The H-FARM Business School interviewed 25 professionals across elite sport for their new study, The Decision Paradox, on why organizations that invest millions in data and AI still leave the final decision to a human.

The study includes quotes from Marco Fassone, former CEO of Milan and executive at Juventus, Napoli and Inter; Giorgio Chiellini, Director of Football Strategy at Juventus; Demetrio Albertini, former President of the FIGC Technical Sector; Bologna player Federico Bernardeschi; and, among others, champions and managers such as Maurizia Cacciatori, Stefano Mancinelli, Linton Johnson, Umberto Pelizzari and Romano Battisti.

Excerpt translated from Italian:

Midway through the season, Michele Aragona, match analyst at Ipswich Town in the Premier League, faces a dilemma no model can resolve for him. The data he builds himself, spending hours every week on the dashboards those models come from, points to one zone of the pitch as the most advantageous to attack. Watching the matches, however, Aragona notices that some teams defend that very zone with an effectiveness the models don’t capture, because it depends on how two defenders who have played together for years move together, and that doesn’t show up in the numbers. Together with the manager, he decides to go against the data, and at the end of the season that choice turns out to be the most effective of the year.

Suggesting which exercise to include in Tuesday’s training, a choice that if wrong gets corrected on Wednesday, is not the same as suggesting the purchase of a twenty-million player or the moment to call a driver into the pits as it starts to rain. […] Michele Aragona quantifies it without embarrassment: “there are matches where I’m in complete agreement with the data, others where I’m at 15–20%; over a season, maybe 50%.” The data doesn’t change from one match to the next: the cost of being wrong does.

Federico Bernardeschi provides the mechanism that divergence was missing: “as you have information, I have it too.” Knowing you’ve gone down the same side three times means knowing the opponent will expect it the fourth time, and so you change. Where both sides read the same data, following your own profile amounts to announcing your next move in advance.

The paradox therefore has two faces, and it’s important to keep them together: alibis disappear in both directions. You can no longer hide behind instinct, because the data exists and proves things; and you can’t hide behind the algorithm, because legal and strategic responsibility cannot be delegated.

A human must stay in the loop, especially for the harder the problems.