The case for AI officiating has arrived. It’s time for racing to stop pretending the technology doesn’t exist
Adam Silver was sitting across from Pat McAfee a few days ago when he essentially announced that professional basketball was done pretending human beings could handle every call in real time.
“We’re going to move to a system like [Hawk-Eye] where that whole category of calls will be automatic,” Silver said, referencing the aftermath of a widely condemned officiating blunder in Game 5 of the Western Conference Finals, a moment where replay clearly showed that Chet Holmgren, not Victor Wembanyama, touched the ball last before it went out of bounds, and the officials refused to change their call even after huddling to look at the evidence together. “Those calls will be done by an AI automated system, with cameras lined around the court. It’ll be instantaneous and automatic.”
Silver was careful to draw a distinction. For plays involving physical contact foul determinations, the judgment calls that require understanding of force, momentum, and intent — he acknowledged that human officials still need to be in the room. “There’s often contact on every play,” he said. “That doesn’t necessarily mean there’s a foul. They’re trying to measure whether that contact is impeding the player, how hard it is. It’s something that can’t just be done on camera.”
That distinction is exactly the right one. And I’ll get back to why it matters enormously for horse racing. But first, let’s understand what Silver is actually talking about — because Hawk-Eye is not new, and the sports world has been living with it for a generation.
What Hawk-Eye Actually Is
Hawk-Eye was developed in 2001 by British engineer Paul Hawkins. It is now owned by Sony and operates at scale across tennis, cricket, soccer, rugby, and a growing list of other sports. The system uses six to ten high-speed cameras positioned at different angles around a playing surface, triangulates the data in real time, and builds a three-dimensional model of ball trajectory or in more advanced configurations, of player and object movement accurate to within 3.6 millimeters.
In tennis, Hawk-Eye’s Electronic Line Calling system has operated on more than 450 courts across over 100 tournaments. The 2025 Wimbledon Championships became the first Grand Slam to dispense with human line judges entirely. When a serve clips the line, you don’t wait for a call, the system sees it, the crowd hears the sound effect, the graphic appears on the big board, and play continues. Instantaneous. Objective. Unchallengeable.
In cricket, Hawk-Eye has been the backbone of the Umpire Decision Review System since 2005, tracking ball movement across more than 1,000 match days annually in 15 countries. The system can project the likely path of a delivery through a batsman’s legs to determine whether it would have struck the wicket — a probabilistic judgment call, rendered with statistical precision, based on the ball’s actual tracked trajectory.
Major League Baseball has been testing its Automated Ball-Strike system, built on TrackMan radar and camera technology, in the minor leagues for years and rolled it into 13 major-league spring training venues in 2025. The system adjusts the strike zone to each individual batter’s size. Umpires still call balls and strikes for now but when a pitcher, catcher, or batter challenges a call, the TrackMan decision appears on the video board and the umpire receives the result through an earpiece. MLB is projecting 2026 as the year the first fully automated AI umpire game becomes a reality.
FIFA has semi-automated offside detection operating at the World Cup level, using connected ball technology with an inertial measurement unit sensor embedded in the match ball, transmitting precise ball contact data in real time. The NFL has acknowledged that AI officiating is coming for boundary lines, ball placement, and out-of-bounds determinations. Roger Goodell has said publicly that “it’s getting harder to officiate” and confirmed the league will move toward automation.
The NBA, as Silver’s announcement confirmed, was already using Hawk-Eye’s optical tracking for goaltending reviews. Now they’re preparing to expand it to cover the entire category of out-of-bounds and line calls — making those determinations automatic, removing them from human hands entirely.
The direction of travel across professional sports is unmistakable. What is being automated first is not the hard stuff. What is being automated first is exactly the category of call that should never have required human judgment in the first place: Did the ball touch the line? Did the player’s foot step out? Did the ball go out last off this player’s hand?
Objective. Binary. Provable by physics.
Horse Racing Has Been Sleeping Through This
Here is where I’m going to say something that the sport’s establishment is not going to want to hear.
Everything I just described, the camera triangulation, the real-time tracking, the objective determination of physical events in space applies directly to thoroughbred horse racing. And we are not using it for officiating. Not in any serious way.
What we are using is slow-motion replay reviewed by stewards who may or may not have agreed on what they’re looking at before they make a ruling, a process that varies from jurisdiction to jurisdiction, racetrack to racetrack, and honestly, steward to steward on any given afternoon.
I have been watching this sport for over fifty years. I have sat in the press box and in the grandstand and I have been at the windows when stewards put up an inquiry sign, and I have watched the crowd hold their breath while a small group of officials in a room somewhere reviewed the same footage the rest of us were watching on the monitors. And here is what I know after all of those years: the biggest problem with stewards is not wrong calls. The biggest problem is inconsistency.
Wrong calls happen. They happened yesterday and they will happen tomorrow. In a sport governed by chaos, multiple horses, narrow paths, sudden acceleration, animals making decisions based on instinct — perfect officiating is not the standard. We have never demanded it and we cannot reach it.
What we can demand is that the same foul, committed by the same type of horse, in the same kind of race, is called the same way every single time.
That is what AI can deliver. That is what it is already delivering in tennis and cricket and baseball and basketball. And horse racing has not seriously engaged with the question of whether it could or should work here.
What AI Can Do in a Steward’s Stand
Let me be direct about where I think the technology is already capable and where it becomes more complex.
The straightforward calls: Did the horse drift off his line into a competitor’s path? Did the jockey’s elbow make contact with another rider? Was there interference between the eighth pole and the wire? Did Paco Lopez raise his wrist above his head in The Preakness? These are spatial events. They exist in three dimensions, they happen at high speed, and they can be captured by multiple high-speed cameras, triangulated, and analyzed frame by frame. The same Hawk-Eye architecture that tells you whether a tennis ball touched the line at 130 miles per hour can tell you whether Horse A crossed into Horse B’s running lane at the three-sixteenths pole. The technology is not in question. Whether anyone in racing governance has the will to pursue it is a different matter.
The harder question: This is where it gets interesting, and where I think we have an honest conversation to have. In horse racing, the standard for a disqualification is not simply that a foul occurred. The standard is whether the foul affected the outcome — whether it cost the fouled horse a better position, a better placing, a better finish. Did the interference cause material impact?
That is a judgment call. It has always been a judgment call. And I would be the first to acknowledge that an AI system reading camera data cannot look into a horse’s heart and know what he would have done if he’d had a clear path. Nobody can know that. What we can agree on — what we can almost always agree on — is whether a foul happened. What divides us is whether it mattered. This is why I personally prefer the “old method” where you commit a foul you go behind the horse you impeded, or a system where the rider is sanctioned and there is no pari-mutual change.
I am not arguing that AI ends the stewards’ inquiry. What I am arguing is that AI should be doing the first part of the job so the stewards can focus on the second part. Remove the question of whether from the hands of subjective human review. Give the stewards a definitive answer on the physical event, and let them apply judgment to the question of impact. That is exactly what Silver was describing in the NBA context: automate the objective calls, free the officials to focus on the calls that genuinely require human judgment.
That is a defensible, achievable, meaningful improvement. And horse racing has done essentially nothing to get there.
The Paco Lopez Exhibit
I do not have to argue this abstractly. We have a live case study that demonstrates with embarrassing clarity what AI-assisted officiating would have prevented.
On May 17, at Laurel Park, Paco Lopez rode Napoleon Solo to victory in the Preakness Stakes. Past The Wire reviewed the broadcast footage. Multiple journalists reviewed the same footage. The images were reviewed by PETA and submitted to HISA. What the footage shows, from the three-sixteenths pole to the wire, is Lopez raising his wrist above helmet level before striking the horse on three separate occasions. HISA Rule 2280(c)(1) prohibits exactly that conduct. It is the same rule Lopez was suspended for six months in September 2025 — the most significant whip violation penalty HISA had ever issued — after a documented pattern of identical violations at multiple tracks.
The Maryland Racing Commission stewards reviewed the same footage and found no violation.
Administrative steward Adam Campola offered an explanation to both Horse Racing Nation and The Racing Biz. Lopez, Campola said, has an unorthodox style. He stays very busy on a horse. Moves around quite a bit.
HISA, having initiated the process of referring the matter to the stewards, announced it would defer to the stewards’ independent review. The stewards reviewed and found no violation. The process was completed. The record shows no written finding, no frame-by-frame analysis, no explanation of how images reviewed independently by multiple journalists were evaluated and dismissed.
I want to be precise about what I am saying. I am not accusing the Maryland stewards of corruption. I am saying that what those stewards did, look at footage and reach a conclusion was a human process subject to human perception, and that human perception in this case produced a result that is inconsistent with what the footage objectively shows.
This is not a hard case for AI. It is not a nuanced judgment about whether contact altered an outcome. It is a geometric question. Is the jockey’s wrist above the line established by the top of his helmet? Yes or no. It is the same question Hawk-Eye answers about a tennis ball and a baseline. It is the same question MLB’s automated ball-strike system answers about a pitch and the edge of the strike zone.
You feed the camera data into the system. You draw a reference line at helmet height. You track wrist position relative to that line in each frame. The system tells you, frame by frame, whether the rule was violated.
It has no relationship with Paco Lopez. It has no awareness of the unorthodox style. It does not know or care that it’s the Preakness Stakes and the purse is two million dollars. It knows where the wrist was relative to the helmet. That is all it needs to know.
The Consistency Problem Is the Integrity Problem
I want to come back to something because I think it deserves to be stated plainly.
When bettors, the people who fund this sport, who make the mutual pools that pay every purse in the country watch a race get decided, or watch an inquiry get resolved, they are not watching a sporting contest. They are watching a financial instrument with a result. The outcome of that inquiry has direct monetary consequences for the people holding tickets.
The willingness to accept inconsistency in officiating, to say it is part of the game, to defer to human judgment even when that judgment demonstrably contradicts what the replay shows is not a defense of the sport’s integrity. It is a threat to it. Because the bettor watching the Lopez footage and comparing it to what the stewards said is not going to conclude that officiating is hard. The bettor is going to conclude that the results are not trustworthy. And once you’ve lost that, you’ve lost everything.
The NBA’s Adam Silver understood this when a single out-of-bounds call in Game 5 of the Western Conference Finals drove him to a national television appearance to announce a coming overhaul of how the league handles objective determinations. The consequences for professional basketball are measured in ratings points and franchise values. The consequences for horse racing are measured in handle and handle has been in structural decline for long enough that everyone in this industry should be asking hard questions about every possible driver of that decline.
Inconsistent officiating that appears and occasionally is immune to the evidence on the screen is not irrelevant to that equation.
Where Do We Start?
I am not naive about the infrastructure challenge. Thoroughbred racing in the United States operates across dozens of jurisdictions, each with its own racing commission, its own stewards, its own enforcement apparatus. There is no Adam Silver who can announce a league-wide policy change and have it implemented by next season. The structure of American racing makes everything harder. Enter Mike Repole.
But 1/ST Racing has already deployed AI camera systems at Santa Anita and other facilities to assist regulatory veterinarians in real-time identification of lame horses. Equibase is collaborating on machine-learning products that provide handicapping insights. The infrastructure for multi-camera AI analysis at major thoroughbred tracks is not a distant science fiction scenario. It is a present-tense capability.
The question is not whether the technology exists. It does. The question is whether the people who govern this sport have the will to require its implementation for officiating purposes — and whether the major racing jurisdictions can be brought to a common standard of what an AI-assisted steward’s inquiry looks like. If the industry will not invest in itself, how does it ask bettors to continue to do so?
The tennis model is instructive. It did not require a global mandate to work. Wimbledon adopted Hawk-Eye and it worked. The Australian Open adopted it and it worked. The US Open followed. The players accepted it because it was more accurate than the line judges it replaced. Nobody wept for the line judges. Nobody argued that the human element of making line calls had intrinsic value that the machines were destroying.
Horse racing can do this. It has the technology available. It has the case studies from other sports. It has the documented evidence — sitting in a PTW article and in multiple other places — that human review of objective footage can and does produce results that contradict what the footage plainly shows.
The question is whether the sport wants to be taken seriously as a regulated competition or whether it would rather keep doing what it has done for a hundred years while wondering why the public is walking away.
Hmmmmmm:
