Using the eye test vs. defensive metrics for Richie Martin

BOSTON - Orioles shortstop Richie Martin darted to his left in a game this week in the Bronx, slid to gather a groundball hit by Yankees outfielder Brett Gardner, and popped back up to his feet and fired to first base to record the out by a half-step.

He earned a full style point for the maneuver.

Six innings earlier, as the Yankees were batting for the first time, Martin took a few steps to his right to handle a bouncer from Aaron Judge and threw to second base, but the Orioles couldn't turn what appeared to be a routine double play.

Martin, in his first season in the majors as a Rule 5 pick, keeps passing manager Brandon Hyde's eye test. Many fans on social media have praised his work and noted that it's worth the light batting average he carries. But the defensive metrics aren't nearly as generous.

Baseball-Reference.com grades Martin's dWAR at minus-0.4 and FanGraphs grades his defense at minus-2.6, UZR (ultimate zone rating) at munus-6.0 and UZR/150 at minus-19.5.

There isn't a faster player on the team than Martin and he's viewed by many as an upgrade at shortstop over Tim Beckham, who moved to third base to start the 2018 season and returned to his old position after the Manny Machado trade.

An eye test argument also can be made that Martin, with seven errors and a .975 fielding percentage, is the club's best option at shortstop. Hanser Alberto hasn't be used at the position this season and Jonathan Villar is inconsistent, with a few dazzling plays made lately but also his share of mistakes. He's committed 10 errors for a .945 percentage.

Crunch the numbers, however, and Villar's minus-0.9 defensive rating, minus-3.1 UZR and miunus-13.8 UZR/150 actually place him ahead of Martin at shortstop. Villar also plays second base and Baseball-Reference.com gives him a minus-0.4 defensive grade.

No one in the group is vying for a Gold Glove.

The Orioles have access to the necessary data. Their front office knows Martin's strengths and weaknesses. They just aren't eager to share it with the world.

Assistant general manager Sig Mejdal, brought to Baltimore by new executive vice president and general manager Mike Elias to drag the Orioles into the analytic age, explained the deficiencies in some of the grading systems. He didn't specifically break down Martin's game.

"I think Baseball-Reference uses a UZR-type of metric, where it tries to estimate the opportunities the player had and then it looks at his successful outs, and prior to Statcast, it was the best we could do," Mejdal said.

"Different models were arguing about how to estimate the opportunities the player had, but the combination of Statcast and then the positioning that's going on in baseball, that has really made the UZR-type of metrics of no use. Right? Because it doesn't know where the player starts, and once you know that, it becomes useless."

A shift, in other words, in how we view the game.

"If the players were starting in about the same spot, if they were being positioned similarly, they could still correlate very well to the improved metric, the Statcast-like metric. But they aren't and it's not even close. And so to ignore that, it really takes a bite out of the usefulness of the publicly available defensive metrics," Mejdal said.

"I can imagine Richie Martin, when he's on the other side of the bag and a ball is hit to where a shortstop would play conventionally, it's likely deducting him a lot and it's likely rewarding him a lot when he makes a play on the other side of the bag and you're just hoping it evens out. But if you look at how often he's on the other side of the bag compared to the shortstop on the team that doesn't shift very much, it doesn't even out and that's a problem, while five or six years ago that wasn't as much less of a problem.

"Those metrics are useful in and of themselves if you had nothing else, but they're much less useful than they were five or six years ago when there wasn't all this shifting. Now that said, there is now a much better way that makes, once you have the Statcast data where the guy was standing, where the ball was hit, the time he had to get to it, the time it took him to accelerate, to get to the ball, the time it took him to exchange the ball and release it, and the time the ball was in the air to the first baseman, when you have all that the UZR is of no use. And so the public doesn't see that data."

Mejdal has found when clicking on the more popular sites that outfield leaderboards are more prevalent than the infield variety. But there also will be a shift in that area.

Meanwhile, the clash between the metrics and what some people are observing and interpreting - which Mejdal doesn't dismiss as useless - will resonate through the sport.

"The eye test is great, though," Mejdal said. "It's easy to see arm strength and it's easy to see athleticism, but sometimes what you don't pick up on the eye test is initial reaction and ability to accelerate."

While Mejdal speaks in general terms more than aiming his critiques at Martin, the rap on the rookie from some talent evaluators while comparing him to plus shortstops has included his slower release and acceleration and a need to improve accuracy.

Jeff Todd of MLBTradeRumors.com, in a write-up earlier this month on Rule 5 picks who stayed in the majors, referred to Martin as "a distinct negative at shortstop" and linked the FanGraphs numbers.

Hyde is more likely to go by what he's observing from the dugout.

"Consistency," he said. "I think he's been really consistent all year. Not that many errors and makes the routine play the majority of the time. I just like the way he plays defense. He plays with energy and he's always in the right spot. You want your middle of the infield guys to be consistent and make the routine play on a nightly basis and Richie's done that for the most part this season."

The offensive components are culled and shared on Twitter on a nightly basis. Exit velocity and launch angles are prime examples. But the defensive side of the sport is evolving.

"They're putting it together," Mejdal said. "It's every once in a while they have something else on the Statcast leaderboards. They've got catcher framing, outfielder jump, catch probability, so for whatever reason they seem to be focusing on the outfielders first.

"But if you look at that, they have outfielder's jumps, they have reactions in burst and route efficiency, so they've broken it down and anybody who has the Statcast data and somebody with some college-level analytical skills, you could do something of that family and then further break down the value and capabilities of the shortstop, just like they do with the outfielders. His reaction and his burst or his ability to accelerate."

Mejdal and his staff will stay busy gathering information and separating what is and isn't applicable, funneling it to Hyde and others in the organization.

The shifting that's frequently done with the Orioles infield, including how Martin is set up on the right side past the bag, must be taken into account with some metrics that are routinely tossed around.

"I guess I would say the public available defensive data relies on all things being equal and those things are that the shortstops are positioned similarly," Mejdal said, "but it can't distinguish between a shortstop who got to the ball because of his skills or one that was positioned there for other reasons. And in the past, that wasn't a giant deal, but now with the different starting points of the shortstops, it is a big deal and it's a challenge to the models and undoubtedly it's creating a lot of noise.

"For the teams that shift a lot or shift a little, it's not only creating noise, it's creating a large bias. And so the accuracy of those have gone down as shifting has gone up and that's all the general public has.

"The Statcast data, it's not a problem. The calculus is much different and when you have that Statcast data you don't rely whatsoever on UZRs."

Mejdal is building his analytics team, with more hires to be made.

"We're not at the final head count, but we have a very capable analytics department," he said. "We have two full-time analysts and additional developers, plus the one (Di Zou) that was here. And we have four or five very skilled interns. They aren't a family member who's gotten in. These are recent graduates, some with advance degrees that are very capable. So we've got close to a dozen people.

"There will be likely another two hires in this next year."

Zou is the holdover, a developer who understood that his input would be valued in 2019 and beyond.

"He stayed with the hopes that the new front office will prioritize analytics a bit more," Mejdal said. "So he stayed with that hope."