The battle between the pitcher and the hitter is the ancient core of baseball. Everything in this team sport emanates from that one-on-one duel at 24 paces contested an average of 76 times each game. The odds are stacked in the pitcher’s favor. The pitcher initiates the action. The hitter does not know what is coming. Measured by keeping the batter off base, the pitcher wins 68% of these duels.

But what if the hitter knows what’s coming? The ancient balance flips. Such knowledge can be gained legally, such as a runner at second base relaying signs to the hitter, or illegally, such as the 2017 Astros misusing in-game technology to steal and relay signs.

This season, a new element is changing how batter-pitcher matchups are won and lost—and by extension, how games are won and lost on a nightly basis. High-speed, multicamera systems with deep learning algorithms are tilting outcomes and setting off full-blown paranoia about pitchers tipping their pitches like never before.

While legal, the practice is raising a question within the game: How much should technology, not just player skill, influence the actual competition? The sideways glance of Yankees slugger Aaron Judge last month before hitting a home run off Toronto reliever Jay Jackson—and the subsequent two-day contretemps it prompted between Toronto and New York—provided the most striking visual clue to an otherwise quiet revolution.

“I spend more time on combating pitch tipping than game planning,” says one MLB pitching coach, who requested anonymity. “It’s not just watching video and picking something up a pitcher does anymore, or position players seeing something at second base with the grip in the pitcher’s hand. [It’s] way more elaborate, and I think destructive to the game.”

Says a high-ranking club executive, “He’s right. It is absolutely eating up huge amounts of time and energy like never before. We came out of a [road] series and we had all of our pitchers who got hit looking at how they held their glove, how they came set ... just paranoid about everything they might be doing that otherwise might not have been able to be seen in real time by the naked eye.

“Even if what teams are doing is legal, the danger is how much further teams are willing to go, and do they cross over that line? That’s what happened with about five teams years ago with sign-stealing. The Astros got caught because they went further than other teams. That’s the worry: How far do people take it?”

Says one bench coach, “It’s really alive. Some teams [are] advanced. ... We are playing catch up.”

A source with MLB said the league is aware of this new paranoia over pitch tipping but, as it found with the Judge home run, is aware of only “legal and permissible activity.”

Toronto Blue Jays pitcher Jay Jackson admitted his grip and his timing may have tipped off the Yankees.

Kirby Lee/USA TODAY Sports

The technology being used has been around MLB for seven years, though it has been more widely adopted in recent seasons. In 2019, for instance, five teams worked with KinaTrax, a multicamera system in ballparks that provides markerless, 3-D motion capture from up to 16 cameras, half dedicated to the pitcher and half to the hitter. Today 15 MLB franchises work with KinaTrax.

With built-in safeguards against actual sign-stealing, the system is designed to be used in real time to assess and improve performance through biomechanical markers and data. For instance, KinaTrax can measure a pitcher’s landing leg flexion, the pronation of the forearm upon ball release, the internal and external shoulder rotation and other biomechanical data points. Through those metrics, for instance, KinaTrax might be able to tell a pitcher is tiring even before the pitching coach. It can be a boon in guarding against pitching while fatigued, one of the dominant causes of pitching injuries.

Those eight camera angles directed toward the pitcher also can be used to provide information about pitching “tells,” the idiosyncrasies that might telegraph which pitch the pitcher is about to throw. Teams devote platoons of analysts to deploy machine learning algorithms to scour those camera angles.

“There are a number of teams that use skeletal overlays on every pitcher with each pitch they throw and figure out the differences, then focus on it in real time,” says the pitching coach. “Any deviation is red flagged to be viewed.

“They also have technology for hitters to view the tip as if they were there in the batter’s box.”

The club executive used this example: In one case, the cameras caught flexion in a pitcher’s forearm muscles as he was gripping the baseball inside his glove to throw a split-finger fastball. The hitter would be unlikely to see that subtle tell from the batter’s box. But a base coach or other team personnel in the dugout, once alerted by artificial intelligence to the tell, could look for the tell from a side angle—usually the “open” side of a pitcher’s delivery—and signal to the batter a split was coming. The signal could be sent by any one of several means, such as posture or an audible signal. One team’s base coach, for instance, would use one hand on the hip for a fastball and both hands down for an off-speed pitch.

An MLB source said it would be virtually impossible to adopt rules to prevent signals from base coaches or dugout personnel to hitters because the simple act of how a person stands could be a signal ... or it could be nothing.

Studying video for tells is perfectly legal, akin to studying which pitch a pitcher favors when ahead of the count with a runner in scoring position. The advancement has come from the machine learning required to scour through so many camera angles as well as the need for a real-time delivery system once those tells are identified.

According to the pitching coach, once teams discover a tell through technology, they “then use elaborate relay systems to get [the pitch information] to the player via the first base and third base coach. Approach is secondary in hitter meetings for certain teams. ... Some teams spend more time on tips than approach.

“I can’t stand [the] direction of [the] game.”

No one in baseball has a problem with old-school, real-time sleuthing to decode what a pitcher might be throwing. In Game 6 of the 2001 World Series, for instance, the Diamondbacks knew every pitch Yankees starter Andy Pettitte was throwing from the stretch, because they noticed he would bow his hands out before coming set at the waist for a breaking ball and hold them closer to his body for a fastball.

PitchCom, the wireless transmission of signals, has removed much of the sleuthing done from second base when it comes to catcher hand signals. But relaying location from second if a catcher sets up too early, for instance, is regarded as fair play (with the onus on the catcher for not being careful enough).

The new method of exploiting tells takes the game in a new direction because it requires machine learning and a delivery system outside of the players who are competing on the field.

“It’s a rather slippery slope,” says the club executive. “I fall on the side of if it’s something you’re picking up in real time, there’s nothing you can do about it. It’s fair game. But this method just doesn’t seem fair to me. And my concern, even if it’s, quote, unquote, ‘legal now,’ is that someone’s going to take it too far.”