From Thousands of Lots to a Better Shortlist

Buying a racehorse is not easy. At a major sale, the buyer is not choosing from ten horses, but from hundreds or even thousands. The problem is not that there is no information. The problem is that there is too much information, and only a small percentage of horses will later become high-level performers.

That is exactly the type of problem ChampionBred was built to help with.

Our idea is simple: use pedigree data, racing results and AI models to reduce a large catalogue into a smaller group of horses with better chances. Not guaranteed winners. Not magic. Just a better probability.

We have now tested this idea by doing something very straightforward: selecting horses first, and then waiting to see what happened after they raced. This is important because it avoids the easiest mistake in bloodstock analysis: explaining success after it has already happened.

When selections are made before the results are known, the question becomes much cleaner: did the shortlist produce more good horses than expected?

Across several auction examples, the answer has been yes. ChampionBred selections have repeatedly shown a higher concentration of high-performing horses than typical auction averages. In one retrospective Inglis Easter test, 6 of 25 selected horses reached Timeform ratings above 100, or 24%. If one related case is included through a full sibling that later became a stakes winner, the figure rises to 28%.

That does not mean every selected horse worked. Of course not. Racing depends on many things beyond pedigree: conformation, soundness, training, temperament, opportunity and luck. A model cannot control those factors, and no serious tool should pretend otherwise.

But the purpose of the tool is not to guarantee the final result. The purpose is to improve the starting point.

If a normal catalogue gives you a low chance of finding a stakes-level horse, and a filtered shortlist gives you a higher chance, that difference has real value. It means buyers can spend more time, inspection effort and professional judgement on the horses that already pass a statistical screen.

This is how we believe AI should be used in bloodstock: not as a replacement for agents, trainers, vets or experienced horsemen, but as a way to focus their expertise where it matters most.

Pedigree is not destiny. But pedigree contains signals. Our job is to find those signals more consistently, test them honestly, and use them to improve the odds.

No crystal ball. No magic. Just higher chances.