The automated ball strike system, and what it actually changed

A hitter takes a pitch on the outside corner, the plate umpire's right hand goes up, and the hitter taps the top of his helmet. He has about two seconds to do it and nobody in the dugout is allowed to help him decide. Fifteen seconds later an animation of the pitch appears on the videoboard, the crowd learns whether he was right, and play resumes with either a new count or one fewer challenge in the bank.

That is the automated ball strike system as major league baseball actually uses it. Almost everyone calls it robot umpires, which is the one thing it deliberately is not. The plate umpire still calls every pitch. The machine is silent unless somebody spends something to make it talk.

The distinction is not pedantry. It is the whole design, and it explains nearly every rule attached to the system, including the odd ones.

Why the automated ball strike system is a court of appeal

Full automation was tried first, and it worked in the narrow sense that the calls were consistent. What it produced was a game with a strange texture. Every pitch was correct and nobody could argue about any of them, which removed a category of tension the sport had been built around for a century and a half. The catcher stopped mattering behind the plate. The umpire became a man who relayed a verdict he had not formed.

The challenge model keeps the umpire as the decision-maker and puts the machine behind him as an appellate court that only sits when a party petitions it. That has three consequences worth understanding before any of the detail.

The first is that the great majority of pitches are still called by a human being with the ordinary human error rate. A club with two challenges cannot fix a game's worth of missed calls. It can fix the two that matter most, assuming it correctly identifies them, which it often does not.

The second is that the system converts umpiring accuracy into a resource that has to be rationed. This is a genuinely new thing in baseball. A missed strike three in the second inning and a missed strike three in the ninth are the same error, but under a challenge system they are not the same event, because spending on one means not having it available for the other.

The third is that the appeal is public and instant. The pitch is redrawn on the board in front of forty thousand people while the pitcher stands on the mound. There is no room off the field, no crew chief on a headset, none of the machinery that makes football's video review feel like an administrative process happening somewhere else. It is closer to the way cricket's referral system works, in that the player owns the decision to ask and everyone watches the answer arrive.

• 2Challenges each club starts with
• 17Width of the zone, inches
• 53.5Top of the zone, per cent of batter height
• 27Bottom of the zone, per cent of batter height

Rule-defined values published by the league. The width is the width of home plate. The vertical bounds are proportions of the individual batter's measured standing height, not fixed distances.

What the automated ball strike system actually measures

Twelve Hawk-Eye cameras are fixed around the perimeter of the field. They track the ball optically, many times per flight, and reconstruct its path as a curve in three dimensions. The data moves off the cameras on a private 5G network rather than over the venue's ordinary connection, which is a small detail that matters a great deal: a review that takes fifteen seconds cannot be sharing bandwidth with the concourse.

The output is a trajectory, not a photograph. That is worth sitting with, because it is the source of most of the confusion about what the system can and cannot do. Hawk-Eye does not look at a still image of the ball beside the plate and judge it. It fits a model to a sequence of observations and then evaluates that model at a chosen point. Everything downstream depends on where the chosen point is.

The reported resolution is fine, in tenths of an inch, and the league has published confidence bounds around it that are smaller than the diameter of a baseball's stitching. The measurement is not the weak link. The definitions applied to the measurement are where all the argument lives.

1. 1Twelve cameras see the ballOptical tracking around the perimeter of the field, capturing the ball repeatedly through its flight rather than at one instant. No sensor is in the ball, the bat, or the plate.
2. 2A trajectory is fittedThe observations become a modelled path in three dimensions. This is why the system can answer questions about points on the flight it never directly photographed.
3. 3The batter's zone is loadedEach hitter has a personal top and bottom, computed once from a measured standing height and applied identically at every ballpark for the whole season.
4. 4The path is evaluated at one planeThe system asks where the ball was as it crossed a flat rectangle standing at the midpoint of home plate. Not where it started, not where it finished, not where the catcher caught it.
5. 5A binary answer is returnedStrike or ball, with no tolerance band and no benefit of the doubt in either direction. A pitch a tenth of an inch outside is a ball.
6. 6The animation is rendered and shownThe same picture goes to the videoboard and the broadcast simultaneously, which is the point: the verdict and its reasoning arrive together and cannot be quietly disputed afterwards.
7. 7The count is corrected and play restartsIf the call is overturned the count changes, and the umpire then decides whether anything that happened on the bases in the meantime was affected by the original call.

The measurement chain behind a single challenged pitch. Every step is either a published property of the system or a direct consequence of the zone definition; the timing figure is the review length the league has described.

The zone is cut from the batter's own height

This is the part that people find hardest to believe, and it is the most consequential rule in the system.

The ABS zone is not a fixed box. Its top edge sits at 53.5 per cent of the batter's height and its bottom edge at 27 per cent. Both are proportions of a measurement taken with the player standing straight up, in socks, by independent testers during spring training. The width is the width of home plate, seventeen inches, and it does not vary.

Subtract one from the other and the zone is always 26.5 per cent of the batter's height, whoever he is. A shorter hitter gets a shorter zone in absolute terms, which is what the rulebook has always intended and what human umpires have never quite managed to deliver.

5ft 6in batter17.5in

6ft 0in batter19.1in

6ft 6in batter20.7in

Worked example with invented round heights. The arithmetic is the published rule: top at 53.5 per cent of standing height, bottom at 27 per cent, so the zone is 26.5 per cent of the batter's height in every case and the width never changes.

Show the numbers

How tall the zone is for three invented batters

Item	Height of the zone
5ft 6in batter	17.5in
6ft 0in batter	19.1in
6ft 6in batter	20.7in

Three inches of difference between the tallest and shortest hitters in that invented set is more than it sounds. A pitcher who lives at the top of the zone is throwing to a target that moves by a hand's width depending on who is standing there, and the target no longer moves when the hitter crouches.

That last clause is the real change. Under the rulebook, the zone is defined relative to the batter's stance as he prepares to swing, which means a hitter who sinks into a deep crouch is entitled to a smaller zone. Under ABS, he is not. The percentages are applied to a standing height recorded weeks earlier, in a gymnasium, wearing no cleats. A hitter can no longer buy strikes back by folding himself in half, and several who had built careers on exactly that adjustment discovered it in the first month.

The measurement protocol was tightened for the same reason. Listed heights in baseball had always been soft, drifting upward through a career the way they do in every sport, because nothing depended on them. Once the number sets the zone, everything depends on them, and the league took the measuring out of the clubs' hands entirely.

Why the plane sits in the middle of the plate

The rulebook zone is a solid. It is the space over home plate, bounded above and below by parts of the batter, and a pitch is a strike if any part of the ball passes through any part of that space. Home plate is seventeen inches deep as well as seventeen inches wide, so the rulebook zone is a three-dimensional wedge that a pitch can clip at the front, at the back, or anywhere between.

The ABS zone is a flat rectangle standing upright at the midpoint of the plate, eight and a half inches from the front edge and eight and a half from the back. The system asks one question at one place: where was the ball there?

Two things follow, and they pull in opposite directions.

A big breaking ball that crosses the front of the plate above the letters and lands in the dirt behind it may be entirely inside the zone at the midpoint. Under the rulebook it clipped the front of the wedge and was a strike anyway. Under ABS it is a strike for a different reason. A curveball at the other extreme, one that arrives high and drops through the back edge, will fail at the midpoint even though it passed through the rulebook zone somewhere. The plane is not more generous than the wedge or less generous. It is differently generous, and the difference falls almost entirely on pitches with vertical movement.

That is the design compromise, and the league made it knowingly. A three-dimensional test is easy to describe and horrible to explain in an animation on a videoboard in fifteen seconds. A single plane produces a picture that a viewer understands instantly: a rectangle, a circle, and whether the circle touched it. The system is designed to be seen, and being seen cost it a dimension.

The other quiet simplification is the width. Experiments ran with a nineteen-inch zone, which is the plate plus roughly the diameter of a baseball on either side, an attempt to approximate the "any part of the ball" language of the rulebook. The published system uses seventeen. The consequence is that the outer edge of the ABS zone is stricter than the outer edge of the rulebook zone by about the width of a ball on each side. If you have ever wondered why a pitch that looked like a strike on the broadcast box gets overturned, this is usually why, and the rulebook strike zone is doing something subtly different from the rectangle you are shown.

The challenge, step by step

The procedure is short and every clause in it is doing work.

1. 1The pitch is taken and the umpire calls itNothing about the call changes. The umpire's judgement is the ruling on the field and remains so unless it is successfully appealed. A pitch that is swung at cannot be challenged at all, because there is no ball or strike judgement to review.
2. 2The pitcher, catcher or batter taps his cap or helmetOnly those three, and only immediately. The window is measured in a couple of seconds, not in the time it takes to look at the dugout. A late tap is not a challenge, it is a player touching his hat.
3. 3He says so out loud to the plate umpireThe tap alone is deliberately ambiguous, so the request is also vocalised. This is what stops an accidental adjustment of a helmet strap from starting a review.
4. 4The umpire checks that nobody helpedA player seen taking a signal from the bench or from a team-mate can have the challenge refused before it is heard. The rule exists because a dugout with a monitor would otherwise be challenging every close pitch, which turns two challenges into a second replay system.
5. 5The system is queried and the answer is drawnRoughly fifteen seconds. The pitch is animated on the videoboard and the broadcast at the same moment, so the verdict is public before anyone can characterise it.
6. 6The call stands or is reversedIf the umpire was right, the challenging club is down one. If he was wrong, the count changes and the club keeps its challenge. Only losing costs anything.
7. 7The consequences on the bases are sorted outIf the original call caused someone to do something, a catcher who stopped throwing because he heard ball four, a runner who kept going because he heard strike three, the umpires have discretion to unwind it. A stolen base gained purely from the wrong call can be taken back.
8. 8Play restarts, and the ledger is updatedThe challenge count is what everyone in both dugouts now watches, because it determines what either club can afford to argue about for the rest of the night.

The published sequence for a single challenge. The two-second window is the practical timing the league has described rather than a number written into the rule; everything else is the rule as announced.

The economy of it is the interesting part. A club begins with two and keeps every one it wins. In principle a catcher with a very good eye never runs out. In practice roughly half of all challenges are upheld, which means the average club is spending its allocation at about the rate a coin would.

From the tenth inning onwards, a club that has none left is handed one at the start of each extra inning. It does not accumulate: a club that goes into the twelfth with a challenge in hand gets nothing extra. The provision exists because a long game under the old arithmetic would otherwise end with both teams exhausted of appeals in exactly the innings where a single ball-strike call decides everything, and what happens in extra innings is already unusual enough without that.

Two challenges is a budget, and budgets change behaviour

Watch a club that has spent both of its challenges in the fourth inning and you will see the pitcher work differently. The edges get less attention. Pitches that would have been thrown at the corner in the hope of a favourable call get moved half a ball inward, because a favourable call is now permanent and an unfavourable one is too.

That is the second-order effect nobody predicted loudly enough. The challenge system does not only correct calls, it changes the price of being on the edge, and the price is different depending on how much appeal budget each side is carrying.

The minor league data on when players spend showed the shape of it clearly. Challenges cluster on full counts and hardly appear on first pitches, which is exactly what a rational actor would do with a scarce resource. A missed strike on 0-0 costs a fraction of a run in expectation. A missed strike on 3-2 ends the plate appearance and may end the inning. Players worked this out without anybody publishing a table telling them to.

The refinement, and it takes a season or two to develop, is that the right time to challenge is not simply the highest-leverage moment. It is the highest-leverage moment where you are also most confident. Those are not the same, and confidence is the harder half, because a hitter's view of a pitch on the inside corner is the worst view on the field.

Why the catcher is the best person to spend one

He sees the ball arrive from directly behind it, at rest, with a known reference frame in the shape of the plate under him and the batter beside him. He catches thousands of pitches a season and has a trained sense of where each one finished. He is also the only one of the three who is not emotionally invested in the outcome of that particular pitch in the way a hitter with two strikes is.

Catchers56%

Batters50%

Pitchers41%

Success rates by the player who made the request, from the major league spring training trial of the challenge system. The league publishes running challenge data by role and by club on its Baseball Savant leaderboard. Roughly half of all requests across all three roles were upheld.

Show the numbers

Who wins their challenges

Item	Challenges upheld
Catchers	56%
Batters	50%
Pitchers	41%

The pitcher is the worst of the three and it is easy to see why. He is sixty feet away, looking down a line where depth perception is nearly useless, and he has just thrown the pitch he believes should be a strike. He is the only one of the three whose judgement is contaminated by intent. Clubs have quietly discouraged their pitchers from tapping.

The catcher's advantage also creates a small internal conflict, because the same man is responsible for presenting borderline pitches persuasively in the first place. A catcher who has just made a ball look like a strike and got the call has no reason to challenge, and one who has made a strike look like a ball has just cost his own pitcher. The framing skill and the challenge skill are two halves of the same job and they occasionally point in opposite directions.

What the automated ball strike system does not touch

The list of exclusions is short and every item on it is deliberate.

Swings are not reviewable. If the batter offers at the pitch, there is no ball-or-strike judgement to appeal, so there is nothing to challenge. A checked swing that the plate umpire rules no swing still goes on appeal to the base umpire, in the old way, with the old arm-out gesture. The single most argued call in the sport was left entirely alone.

Hit-by-pitch stays with the crew. So does catcher's interference, so does the batter stepping out of the box, and so does every call on the bases, which remains inside the ordinary video replay system with the manager's own separate allocation. Both systems can be triggered on the same play, and when they are, the ball-strike question is settled first and the replay on the bases is heard afterwards. The order matters, because the strike-ball answer can change what the play on the bases even was.

Nothing can be challenged when a position player is pitching. This looks arbitrary until you remember why position players pitch: the game is a blowout and everyone wants to go home. Allowing challenges there would let a losing club spend fifteen seconds a pitch to no purpose, and the rule is a small piece of anti-farce legislation of the same family as the restrictions on when a position player may pitch at all.

And the plate umpire is not being scored in public by the system on every pitch. He is being appealed against on a handful. This is a meaningful difference from the way umpire accuracy was discussed before ABS existed, when a pitch-tracking overlay on a broadcast produced a running indictment that nobody in the stadium had asked for and no player could act on.

The pitches that get overturned, and the ones that never do

Not all borderline calls are equally likely to be reversed, and the pattern is a direct consequence of the geometry rather than of umpire weakness.

Start with a pitch that misses laterally. A fastball six inches off the outside corner is not close under any definition and nobody wastes a challenge on it. A fastball an inch off is close, but the umpire's horizontal judgement is the strongest part of his game, because he is looking straight down the line of the plate with the catcher's glove and the plate's own edges as references. Human horizontal error is small. There is not much for a challenge to win.

Vertical calls are the opposite. The umpire has to judge the top of the zone against a moving human being who is bending, striding and rotating, and he has to do it against a boundary that the rulebook defines with a phrase rather than a number. He is estimating. ABS is not. The gap between an estimate and a computation is where nearly all the overturns live, and it is why a hitter who has just taken a high fastball is far more likely to be worth listening to than one who has just taken one on the corner.

Then there is the breaking ball, which is its own case. A pitch with heavy vertical break has a different height at the front of the plate, at the midpoint and at the back, and the single-plane test picks one of those three. A sweeper that finishes well outside can be a strike at the midpoint. A big overhand curveball that looks like it drops through the zone can fail at the midpoint because by then it has already passed under it. Pitchers who have spent years designing pitches around a shape rather than a location now have a precise definition of where that shape is being measured, and the ones who adjust fastest will be the ones who understood that the measurement point was the variable all along.

Who else can afford one

Twelve tracking cameras, a private network and the staff to run them are not an expense a college conference absorbs casually, and they are wholly out of reach for high school baseball. This is the least discussed and most consequential feature of the whole thing: the rule is now different at the top of the sport from the way it is everywhere else, and not because anybody wrote two rulebooks.

A player coming up through amateur baseball learns a zone defined by a stance and adjudicated by a person. He arrives in professional baseball and finds a zone defined by a percentage of his standing height and adjudicated by a fitted trajectory. Those are not the same skill. Learning to take the high fastball because the umpire in your league does not call it is now a habit that will cost you, and unlearning it takes time that nobody has budgeted for.

Some conferences have run the challenge system at their end-of-season tournaments, where the games are concentrated in one venue and the equipment can be installed once. That is the realistic diffusion path: the machinery follows the money and the fixed venues, so it will reach championship events long before it reaches an ordinary Tuesday. For most of the people who play baseball, the automated strike zone will remain something that happens on television.

The road from the independent leagues to the majors

The system did not arrive fully formed, and the shape of the testing explains the shape of the rule.

It was first put in front of paying spectators in the independent Atlantic League, whose willingness to be baseball's laboratory has given the sport several of its recent rules. The early version called everything, using a different tracking technology from the one now in use, and it was crude. The Arizona Fall League, where the best prospects play in front of scouts, took it next. Then low-A, then the Florida State League, where the challenge model itself was first tried.

Triple-A is where the two models were put side by side. For a period, some games were called entirely by the machine and others used the challenge format, and the league had a controlled comparison of the two in front of the same players in the same parks. The challenge model won that comparison, and Triple-A moved to it exclusively partway through a season, which is not something a league does unless the evidence is one-sided.

The major league sequence after that was spring training first, in a subset of ballparks, then an appearance in the All-Star Game, which is the sport's traditional venue for showing a rule to a national audience with nothing at stake. Then the competition committee approved it and it took effect across spring training, the regular season and the postseason at once.

Full automation, incidentally, is not a hypothetical. It has been used at the top level of professional baseball in South Korea, where the KBO League calls every pitch by machine and relays the verdict to the umpire. The two paths exist side by side and the sport has a real comparison rather than a thought experiment.

Precision, and the argument about the buffer that is not there

The system reports positions in tenths of an inch. The league has published confidence figures showing that a reported location is very close to the true one, with the typical error small enough to be invisible on any animation a viewer will ever see.

None of that is the contentious part. The contentious part is that the answer is binary and there is no tolerance band.

A human umpire operates with an implicit buffer. So does the system that evaluates human umpires, which has always allowed a margin around the zone edge before counting a call as wrong, on the sensible ground that nobody should be graded to a precision finer than the measurement. ABS has no such margin. A pitch whose modelled centre sits a tenth of an inch outside the rectangle is a ball, and the animation will show a circle apparently kissing a line, and the crowd will boo, and the pitcher will look at the sky.

There is a reasonable argument that a buffer should exist, and a reasonable argument against. The case for is that a system with a known error should not resolve differences smaller than that error, because on those pitches it is not adjudicating, it is generating a coin flip with extra steps. The case against is that any buffer has to have an edge of its own, and that edge will produce exactly the same complaints one ball's width further out, with the added problem that the rule now contains a number nobody can justify from first principles.

The league went with the sharp edge, which is the right call for a system whose main product is not accuracy but the visible appearance of accuracy. A rectangle either contains a circle or it does not. A rectangle with a fuzzy border around it is not something you can show on a videoboard between pitches.

What this did to the rhythm of a game

The fifteen seconds are not free. A challenge is a stoppage in a sport that spent the previous few years systematically removing stoppages, and the two rules pull against each other on purpose. The clock speeds the game up; the challenge stops it, briefly, at moments that are almost by definition tense.

Whether that is a net loss depends on what you think a break in play is for. Fifteen seconds of a pitcher walking round the mound is dead time. Fifteen seconds of forty thousand people watching an animation resolve is not, and the noise when it lands the right way is one of the few genuinely new sounds in baseball. The stoppage has content.

The more interesting cost is to the argument. Baseball's oldest performance is a manager coming out to complain about a strike zone, and there is now a category of complaint that has been settled before it can be made. That is not a small subtraction from the game's texture, even for people who think the calls should be right. The sport traded some of its theatre for accuracy, in a limited quantity, twice per club per night, and it did so with its eyes open. Whether the trade was worth it is a question about what baseball is for, and no tracking system is going to answer it.

Four things to check on any challenge

When the next one happens, these are the things worth watching, in the order they tell you something.

Who tapped. A catcher's challenge is a different proposition from a pitcher's, and the difference is large enough to see over a season. If the pitcher is doing the tapping, either the club is undisciplined or the catcher has already lost confidence in his own read.

What the count was. A challenge on 0-0 is either a very strong read or a wasted asset. On a full count, in a close game, it is exactly what the budget exists for.

How high the pitch was. Overturns cluster at the top and bottom of the zone, because that is where the rectangle's edges have been redefined as proportions of a height rather than judged against a stance. A pitch three inches off the outside corner is not being overturned; a pitch at the letters might be.

What the score and the inning were. A club that spends both challenges before the sixth inning has told you something about how it values the present against the future, and if the game is tied in the ninth, it will have told you something else.

Two of those you can see live. The other two you have to remember, which is a reasonable description of what watching baseball has always required.