Prediction Markets Put a New Test on Who—and What—to Trust

Imagine the following scenario: 9:07 a.m. on Super Bowl Sunday, 2026. Emma Rodriguez set her coffee mug down with more force than she intended. “Can you run that again?” she asked her analyst Josh Lin. Rodriguez, vice president of strategy at a media analytics firm, had already locked in her prediction for Super Bowl LX, giving the Seattle Seahawks a 57% chance of winning against the New England Patriots. That percentage was an important part of the forecasts her company used to advise advertisers, broadcasters and sports sponsors on advertising buys, considering factors such as audience retention and second-screen engagement.

Across the table, Lin refreshed his screen. “It’s real,” he said. “Seahawks just jumped from 56 to 61 on Kalshi. And Polymarket was already at 63.” That was a four-to-six-percentage-point difference from Rodriguez’s internal model, which had been carefully calculated based on historical performance, matchup efficiency and in-game momentum simulations. Now, the market was telling a different story, and a client was already texting about reallocating resources before kickoff. Rodriguez scowled. “Those are betting platforms!” she said.

Lin replied, "Some might say they are prediction markets.”

That distinction, and the managerial dilemma it creates, is at the center of “Belief at Scale: Prediction Markets & Managerial Decision-Making,” a new case co-authored by four University of Virginia Darden School of Business scholars: Dean Yael Grushka-Cockayne, professor Anthony Palomba and Asa Palley, and postdoctoral researcher Junnan Wang.

The case arrives at a moment when prediction markets have shot up in popularity. Kalshi was valued at $11 billion after a $1 billion Series E round in December 2025. Polymarket, backed by Intercontinental Exchange with $2 billion committed, recorded more than $3.7 billion in monthly trading volume in the same period. CNN struck an exclusive data partnership with Kalshi, while Major League Baseball named Polymarket its official prediction platform.

“If you talk to proponents of these markets, they will say, ‘This isn’t just gambling, this is trading on information,’” says Asa Palley, associate professor of business administration in the Data Analytics & Decision Sciences (DADS) area at Darden. “It’s an efficient mechanism for distilling the right probability for an uncertain event, similar to the way trading activity in the stock market identifies what a company is worth.”

The case, focusing on a media analytics firm’s game-time dilemma, asks students to wrestle with how accurate these prediction platforms are, and how companies should take them into account in determining their forecasts. The case also provides an opportunity for students to learn more about how these prediction markets operate and the nature of the implied probabilities they generate.

A Decision-Making Conundrum

Prediction platforms such as Kalshi and Polymarket go way beyond sports contests, allowing users to trade contracts on a variety of political and cultural events—including election outcomes, court decisions, tech product releases, economic indicators, award show winners and music chart ratings. A contract on one of the sites pays $1 if an event occurs, and nothing if it doesn’t. As users buy and sell those contracts, the price fluctuates to indicate a percentage chance that outcome will occur.

Prediction platforms aren’t exactly new. Starting in 1988, a trio of University of Iowa economists launched the Iowa Electronics Market, allowing participants to predict the outcome of U.S. elections, with bets capped at $500. Its track record was impressive: in five presidential elections from 1988 to 2004, the Iowa Electronics Market outperformed traditional polls roughly 74% of the time. In subsequent years, other sites such as InTrade and PredictIt followed.

Kalshi, however, was the first exchange to be licensed by the Commodity Futures Trading Commission (CFTC) as a federally regulated predictions market in 2025, and Polymarket followed soon after. Bolstered by the CFTC’s official approval and the widespread adoption of online sports betting platforms such as DraftKings, prediction markets have experienced a surge in popularity.

“Why is there so much money appearing in these markets? Certainly, one argument would be that they are fun,” Palley says. “People can engage in them like a hobby. On the other hand, they have been criticized as potentially addictive.”

Recently, the sites have been heavily criticized for allowing bets on controversial topics such as the war in Iran and are suspected of being manipulated by users who place big bets that can shift odds by leveraging insider political information. There’s no denying the impact that the sites have had. News outlets cite them as a legitimate way to take the pulse of public opinion. Yet for media analytics firms and other organizations, prediction markets present a conundrum.

“Firms are making dynamic decisions based on quantitative models that generate a probability and advise their clients on strategy, and now this market shows up,” Palley says. “How much faith should they put in that? How should they integrate it into their decision-making process?”

Essentially, companies have three options: 1) stick with their internal predictive models, which might be seen as more transparent, reproducible and defensible in client meetings, but are slower to update and unable to incorporate real-time market signals; 2) defer to the market, which is efficient in incorporating many opinions but also volatile, unauditable and lacking proprietary characteristics; or 3) blend the two approaches.

“The case frames the protagonists’ situation not as a prediction-accuracy problem, but as a decision-architecture problem,” says Anthony Palomba, assistant professor of business administration in the DADS area. “If rival organizations are integrating Kalshi or Polymarket signals into real-time decisions and you are not, are you at a systematic informational disadvantage? And if prediction markets are already embedded in CNN's on-air programming, in MLB's official data infrastructure, and in Dow Jones's digital properties, what does it mean for firms that treat these platforms as noise?”

The case guides students through those issues as they consider how heavily to weigh internal predictions against prediction markets in order to generate the best possible forecast.

 “The question is not about math,” says Palomba. “It is about what a firm believes concerning the relative quality of two different information sources, and what it is willing to defend in a client meeting.”

As part of the exercise, the case draws on various theoretical traditions, such as the “wisdom of the crowds,” which holds that large groups of amateurs can be more effective at making predictions that a small number of elites, provided certain conditions are met. It also examines the efficient market hypothesis and its behavioral critiques, as well as the literature around forecasting research.

Each offers a different lens on the central question of when a firm should trust a market over its own model. Specific events around a prediction can also be a factor in assigning weights, according to Palley.

“If the starting quarterback sprains his ankle just before the game, that would have a huge effect on the outcome, but the analytics model is not going to see that,” he says. “If you see a prediction market suddenly jump from 30% to 70%, it probably means something significant happened, and so you might want to give the momentum shift more weight.”

In the situation depicted in the case, the protagonist and her team chose a 50/50 blend of market and internal model to create their forecast before the kickoff. But the actual numbers students ultimately reach are less significant than their reasoning behind defending their choice.

“It’s not necessarily that there is a right answer, and are you going to get there or not?” Palley says. “There are multiple perspectives that could be justified, and the most important thing is that you have a solid framework that you can use to support and explain your course of action.”

In other words, both internal prediction models and market forecasts could have their place in the final outcome, provided that their inclusion is reasonable. “The question for firms is not which one to trust,” says Palomba. “It is how to combine them deliberately rather than defaulting to one and hoping for luck.”

This article is adapted from the case “Belief at Scale: Prediction Markets & Managerial Decision-Making,” co-authored by professors Yael Grushka-Cockayne, Anthony Palomba and Asa Palley, with postdoctoral researcher Junnan Wang, and published by Darden Business Publishing (July 2006)