The Road to Paris 2024: running races and running algorithms
Heymans never grew up with a conventional Olympic dream. In fact, he only started running at college, and decided to start professionally just two years before Paris 2024.
“I only started running at 18,” he recalled. “It came step by step. But two years before graduation [in 2022], I realised I can reach the World Championships and the Olympics. But to do it, I had to commit entirely. No job, no distractions.”
But deciding to qualify for the Olympic Games in just two years is basically like trying to bake a perfect soufflé in a microwave. That’s when Heymans turned to technology for help.
“I wrote an algorithm to optimise everything because I had just two years,” he said. “It wasn’t enough to simply run fast. He needed to run smart, something that came naturally to Heymans, who holds his Master’s in bioengineering and nanotechnology.
According to Heymans, the World Athletics qualification system allowed 42 athletes to qualify for the 5000m event. Half qualify by achieving a very fast time (time standard), while the other half qualify by accumulating points based on performances at different ranked events throughout the season to end up in the top 42. That’s where Heymans saw an opening.
By scraping World Athletics data, plotting trends, and running simulations, he identified a crucial marker: not all races were created equal. Some events awarded more ranking points for a modest time in a high-prestige race than for an elite performance in a lower-tier event.
“I started by going on the World Athletics website, looking at the world rankings. I created a graph to see how the ranking evolved over time,” he said. “Then I wrote a machine learning algorithm that predicted how many ranking points I would need in 2024 to be inside the top 42.”
Heymans calculated that with his base 5000m time of 13 minutes and 15 seconds, he needed around 1215 points to become an Olympian.
This calculation turned out to be shockingly accurate; his predicted score was within a few points of the actual cutoff. From there, the algorithm gave him a blueprint for success.
He asked ChatGPT to determine precisely which races would boost his ranking points most effectively. “I asked: ‘Which races will maximise my chance of hitting 1215 points?’ The AI considered three factors: the race’s prestige, his expected finish position, and achievable times.
“Basically, it’s a mathematical optimisation problem,” the Belgian explained to the “nerds” among us in his detailed YouTube video of the process.
“It told me: run indoors, choose gold and silver-level races, and prioritise events where I could finish in the top three rather than just chasing fast times.”
His plan raised eyebrows. “People told me I had to run 15 races to qualify. I said, ‘No, I need two perfect races.’” When he finally put the algorithm’s plan into action, it worked better than he ever imagined.
“In 2022, I wasn’t even ranked,” he said. “Then, in just three races, I went from nowhere to 31st.”
Ultimately, Heymans didn’t rely solely on rankings; he also secured qualification outright by running a Belgian national indoor record of 13:03.46 at the BU John Thomas Terrier Classic in Boston on 26 January 2024, beating the Olympic entry standard of 13:05.00.