Surveillance and Market Integrity
The surveillance challenges facing prediction market exchanges and intermediaries are rapidly becoming more complex. Unlike traditional markets – where decades of infrastructure exist around insider trading, restricted persons and transaction monitoring – event-driven markets often involve broader participant networks, less standardized data and new forms of information asymmetry. The question is no longer simply whether misconduct can occur, but how quickly firms can identify it, investigate it and act before it undermines market integrity.
At FORWARD, that theme carried into a surveillance-focused panel, featuring:
- Eric Haynes, Head of Surveillance, Robinhood
- Scott Sadin, Co-Founder & Co-CEO, IC360
- Jeff Bell, moderator, Chief Operating Officer, Eventus
Through a series of real-world examples – from sports betting and restricted participant lists to event contracts tied to weather, entertainment, corporate outcomes and bank failures – the panel examined how surveillance programs need to evolve as prediction markets scale.
Surveillance Starts with the Contract
Bell opened by positioning surveillance between product design and enforcement. Product design determines how a contract is structured and who may be restricted from trading it. Enforcement addresses what happens when misconduct occurs. Surveillance sits in the middle, helping firms identify suspicious behavior, reconstruct activity and decide when to escalate.
Prediction markets cannot be surveilled as a single, uniform category.
“Think about the classification of the type of event contract,” Bell said. “It’s not a one-size-fits-all market. Everything needs to be considered in context. The data is hugely important and does feel to me like a gap.”
The diversity of these contracts creates distinct surveillance challenges. A weather contract may involve people with better information, such as those connected to the National Weather Service, but they are not creating hurricanes. Sports contracts are different: a pitcher, catcher, coach, manager or someone connected to the locker room may influence or know something about the outcome of the next pitch. A corporate contract may involve familiar insider trading concerns. A bank failure contract raises another risk: if the market moves, that signal could potentially influence the behavior it is measuring.
Restricted-Person Controls Need to Scale
The panel then turned to one of the most difficult operational questions in event-driven markets: identifying who should be restricted from trading.
Haynes explained that Robinhood relies heavily on contract rules to identify prohibited participants, then works through occupations, employers and other indicators that may signal whether a customer has access to information or influence over an outcome.
“We identify who would be restricted, and then we brainstorm on occupations and employers that might fit those restrictions,” Haynes said. “So it’s a very manual process that we’re employing right now.”
As event contracts multiply, that manual process becomes harder to sustain. Haynes said Robinhood is exploring AI to help identify connections between customers and underlying event contracts.
Sadin added that sports markets show how broad the insider universe can become. In collegiate sports, high-profile athletes may have large networks of friends and family around them. A player may not have malicious intent when sharing that they are not playing, but that information can become valuable once it enters the market.
“You do not have to have any malicious intent to say, ‘I’m not playing on Friday or Saturday,’” Sadin said. “But that information is now extremely valuable, and that edge is often then reflected in the market.”
That is why restricted-person controls are not only a policy issue, but an infrastructure issue as well.
Detection Depends on Data, Behavior and Context
Surveillance requires more than volume and price monitoring. Sadin emphasized that anomaly detection in sports and event-driven markets starts with “data in” – not just market activity, but transaction-level detail, participant profiles and information about the people connected to the event itself.
“Data in, in the sports world, means it can’t just be volume that you’re looking at as a potential signal for anomalies or abnormalities,” Sadin said. “When you open up and look at the number of participants that could impact the outcome of an event contract with respect to a sports match, it’s quite vast.”
That may include players, officials, referees, judges, umpires and others. Surveillance teams need behavioral baselines, peer comparisons and indicators that may reveal manipulation or misuse of information.
Haynes said suspicious activity is often less about one successful trade than about size, frequency and deviation from normal behavior.
“If you have a customer and it’s the same pitcher game after game, and that individual customer has a 95% win rate on those contracts, well, then I’m interested,” Haynes said. “Or I’m interested in the one who out of nowhere puts down $20,000 on the next pitch and gets it right.”
In equities, teams may begin with activity and work backward to the individual. In event contracts, identity, relationships and proximity to the event may need to be part of the review from the start.
“Normal” Is a Moving Target
Traditional markets offer mature reference points: average daily volume, earnings windows, historical trading patterns and established market structure. Event contracts may involve a 15-minute market tied to a single event.
Bell said surveillance will need to evaluate behavior on multiple axes.
“I think it’s going to ultimately come down to not just trader profiling or account profiling, but also contract profiling and thinking about what’s normal on both axes,” he said.
Sadin summarized the challenge directly: “Normal is never normal. It’s always changing.” Products change, participants enter and exit, market structures evolve and bad actors adapt.
Reconstruction and Documentation Matter
Bell also highlighted a major infrastructure gap between traditional financial markets and event contracts. In traditional markets, regulatory reports and databases like EDGAR help firms reconstruct what happened. Event contracts do not yet have the same standardized information architecture.
That makes it all the more critical for firms to create a defensible record of their decisions. Bell stressed that firms should document their process, even where responsibilities are still evolving. If a market attracts attention or unusual activity, compliance teams should understand what happened on their platform, review the activity and memorialize the steps they took.
Ultimately, the panelists agreed on the urgent need for prediction market surveillance infrastructure capable of evolving alongside the market itself.