Why most Polymarket traders lose

By the Cent Signals editorial desk. Reviewed June 17, 2026.

A common observation about Polymarket is that most accounts end up down. That is not a moral judgment about the traders involved, it is a consequence of how the market is built and what the public data shows. This guide describes the structure and the findings, drawing on a 2026 study by Akey, Gregoire, Harvie, and Martineau, "Who Wins and Who Loses in Prediction Markets? Evidence from Polymarket". It explains why the pattern exists rather than what anyone should do about it, and like everything on this site it is description, not direction.

A zero-sum structure

Every Polymarket contract is a binary claim that pays one dollar if it resolves YES and zero if it resolves NO. Unlike a share of stock, there is no equity premium, no dividend, and no passive return that accrues simply for holding. The total value paid out at resolution is fixed by the outcome, so the pool of money is not growing on its own.

That makes the activity zero-sum before costs: every dollar one trader gains on a contract is a dollar another trader loses on the opposite side of the same claim. The average participant therefore cannot come out ahead in aggregate, and once costs such as the spread are added, the average outcome sits below zero by construction.

Gains go to a small minority

The study examined roughly 67 billion dollars of volume and found the gains concentrated in very few hands. The top 1 percent of profitable users captured about 76.5 percent of all profits, and roughly 69 percent of all users ended with a loss. When winnings concentrate that heavily at the top, the rest of the distribution is, by arithmetic, on the other side of those gains.

This is the mechanical link between the zero-sum structure and the headline result. A market that pays a small group of accounts the large majority of its profits is a market in which most accounts must have funded those profits.

Takers pay the spread on every order

Most losing traders in the study used market orders, which cross the bid-ask spread and pay it on each trade, while those posting resting orders earned that same spread. The cost is small per order but accumulates with activity. The authors found that removing even the minimum-tick portion of the spread cost would have moved about 18.5 percent of losing accounts, roughly one in five, into non-negative territory.

The spread did not explain everything, though. It could not account for the deepest losses, which the study traced to systematic forecasting errors rather than to transaction costs. For the distinction between the two sides of every order, see makers vs takers.

Extreme-price exposure

Losing accounts in the study skewed toward contracts priced near the extremes, at a few cents or in the high nineties. Those are precisely the prices where the long-documented favorite-longshot bias appears: across large samples, tail outcomes have tended to occur less often than their price implies. Concentrating positions there exposes an account to that historical gap.

The bias is a statistical tendency rather than a verdict on any single market, and the effect is measurably weaker on low-fee venues. For the evidence behind it, see the favorite-longshot bias.

Churn and selection

The population of traders turns over quickly. The study reported that about 44 percent of users stopped trading within a month and up to 66 percent within six months. Most participants exit early, often after a loss, rather than staying long enough to compound any edge they might have.

That churn shapes the aggregate figures. A market dominated by short-lived accounts that leave after a setback will show a very different profit distribution from one populated by participants who trade steadily over years, and the loss rate reflects who actually stays in the data.

What it does not say

These figures describe an aggregate distribution across many accounts, not the fate of any individual account, and they say nothing about how a particular trader will fare. The same body of work also finds that Polymarket prices are well calibrated overall, meaning that as implied probabilities they track realized frequencies closely across most of the range. A well-calibrated market and a population in which most accounts lose are fully compatible under a zero-sum structure with costs.

Nothing here is advice, a tip, or direction to take a position. It is a description of public findings. For who sits on the winning side of the distribution, see who actually wins on Polymarket, and for the cost mechanics behind it, see makers vs takers.

Frequently asked questions

Do most people lose money on Polymarket?

In the largest study of the platform to date, covering roughly 67 billion dollars of volume, about 69 percent of users finished with a loss. The gains were heavily concentrated, with the top 1 percent of profitable accounts capturing around 76.5 percent of all profits. So in aggregate the typical account ended down, which is what a zero-sum structure with costs produces.

Why is prediction-market trading described as zero-sum?

Each Polymarket contract is a binary claim that resolves to one dollar or zero, and every position has a counterparty on the other side of the same claim. There is no equity premium, dividend, or passive return as there is with stocks, so one trader's gain is funded by another trader's loss. Before costs the average participant nets to zero, and after the spread and fees the average lands below zero.

Does the spread really matter that much?

The study found that traders who placed market orders paid the bid-ask spread on each trade, while those who posted resting orders earned it. Removing even the minimum-tick portion of that spread cost would have moved about 18.5 percent of losing accounts, roughly one in five, into non-negative territory. It did not explain the deepest losses, which the authors attributed to systematic forecasting errors rather than to costs.

Do most Polymarket users keep trading?

No. The study reported that about 44 percent of users stopped trading within a month and up to 66 percent within six months. Most participants exit quickly, frequently after a loss, so the population is dominated by short-lived accounts rather than by people who trade repeatedly over a long period.