AI enters the paddock as Pythia Sports brings data to breeze-up horse sales
A new wave of technology is entering the traditional world of horse racing auctions. Pythia Sports has launched a proprietary data model that applies artificial intelligence and biomechanics to breeze-up sales, offering buyers deeper insights into a horse’s performance potential before the gavel falls.
Bringing structure to instinct-led sales
Breeze-up sales allow buyers to watch two-year-old horses gallop before bidding. These short sprints – known as “breezes” – have traditionally been judged by eye. Pythia aims to shift this process with technology that evaluates over 100 performance factors, including stride, speed, and biomechanics.
While the horse racing industry has historically been slower to embrace data-driven tools compared to other sports, Stephen Davison, Head of Commercial Operations at Pythia sees a shift underway.
“I think there’s definitely starting to be a lot more openness to people using data to inform their decision making. You only have to look at where football was 10-15 years ago to where it is today. The obvious examples are Brighton and Brentford who were using their own player data models to find value in the transfer market way before any other top level club. It is only natural that you can get a cross over into other sports and other industries. We’re hoping we can play our part in revolutionising the way horse racing and bloodstock views the use of data within the sport.”
How the model works
According to Stephen Davison, the model breaks down each breeze into three core sections: time, speed, and stride. Every stride is tracked as the horse accelerates, giving a full picture of how the animal moves, adapts, and reaches top speed.
“We have biomechanics as well, that looks at the horse’s walk and takes measurements of a horse’s conformation and its gait. That works in conjunction with historic racing performances to learn what the most defining physical features are in an elite horse.”
The AI model is trained using historical breeze-up sales data going back ten years, allowing it to identify subtle patterns and correlations that might be overlooked by the human eye. Davison added:
“For example we can go back ten years to look up breeze-up data to say, ‘this horse did this breeze, this stride and these biomechanics’ and then the machine learning model can start picking up stuff we’ve missed or things we’ve maybe underestimated or overestimated.”
The system generates a single rating for each horse. Buyers can use this score to compare animals with greater confidence and accuracy. Pythia debuted the system at the Newmarket breeze-up sale, where it attracted interest from trainers, bloodstock agents, and buyers.