tatis Jr. fourth from the bottom at minus-14 OAA, we were surprised, because he's one of baseball's most exciting young talents and capable of making some breathtaking plays. We watched a lot of his video in an attempt to break this down, but eventually the standard stats did the best job of telling the story: While 26 shortstops played more innings than his 731 1/3, no one in baseball at any position had more throwing errors than his 14. (Those plays alone cost him over 9 OAA. On plays where he didn't make an error, he was a more reasonable -4 OAA.)

That makes the poor rating track, though it also provides hope that simply by cleaning up the easy mistakes, he can be elite.

We thought Gio Urshela, owner of an excellent defensive reputation, would rank higher than his -2, though he also came in with a -4 DRS and a -2.5 UZR, so perhaps that aligns.

There are, surely, endless stories behind each of these metrics. You can find the new infield defense OAA numbers at Baseball Savant, and they'll be updated daily once the new season gets going.