Prediction Market Surveillance: A View from the Trenches

It feels like prediction markets are everywhere these days. As prediction markets expand and trading volumes continue to rise, the industry’s long-term success depends on both established players and new entrants adopting rigorous market surveillance systems. And, over the course of the last year, it has become clear that although many principles from equity-market surveillance are applicable, prediction markets have several distinctive features that must be accounted for to ensure any surveillance system operates effectively.

Equity Market Surveillance

Market surveillance in equities has developed over the past 50 years into a sophisticated and well-established discipline. Today, the core principles of equity market surveillance are embodied in systems such as FINRA’s SONAR and the SEC’s ATLAS surveillance platform. These tools are designed to detect potentially illicit trading ahead of market-moving corporate events. For example, when a stock experiences a sudden and significant price movement, such as a 40% move following the announcement of FDA trial results, these systems can identify individuals who traded shortly before the event and assess whether their pre-event activity differs meaningfully from their historical trading patterns.

Consistent with the SEC’s focus on trading activity ahead of market-moving events, a review of SEC insider-trading cases over the past two decades suggests the agency has a keen ability to identify individuals who trade unusually large amounts of out-of-the-money options shortly before significant news is released. This is the classic “YOLO” trade (you-only-live-once): max out your line-of-credit, liquidate your retirement savings, take out a car loan, and bet it all on a specific ticker just before material non-public information (“MNPI”) becomes public. On its face, this type of conduct appears relatively straightforward to detect. Prediction markets, however, are not equity markets, and their differences create important challenges for market surveillance.

Differences in How Individuals May Seek to Monetize MNPI

Stock prices are influenced by many factors, including Federal Reserve interest-rate decisions, geopolitical developments, earnings announcements, forward guidance, executive turnover, and other market-moving events. Rarely does anyone possess material non-public information about the stock price itself; no one can know with certainty where a stock will trade at the end of the day. Instead, individuals may possess MNPI about one or more factors that influence––but do not solely determine––the stock price.

As a result, individuals seeking to monetize MNPI in equity markets can only do so indirectly. A rational trader generally waits to trade until shortly before the MNPI becomes public because, until then, the trader bears the risk that the stock price may drop for reasons unrelated to the MNPI. Trade too early, and the price may move against the trader because the CEO resigns, the Federal Reserve changes interest rates, or geopolitical events disrupt the market. The SEC’s “hack-to-trade” cases illustrate this point: even when someone hacks a newswire service and obtains an earnings release before the public, trading on that information does not always produce a profitable result.

This equity market dynamic is an important difference from prediction markets. Many prediction market contracts are resolved by reference to a specific corporate performance metrics (e.g., Sweetgreen’s quarterly profit margin, DoorDash’s quarterly delivery volume, or SpaceX’s monthly launch count) or to discrete political events (e.g., whether Pam Bondi or Tulsi Gabbard will step down). In these markets, MNPI can involve knowledge of the market outcome itself, namely whether the contract will resolve to $0 or $1.

An insider who knows the relevant outcome before the public has a direct means of monetizing their MNPI. If the insider knows a company’s gross margin, for example, they can simply buy a contract that resolves based on that metric. This eliminates the intervening factors that would otherwise affect stock prices and pose a risk to the insider. In this setting, the economically rational strategy will often be to trade immediately after learning what the contract outcome will be. If the insider delays trading, the principal risk is that information leakage or shifts in public sentiment will erode their informational advantage.

This distinction has important implications for prediction market surveillance. In some markets, MNPI regarding the resolution of an event only becomes available shortly before the publication of such information takes place. For example, the outcome of an election or the Federal Reserve’s interest rate decision might be known to insiders only shortly before the information becomes public. Traditional surveillance techniques design to detect anomalous trading shortly before price movements are likely to capture insider activity in these types of markets. However, monitoring for front-running shortly before market-moving events in other types of markets may not always be effective for detecting insider trading in prediction markets. Instead, surveillance may need to place greater emphasis on ex post analysis after a market has resolved.

Differences in the Concept of Materiality

One of the critical elements of market surveillance is determining what magnitude of price movement should be considered “economically material.” In equity markets, a sudden 40% increase in the stock price of a liquid mid- or large-cap company would generally be viewed as material. Prediction markets, however, present a more complicated case because contracts trade between $0 and $1. If a contract is trading at $0.05, is a 40% increase to $0.07 material? If it is trading at $0.50, is a 40% increase to $0.70 material? And once the price reaches $0.80, a 40% increase is no longer even possible.

The binary structure of prediction market contracts therefore complicates traditional concepts of economic materiality, which are often based on percentage changes in price. Surveillance thresholds anchored solely to fixed price changes, whether measured in percentage terms or cents, are insufficient. Instead, assessing materiality may require quantifying the benefit that specific trades provided to the trader.

Differences in Interpreting Consistent Winners

In addition to monitoring trades placed shortly before material events, the SEC also evaluates what is commonly referred to as a trader’s “win rate,” or the percentage of trades on which an individual earns a profit. The intuition behind this metric is similar to that used in casino games: if someone plays roulette 100 times and wins 90% of the time, the result appears suspicious. Win-rate analysis is designed to detect a different form of insider trading than the one-off YOLO trade referenced earlier. Rather than identifying a trader who places a single outsized bet, win-rate surveillance is aimed at detecting traders who repeatedly profit from MNPI, for example, law firm employees trading across multiple client merger announcements. The SEC’s Artemis surveillance platform is designed to identify these types of “serial cheaters.”

In equity markets it is commonly thought that the benchmark win rate, for a strategy without MNPI, should be below 70% (generously). But in prediction markets, the price of the contract signifies the probability of the particular outcome. For example, if the contract is priced at $0.90, the market believes the outcome will occur with a 90% probability. If an individual buys contracts across 10 different events, each priced between $0.90 and $0.99, the individual’s expected win rate exceeds 90%. Indeed, a well-known trading strategy in prediction markets is called “bonding.” Bonding entails buying for $0.98 a contract that will expire at $1, and holding until expiration. Here the individual locks up their capital until expiration, but clips $0.02 or about 2%. Someone engaged in a bonding strategy will have very high win rates across a large number of markets. Hence, looking only at simple win rates can be misleading.

Conclusion

As prediction markets expand and trading volumes reach record levels, rigorous insider-trading surveillance is not optional; it is essential to the credibility, integrity, and long-term viability of the industry. Platforms that fail to build surveillance systems tailored to prediction markets risk importing equity-market concepts in ways that are incomplete, ineffective, or misleading. Effective surveillance must account for the distinctive structure of these markets by looking beyond trades placed shortly before large price moves, conducting ex post reviews of unusually profitable positions after contracts settle, applying flexible measures of materiality tied to trader-level gains and losses, and adjusting win-rate analysis to reflect the probabilities implied by contract prices. If prediction markets are to grow into trusted, durable, and well-functioning markets, incumbents and new entrants alike must treat the detection and deterrence of insider trading as a core market-design obligation.