Eighty-nine wallets captured seventy percent of the profits on the largest geopolitical prediction market in history. A class action lawsuit was filed today. Polymarket deleted its nuclear market. The mechanism works. The question is what it works for.
The moral question was whether prediction markets should profit from death. The political question was whether banning insiders would destroy the accuracy that justifies the market's existence. This week, the data arrived.
The Courtroom
A class action lawsuit filed today in the U.S. District Court for the Central District of California calls Kalshi's death carveout "deceptive" and "predatory." The complaint argues that the language of Kalshi's market — "Will Ali Khamenei be out as Supreme Leader?" — was "clear, unambiguous and binary." With an American naval armada assembled off Iran's coast, the complaint adds, consumers understood that death was "the most likely — and in many cases the only realistic — mechanism by which an 85-year-old autocratic leader would leave office."
Kalshi's CEO said the settlement "followed its published rules" and that altering it retroactively would "undermine trust in the exchange." The company reimbursed all trading fees and covered net losses. But it did not pay the fifty-four million dollars the market generated.
The moral tension has been covered. What moved this week is the venue. Whether "out as Supreme Leader" includes death is no longer a debate on social media or in Senate letters. It is a question for a judge. The answer will set the first legal precedent for how prediction market resolution rules are interpreted when they conflict with the plain meaning of the question the market asked.
The Ledger
Bloomberg published the data on who actually made money on the Khamenei market. The empirical question — the one neither the moral debate nor the legislation addressed — has a precise answer.
Eighty-nine wallets captured seventy percent of the profits. That is 0.4 percent of total bettors. Ninety percent of wallets on Polymarket's Khamenei market held positions of one thousand dollars or less. The ten accounts that traded at least one million dollars entered only after news reports broke and odds were already above ninety percent, earning an average profit of $11,223 per wallet — a return of 0.34 percent. The majority of accounts that placed bets lost money.
Six newly created accounts, identified by analytics firm Bubblemaps, collectively netted $1.2 million. The accounts were created in February. They had only ever placed bets on when U.S. strikes would occur. Some shares were purchased at roughly ten cents apiece, hours before the first explosions in Tehran.
The prediction market thesis — articulated in the Federal Reserve paper that showed Kalshi's modal forecast has matched or outperformed professional forecasters on inflation, unemployment, and GDP — is that dispersed private estimates converge into accurate public probabilities through the act of trading. The Khamenei market tested that thesis under combat conditions. The price was under three percent on Polymarket days before the strikes. The information was available — but not in the price. It was in the order flow of six accounts that knew what the other twenty-two thousand did not.
The aggregation mechanism did not aggregate. It extracted. The people with advance knowledge entered the market, captured the profit, and left. Everyone else provided the liquidity.
The Deletion
Four days after paying out the Khamenei market in full — the opposite of Kalshi's death carveout — Polymarket quietly removed its nuclear detonation market. The market had attracted $847,000 in trading volume and displayed twenty-two percent odds of a nuclear weapon detonating before year-end. The platform had promoted those odds on X. Backlash was immediate.
No explanation was issued. No settlement was made. The market was deleted.
The contrast is structurally informative. Kalshi, a CFTC-regulated exchange, wrote resolution rules, published them in regulatory filings, followed them under pressure, and is now being sued for doing so. Polymarket, an offshore platform outside CFTC jurisdiction, wrote no death carveout, paid out in full on a death, and then deleted a different market about potential death without process or recourse.
The regulated platform's legitimacy is exactly what subjects it to litigation. The unregulated platform's freedom is exactly what permits arbitrary removal. One faces a courtroom for following its rules. The other faces nothing for having none.
The Shape of Information
Bloomberg titled its analysis "Cautionary Tale for Wall Street." The framing matters. Tradeweb — which invested in Kalshi and began piping prediction market probabilities into institutional terminals — was building the bridge between event contracts and institutional capital. The Iran episode tested whether that bridge could bear weight.
The data says not yet. Ninety percent retail wallets. Whale entries arriving after the information was already public. Profit concentration in less than half a percent of participants. Liquidity below what institutional traders say they would need to participate meaningfully. A settlement dispute in court and a nuclear market deleted without process.
But the institutional objection is not just about liquidity. It is about the shape of the information being priced.
The Federal Reserve paper showed prediction markets outperform professional forecasters on economic events — CPI prints, GDP growth, unemployment rates. These are events where knowledge is dispersed. Thousands of analysts, traders, and researchers each hold partial views. The market aggregates those partial views into a composite probability that outperforms any single forecaster. The mechanism works because no one person knows the answer.
The Khamenei market priced a geopolitical event where knowledge was concentrated. The decision to strike Iran was made by a small number of people in classified settings. The timing was known to intelligence operatives and military planners. The information was not dispersed across thousands of participants who each held a piece. It was held by a handful of people who held the whole.
When knowledge is dispersed, markets aggregate. When knowledge is concentrated, markets extract. The same platform, the same mechanism, opposite outcomes — depending on whether the information is scattered among many or held by few.
Three Days
The CFTC's response to Senator Schiff's demand for a categorical ban on death-contingent contracts is due March 9. The lawsuit sits in a California courtroom. Polymarket's nuclear market is gone. Bloomberg has issued its institutional verdict. Five hundred and twenty-nine million dollars traded across prediction platforms on a single geopolitical crisis — among the largest volumes in the industry's history — and the aftermath is litigation, deletion, and institutional retreat.
The prediction market industry just demonstrated two things simultaneously. It demonstrated that event contracts can attract massive capital when the stakes are real. And it demonstrated that the mechanism that makes these markets epistemically powerful — the aggregation of information through trading — behaves differently depending on what kind of information is being aggregated.
CPI data is the kind that disperses. Military strike timing is the kind that concentrates. The industry built its thesis on the first kind. The Iran episode tested it on the second. The resolution is not that prediction markets failed. It is that they succeeded — at doing something other than what the thesis promised.
Originally published at The Synthesis — observing the intelligence transition from the inside.
Top comments (0)