On 24 August 2026, Stuart Crowley, a commercial director who writes about prediction markets, published the first instalment of his interview series “Prediction People”. His subject was Émile Servan-Schreiber, who has built and studied prediction markets since the early 2000s, first with NewsFutures and now with Hypermind. Servan-Schreiber shared the interview the same day as a “deep dive” into the real mission and lasting promise of prediction markets.
Crowley notes that Servan-Schreiber came to the field through psychology rather than trading. He is less interested in the wager than in the people: why they take part, why some become consistently good, and what makes a crowd smarter than its members.
A leaderboard from 2002
The detail Crowley builds the piece around is a small one. Servan-Schreiber told him he had recently opened a NewsFutures leaderboard from 2002 and recognised some of the names. Nearly 25 years later, the same people are still forecasting on Hypermind, with no jackpot waiting for them.
Whatever their motivations are, it’s not the money. More likely a sense of community and a pursuit of excellence.
Émile Servan-Schreiber, in “Prediction People”
Crowley expected real money to sharpen people’s judgement. Servan-Schreiber pointed instead to a long record of virtual-money forecasting: NewsFutures in sport, the Good Judgment Project in geopolitics, and Hypermind in public health and policy. He also cited Hypermind’s analysis of 1,141 markets over 12 years, which found its play-money forecasts comparable with the large real-money platforms on accuracy and calibration. Calibration means that events given a 70% chance happen roughly seven times in ten.
Mechanism and business model
So what does the money add? Servan-Schreiber’s answer was:
The only thing that real-money wagers adds to the equation - but it’s a very big thing - is an obvious business model.
Émile Servan-Schreiber, in “Prediction People”
Crowley admits he had treated the forecasting mechanism and the commercial model as one thing. A real-money venue earns from trading activity. A play-money forecasting business has a harder job, because someone outside the crowd has to value the answer enough to pay for it.
That distinction does not make Servan-Schreiber dismissive of today’s platforms. According to Crowley, he credits Kalshi and its co-founder Tarek Mansour for the long regulatory work that established a legal real-money market in the United States. His concern is the story the industry tells about itself. He worries that the current wave of entrepreneurs is losing sight of the original promise of prediction markets, which he described as “mobilising collective intelligence on consequential problems” rather than sports betting.
Building his own competitor
On AI, Servan-Schreiber told Crowley that well-designed AI forecasters have caught up with prediction markets on public benchmarks. He pointed to recent research with Younes Jeddi and José Segovia-Martín, his co-authors at UM6P’s School of Collective Intelligence, on how machine forecasts increasingly resemble human crowd forecasts. Hypermind’s response was to spin out The Forecasting Machine. Crowley describes this as a founder deliberately cannibalising his own business: if AI can do part of the job better, build around it rather than protect the old product.
People still have a role, in Crowley’s account of the conversation: deciding what is worth asking, fixing a badly framed question, adding information a model cannot see, and challenging answers that look tidy but do not hold up.
Governments, and the limits of tradability
Asked for an underrated use case, Servan-Schreiber chose government. He pointed to Glimt, a Swedish public forecasting platform on the war in Ukraine. No one there plays for cash. The value, as Crowley describes it, comes from asking people to turn an opinion into a probability and then live with the result as events unfold.
I think it has the potential to make our democracies much smarter. They need it!
Émile Servan-Schreiber, on governments using prediction markets
He is not keen on making every event tradable. Crowley’s example is trading on whether a CEO will resign while sitting near the people deciding it. Servan-Schreiber’s warning, as Crowley puts it, is against taking the “truth machine” label too literally. Society still has to decide which information advantages it is willing to let people turn into cash.
