According to an analysis by the Wall Street Journal, over 70% of Polymarket users lose money. Losses are often modest for the average user — between $1 and $100, but significantly heavier for the underperformers: the bottom 10% show an average loss of approximately $4,000 each. On Kalshi, a similar imbalance exists: according to the platform, there are 2.9 losing users for every winning one.
Another compelling figure: 67% of Polymarket's profits are captured by just 0.1% of accounts. In other words, fewer than 2,000 accounts have reportedly amassed nearly half a billion USD in profits, out of a sample of 1.6 million active accounts since November 2022, while the platform is estimated to have at least 2.3 million total accounts.
Standard-bearers for the democratization of "event finance," prediction markets ultimately resemble an arena where amateurs provide liquidity and professionals harvest the margins.
It must be noted that prediction markets are no longer a crypto curiosity or a playground for US politics enthusiasts. It is a massive, liquid, and highly publicized market, fueled by social media, influencers, sports bettors, crypto traders, news junkies, and quantitative firms. In April, cumulative trading volumes on Kalshi and Polymarket reached $24.2bn, compared to $1.8bn a year earlier.
token terminal
Follow politics? Bet on elections.
Love sports? Bet on matches.
Know the markets? Bet on rates, inflation, or bitcoin.
Watch American TV? Bet on the words that will be spoken on air.
Know your viruses? Bet on Hantavirus.
Will there be an hantavirus pandemic in 2026?
The Wall Street Journal investigation shows that the professionals in these markets are not there to play around. By heavily utilizing third-party data feeds, AI agents, and algorithms to predict price fluctuations and manage risk faster than any human, these "Trader 3.0" entities can execute up to 60 trades per minute and modify their orders 30 times per second.
On the other side is the average user who clicks "yes" because they believe an event will occur. The most revealing segment is that of "mention markets."
The trap of "mention markets"
The principle: betting on whether a public figure will say a specific word during a public appearance.
Will a president say "cartel"?
Will a candidate say "inflation"?
Will a celebrity say "rapper"?
Will a politician utter a specific word in a speech?
At first glance, these bets seem fun, accessible, almost ludic. There is no need to be an economist, statistician, or high-frequency trader. One simply needs to watch a show, know a public figure, or follow an intuition. But the figures show that these markets are among the most treacherous.
In February, Kalshi users wagered nearly 181 million USD on mention markets. For the State of the Union address alone, over 28 million USD was committed to the possibility of Donald Trump saying words like "cartel," "Somalia," or "hockey." Volume is exploding, but profitability for amateurs is collapsing.
Based on an analysis of over 35,000 completed mention markets on Kalshi, "yes" contracts displayed at a 50% probability actually pay out only about 40% of the time. In other words, users are overpaying for events they overestimate. On average, traders who bet "yes" at the first listed price — a typical retail behavior — lose 11% of their stake.
The long shot bias
This phenomenon is known as "long shot bias." Bettors overestimate improbable events. The more spectacular the potential gain appears, the more the brain tends to assign too much weight to the rare hypothesis. A low-probability contract seems attractive because it can yield high returns. But if it is systematically overpriced, it becomes a trap. This is exactly what is happening in mention markets.
The user sees a bet for a few cents. They think: "Why not? This word might come up." They imagine the winning scenario. They think of the potential multiple. They underestimate the actual probability of failure. And above all, they do not necessarily see that the price already incorporates an unfavorable margin.
A blurred line with gambling
Platforms insist they are not casinos. They present themselves as financial markets, sometimes regulated, sometimes compared to futures or options. In the United States, Kalshi and the American branch of Polymarket fall under the framework of the Commodity Futures Trading Commission. This does not prevent certain atypical cases. For instance, the regulator lifted anonymity for the first historic case of insider trading: a US soldier was charged in April 2026 for betting on the outcome of a secret Venezuelan military operation, pocketing $404,000 in illegitimate profits. Federal justice has initiated parallel civil and criminal proceedings against him. In capital cities, legislators are considering drastic bills to ban certain bets related to war or politics.
On the other hand, proponents of an optimistic outlook point out that these markets are not casinos. Research from the Federal Reserve Board suggests they provide economic indicators that are more responsive than traditional methods. In testing the Kalshi platform, Fed economists even found that it anticipates inflation better than the Bloomberg consensus, and its forecasts on the eve of Federal Reserve decisions have been perfect, outperforming traditional futures contracts. For them, Kalshi serves as a valuable "high-frequency" index for researchers and policymakers.
Ultimately, prediction markets are neither entirely casinos nor simple forecasting tools: they are both at once, depending on how they are used.
