Back in the day, a word circulated in the world of horse racing called “the hunch,” and it seems to be slowly fading away now, thanks to AI. Artificial intelligence has changed the way we analyze things, especially when you’re handling a lot of data. For background, see our recent coverage.
Horse racing produces an extraordinary amount of information. It is simply impossible for human beings to find patterns and connections between the hundreds of thousands of data points that each race produces. We’re talking about pedigree, training history, racing record, veterinary history, section times, stride length, trainer connections, weather, surface, and the list goes on and on.
Plus, horse racing now has modern sensors that collect even more valuable information, like the horse’s vitals during a race, its speed, acceleration, stride length, movement, post position, etc. Now multiply that by thousands of horses and years of competition, and you have a huge database that you’re scared to even touch, let alone analyze it.
The good news is that AI has a solution. After all, ChatGPT has a solution and agrees with you on everything. But this is much deeper.
This smart tech can now dig through all the data points and find patterns in huge horse racing databases. The result? Well, horse racing should be more predictable.
So, how does this change the way we analyze and predict horse racing? Is it easier to predict the winner nowadays? Let’s find out.
AI Is Learning What a Healthy Stride Looks Like
We’ll start with the biggest change that technology brought to the world of horse racing. Wearable sensors nowadays can record tiny changes in a horse’s movement while it gallops. These algorithms then compare those measurements with previous movement and with data gathered from large populations of other racecourses.
And this technology has been proven to work. The American Association of Equine Practitioners put this idea through a major real-world test last year. More than 700 two-year-old Thoroughbreds participated in a prospective study involving wearable systems from different companies like Alogo, Arioneo, Stable Analytics, and the popular StrideSafe.
So, what did the results say? Well, horses receiving yellow or red warning classifications from the sensors were roughly twice as likely to experience a documented musculoskeletal issue as horses receiving a green classification.
That does not mean an algorithm can diagnose an injured horse. The system still has some flaws, and it cannot predict everything. But it is a start that changed the way owners, trainers, and even bettors analyze horse movement and their form.
Even Keeneland Has Put Generative AI Into Race Day
The change has already reached one of America’s most traditional racecourses. Yes, not many people know, but AI is already in some of the biggest horse racing events in the United States.
In April 2025, Keeneland introduced generative AI features into its Race Day app, describing itself as the first racecourse to offer AI-driven insights within a track app. The system generated summaries using program information and past-performance data to help users understand runners and races.
This makes the sport even more fun for the spectators. This year, the Breeders’ Cup is coming to Keeneland, and maybe they will include their AI tech once again. People will be able to get all the AI-driven insights instantaneously on the track, then use them as an advantage when looking through the 2026 Breeders’ Cup odds.
That may sound minor compared with predicting injury, but it addresses one of racing’s oldest problems.
The sport contains a ridiculous amount of information.
Experienced followers understand terms involving form, sectionals, surfaces, class levels, and pedigrees almost automatically.
Thousands of Races Can Teach an Algorithm Things One Veterinarian Cannot See
The power comes from scale.
A 2025 study published in the Journal of the American Veterinary Medical Association analyzed sensor data covering 28,481 races by 11,834 Thoroughbreds between 2021 and 2024. Horses were assigned movement-based risk scores from one to six using accelerometer-derived data and an algorithm.
Only 0.4 percent of starts involved horses receiving the highest risk score of six, yet that group accounted for 4 percent of the fatal musculoskeletal injuries studied. The researchers calculated that horses receiving the highest score had a substantially greater probability of fatal injury than those receiving the lowest score.
That is a useful demonstration of what machine learning can do particularly well.
AI Can Analyze What Happens Inside the Race Too
For decades, racing analysis relied heavily on finishing positions, margins, overall times, and the observations of experienced race readers.
GPS changed that by allowing individual horses to be measured throughout a race.
AI and computer vision can make those datasets richer.
Equibase, the official database for North American Thoroughbred racing has worked with Total Performance Data on GPS information, including stride frequency and stride length. Equibase has also experimented with machine learning and optical tracking that recognizes individual horses through elements such as jockey silks, helmets, saddlecloth colors, and numbers.
The goal is to improve the accuracy of positional information and create richer visual representations of what happened during a race.
Machine Learning Can Look for Problems Across Several Races
AI also becomes powerful when it examines changes over time rather than one isolated performance.
University of Melbourne researchers found that declining speed and stride length across successive starts were associated with musculoskeletal injury. In their earlier work, more pronounced changes began appearing approximately six races before an injury event.
Researchers subsequently tested machine-learning models using race history and stride characteristics to see whether injury, enforced rest, or retirement could be predicted.
So, this made analysis easier and somewhat made the sport more predictable. But this doesn’t mean that AI can pick the winner of an upcoming race with 100% accuracy all the time. The AI is in charge of coming up with a mathematical probability, and decisions still come back to the humans.
