On Polymarket, “human” accounts lose and those who trade in bursts win. Galaxy Research scrutinized 2.9 million individual accounts: 69.2% ended below the break-even threshold, for $338.9 million in cumulative losses.
Conversely, accounts deemed automated according to Galaxy's threshold earned $246.8 million. Another study had already dispelled the myth of the bettor who is right.
Polymarket: 69.2% of individual accounts lose money
Will Owens, an analyst at Galaxy Research, published the study on October 1 with the oracle Stork. It covers the entire on-chain history of the international platform since 2020: 1.27 billion orders placed by 3.07 million wallets, for $82.8 billion in volume.
“Retail” does not mean small carrier. The criterion is pace: a well-endowed, hand-clicking trader enters the sample. Galaxy counts addresses, not identities. The same human can therefore hold several wallets.
The median account lost about $3, roughly half a percent of everything they bet. The middle half of the accounts range from a loss of $36.64 to a gain of $0.40. At the ends, the gap widens: the most losing 1% left at least $4,804, the most winning 1% pocketed at least $3,381. One in ten accounts lost at least 90% of their stake.
Automated accounts according to Galaxy: $246.8 million in winnings
Galaxy excluded 125,429 accounts (4.1% of the total) that exceed 50 orders per day of activity. This small group accounts for 80.8% of orders and 41% of volume.
Together, they earned $246.8 million. There are reward hunters, but also real market makers and arbitrageurs. Galaxy warns that its threshold is a choice of method, not proof that they are robots. And the two amounts do not offset each other: 246.8 million on one side, 338.9 on the other, it is not the same fund.
Losing makes you drop out: 15.2% of losers leave for a month
After a losing position, 15.2% of accounts do not open another position within 30 days. After a gain, they are only 6.1%. A loser is therefore 2.5 times more likely to give up. Be careful though: a “left” account may have just changed wallet.
No rushing forward to rebuild, however. At a comparable entry price, it is the winners who most often increase the size of their positions, especially on contracts quoted above $0.50. After a gain as after a loss, the median risk taking does not change.
Sport, politics, tech: not all specialists are equal
44.1% of traders concentrate more than 60% of their activity on a single theme. These specialists are doing a little worse than generalists: 28.1% of profitable accounts, compared to 30.4%.
Blame it on sport. It brings together 47% of specialists and has the worst success rate, at 25.1%. Politics and culture also fall under the bar. Conversely, finance specialists reach 36.8% and those in tech and science 41.2%, on a smaller sample. Relying solely on OpenAI model outputs implies an advantage. Betting on the NFL on Sunday, much less.
Profitable accounts also bet bigger: $13.96 on median, compared to $10 for losers, and the difference holds for the same level of activity. They trade more often. On the duration of detention, Galaxy, however, finds no clear link.
The noise of the punters pays the prediction
The bill may increase. Polymarket introduced taker fees (paid by the person taking the liquidity) in January on crypto markets, then on most categories in March. A single catch at even odds now costs several times the half percent that the median account has lost over its entire history.
Galaxy does not draw an argument against predictive markets from this. Informed capital needs a less informed flow in the face. It is this noise that finances the forecast, and Polymarket knows it: its partnerships with LeBron James or Eli Manning target less selective takers, not seasoned market makers.
