Insights · Method

How Data Spotted a Triple Crown Winner That Horse Racing Experts Passed On

A data analyst found a Triple Crown winner by ruling out everything else. Our own contact form test showed why the audience gets the last word.

By Bartholemu & Kiran Gopinath, Certified Executive Coach · November 2026 · 9 minute read

Experienced people trust experience, and most of the time that works. The people who buy racehorses are some of the best judges of animals on earth. For generations they paid millions for famous bloodlines, and one of them nearly let a Triple Crown winner walk out of a sale ring for $300,000.

What he needed was not a sharper gut. It was a habit of being wrong cheaply, on purpose, and keeping score.

What did one analyst see in Hip 85 that the room missed?

In August 2013, a bay colt entered the Fasig Tipton yearling sale at Saratoga as Hip 85. He was bred and owned by Ahmed Zayat. Zayat's adviser, Jeff Seder, who holds a Harvard MBA and law degree and runs a horse selection firm called EQB, had told him to keep the colt. Bidding stalled and the colt missed his reserve, so Zayat bought him back through an agent at the posted minimum of $300,000. Auction rules would not let him simply pull the horse (Squash Magazine, Have Heart, Wikipedia).

On the usual measures, the colt was ordinary. Seder's measurements said otherwise.

Those figures come from a Squash Magazine profile of Seder. Two years later, that colt, now named American Pharoah, won the Kentucky Derby, the Preakness and the Belmont, the first Triple Crown in 37 years.

Why trust a heart? Because he had thrown out everything else.

Seder's whole approach was to let data drive the insight instead of tradition. He did not ask what buyers believed made a champion. He asked what could be measured, and he let the numbers decide. He emphasizes physical data over race results, and argues that rivals who lean on race times alone are cutting corners (Inverse).

So he did not start with the heart. He started where everyone starts, with pedigree, and found it related to racing success only weakly. Then he went looking for something better, and most of what he tried failed. Nostril size told him nothing. Leg muscle size told him nothing. Manure size told him nothing. Fast times in short workouts rewarded sprinters, even though the big races run three quarters of a mile or longer. His own conference slides add stride length, a so called health gene, herd dynamics, cow bones, horseshoes and a "magic feed" to the pile of ideas that did not hold up (Squash Magazine, Have Heart, Inverse, Seder's HorseTech slides).

Each failure did real work. It removed a competing explanation, so by the time the left ventricle stood out it had survived against everything that did not. The existing data could not answer his question, so he built a portable ultrasound and created his own dataset, scanning tens of thousands of animals over decades.

What turns a hunch into data science?

Four habits, and they apply well beyond horses. These are the principles from his published slides and interviews. His actual model is proprietary, so I cannot tell you more than the sources do.

1. Compare fairly. A big horse has a big heart. He ranked each heart as a percentile within horses of the same sex, height and weight, so the measure rewarded the heart and not the size.
2. Measure cleanly. His slides stress that technicians needed years of training, because a small error in probe angle produces a large error in the data.
3. Use enough data. The slides describe 5,431 yearlings and 2,003 two year olds, more than 7,400 horses, so that luck stops being a good explanation.
4. Distrust your best result. When Seder saw a huge left ventricle at Saratoga, he knew some horses have enlarged hearts from disease. The finding had to survive his own suspicion first.

I did a small version of this with a contact form

Twenty plus years of experience told me a longer contact form means fewer submissions. The industry agrees: CXL's review of form tests calls cutting fields a best practice and reports one test where going from nine fields to six dropped conversions 14%.

Because I believe in testing and learning, we created an A/B test. Half of the visitors saw a short form with only the basics: name, email, zip code and a comments box that was not required. The other half saw a longer form with five additional fields, including drop downs specific to the client, their services and what customers would be contacting them about.

Visitors who saw the long form submitted it 60% more often than visitors who saw the short form. Of those people, 33% followed through on the lead and converted.

The reason, in our read, was the audience. Our target segment was older adults, and for them a longer form signaled that the organization was truly interested in them. That is an explanation the test suggests but did not prove.

What does this change about how you use data?

It changes who gets the last word. Seder compared each heart to its own size group because the overall average hid the answer. The same idea applies to your customers. Regardless of what the data says holistically, keep your target audience in mind, because a result that holds for everyone can be wrong for the people you serve. I made the same argument about product design in the OXO peeler post: design for your hardest customer, not the average one.

It is also the reason to test every time, even when you feel sure. Nobody would have guessed that two free punches on a car wash loyalty card would change much, since customers still had to buy the same eight washes. In a study of about 300 cards over nine months, 34% of customers with the head start card earned the free wash, against 19% with a blank card. That study was not ours, but you can read all about it in the car wash loyalty post. The lesson is the same as the horse and the form: your gut can only suggest, and the test is what tells you.

Experience tells you where to look. Data tells you whether you were right. Your audience tells you how to read it.