A group of three Italian physics students from the University of Salento claim they won over €50,000 in the lottery thanks to an AI programme they developed. A claim that goes against all mathematical logic and the principle of randomness.
The urge to predict future draws is nothing new. People often study past lotto numbers before making their choices, and many websites analyse the history of previous spins for online roulette. If the number 9 hasn’t appeared in 21 days, it must be bound to happen any second now, right?
No: every draw is an independent event, meaning past results provide no information about future outcomes. Each number has exactly the same probability of being drawn at every turn. No combination is favoured unless the machine is defective—something very rare and heavily regulated. To put it simply: no draw carries a memory of previous results.
The gambler’s fallacy
Yet believing otherwise is a common misconception. So common, in fact, that it has a name: the gambler’s fallacy, also known as the Monte Carlo fallacy, referencing the famous incident in 1913 when a roulette wheel spun black 26 times in a row. The gambler’s fallacy is the belief that if an event has occurred less frequently than expected in the short term, it is somehow “due” to occur soon to balance things out.
AI’s fallacy?
These mathematical facts did not discourage the group of physics students, who decided to feed historical probabilities into an AI in hopes it could predict winning numbers. But instead of instructing the programme to play numbers that hadn’t been drawn in a long time, they told it to do the opposite. This decision was based on advice from a betting shop owner, Diego Manca.
“The three of them,” explains Manca, “came to me just under a month ago asking for information about ‘late’ numbers. They wanted to know whether, in my experience, it was better to bet on numbers that hadn’t been drawn for a long time. I suggested that, based on my experience, it might be more profitable to bet on numbers that are drawn most frequently on a given wheel. They took my advice into account and developed a mathematical system based on it.”
“It corrected our aim,” the three students explained, “leading us to focus on the frequent numbers, contrary to the usual strategy with late numbers.”
50.000 euros of winnings
After testing the programme for the first time and winning €4,500, they used it for a larger operation, winning another €43,000. Of course, these wins could very well be pure coincidence.
“We know very well that the lottery is a random system,” the students told La Repubblica. “But we were curious to see if, thanks to the AI system, we could identify trends that might be useful. We saw an opportunity to turn this curiosity into a small experimental project.”
