a16z: The Super Bowl Moment for Prediction Markets
TechFlow Selected TechFlow Selected
a16z: The Super Bowl Moment for Prediction Markets
NFL fans watch the Super Bowl while closely monitoring trading activity in prediction markets.
Navigating Web3 tides with focused insightsBy Scott Duke Kominers, a16z
Translated by Saoirse, Foresight News
On February 8 (U.S. time), or February 9 at 7:30 a.m. Beijing time, hundreds of millions of American National Football League (NFL) fans gathered around their screens to watch the Super Bowl—many simultaneously monitoring another screen: the real-time trading activity on prediction markets, where bets span everything from the eventual champion and final score to individual quarterbacks’ passing yardage.
Over the past year, trading volume in U.S. prediction markets reached at least $27.9 billion, covering an astonishing breadth of outcomes—from sports results and economic policy decisions to product launches. Yet the fundamental nature of these markets remains hotly contested: Are they trading venues—or gambling? Are they news tools that aggregate collective wisdom, or scientifically validated instruments? And is their current development path truly optimal?
As an economist who has long studied markets and incentive mechanisms, my answer begins with a simple premise: Prediction markets are, first and foremost, markets—and markets are core tools for allocating resources and aggregating information. The operational logic of prediction markets involves issuing assets tied to specific events: when an event occurs, holders of the associated asset receive a payout. Participants trade based on their personal beliefs about the likelihood of those events, thereby unlocking the market’s core value.
From a market-design perspective, information drawn from prediction markets is far more reliable than listening to a single sports commentator—or even consulting Las Vegas betting odds. Traditional sportsbooks aim not to forecast game outcomes, but rather to “balance betting money” by adjusting odds so as to attract wagers toward whichever side currently has less action. Las Vegas bookmakers want bettors to wager on long shots; prediction markets, by contrast, enable participants to trade based on their genuine beliefs.
Prediction markets also make it easier to extract meaningful signals from vast amounts of information. For instance, if you want to forecast the likelihood of new tariffs being imposed, inferring this from soybean futures prices would be highly indirect—since futures prices reflect many overlapping influences. But posing the question directly in a prediction market yields a much more intuitive answer.
The earliest precursors of prediction markets date back to 16th-century Europe, where people even placed bets on “who the next pope would be.” Modern prediction markets, however, are grounded in contemporary economics, statistics, mechanism design, and computer science. In the 1980s, Charles Plott of Caltech and Shyam Sunder of Yale established the formal academic framework; shortly thereafter, the first modern prediction market—the Iowa Electronic Markets—was launched.
The mechanics of prediction markets are remarkably simple. Take, for example, the proposition: “Will Seattle Seahawks quarterback Sam Darnold throw a pass inside the opponent’s one-yard line?” A corresponding tradable contract is issued: if the event occurs, each contract pays its holder $1. As traders buy and sell the contract, its market price becomes interpretable as the implied probability of the event—i.e., the collective judgment of market participants. For instance, if the contract trades at $0.50, the market implies a 50% chance of the event occurring.
If you believe the event’s true probability exceeds 50%—say, 67%—you can buy the contract. Should the event materialize, your $0.50 purchase yields $1.00, generating a gross profit of $0.67. Your purchase pushes up the contract’s market price—and thus the implied probability—sending a signal to the market that someone believes the current consensus underestimates the event’s likelihood. Conversely, if others think the market overestimates the probability, selling will lower both price and implied probability.
When functioning well, prediction markets offer significant advantages over other forecasting methods. Polls and surveys merely reveal opinion shares; converting them into probabilistic forecasts requires statistical techniques to link sample responses to broader populations—and such surveys yield only static snapshots in time, whereas prediction markets continuously update as new participants enter and new information emerges.
More crucially, prediction markets feature clear incentives: participants are “skin-in-the-game.” They must carefully weigh the information they possess and commit capital—and risk—only in domains where they hold comparative expertise. In prediction markets, individuals can convert their private information and professional knowledge into financial returns—a powerful motivator to deepen domain-specific understanding.
Finally, prediction markets cover vastly more ground than alternative tools. Someone possessing information relevant to oil demand, for example, could profit by going long or short crude oil futures—but in reality, many outcomes we wish to forecast cannot be captured via commodity or equity markets. Consider recent specialized prediction markets designed to synthesize diverse judgments about the timing of solutions to particular mathematical problems—a piece of information critical to scientific progress and a key benchmark for measuring AI advancement.
Despite their clear advantages, prediction markets still face numerous hurdles before realizing their full potential. First, infrastructure-level challenges persist: How do we reliably verify whether an event has occurred—and reach consensus across the market? How do we ensure transparency and auditability in market operations?
Second, market-design challenges remain. For instance, participants with relevant information must actively engage—if everyone is uninformed, market prices convey no useful signal. Conversely, if participants holding diverse pieces of relevant information choose not to trade, market valuations will become biased. The UK Brexit referendum prediction markets stand as a classic counterexample.
Moreover, insiders with absolute private information pose fresh risks. Suppose the Seahawks’ offensive coordinator knows definitively whether Sam Darnold will throw inside the one-yard line—or can even influence the outcome directly. If such individuals participate, market fairness collapses. Potential participants, sensing insider manipulation, may rationally exit—triggering market failure.
Prediction markets also face manipulation risks: tools originally designed to aggregate public judgment may be weaponized to steer public opinion. A political candidate’s campaign team, for instance, might deploy campaign funds to artificially inflate the market-implied probability of victory—creating a false impression of inevitability. Fortunately, prediction markets possess some self-correcting capacity: if a contract’s implied probability strays too far from reason, arbitrageurs will step in to push it back toward equilibrium.
Given these risks, prediction market platforms must prioritize operational transparency—explicitly disclosing rules governing participant management, contract design, and market operations. If these challenges are successfully addressed, prediction markets are poised to play an increasingly vital role in forecasting across domains.
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
a16z
