SSRN Research Paper: Polymarket’s Pricing Accuracy Comes From 3.14% Informed Traders

According to an academic working paper published on SSRN on April 20, a research team from London Business School and Yale University analyzed Polymarket’s complete trading records and concluded: the platform’s pricing accuracy is mainly driven by 3.14% informed traders, rather than the collective wisdom of all accounts. The study covers 98,906 events, 210,322 markets, involves 1.72 million accounts, and totals $13.76 billion in trading volume.

Research Method and Core Findings

(Source: SSRN)

According to the paper, the researchers used a “Sign Randomization Test” statistical method, grouping traders based on the source of their profits (real skill vs. random luck).

The results show that on Polymarket only 3.14% of accounts meet the standard of “skill-based winners”: they can consistently profit (with robust out-of-sample test results), participate on average in 79 markets, and their position direction always matches the final market outcome. The remaining 96% of accounts either break even due to luck or lose money. According to the paper, for every one-percentage-point increase in the net buy-in volume of skill-based traders, the probability of correctly predicting the final outcome increases by 8 basis points; although the lucky winners’ accounts have positive balances, they show no meaningful predictive ability in either of the two tests.

Skill Persistence: Better Than Mutual Funds

According to the paper, the researchers randomly split events into a training set and a test set: in the training set, traders categorized as “skilled” had 44% still remain in that classification in the test set; among “unskilled” losing traders categorized as such, 51% still retained the original classification. By comparison, in the parallel test, only 10% of mutual funds categorized as “skilled” could keep their classification. The paper notes that prediction markets exhibit unusually high persistence on both the skill and anti-skill sides.

Insider Trading Identification and CFTC Action

According to the paper, the researchers identified 1,950 accounts that meet timing and confidence standards, suggesting trading may have been based on non-public information; on average, each account profited about $15,000.

One well-documented case involves three accounts: in the hours before the start of a U.S. secret military operation on January 3, 2026, the three accounts built positions in contracts related to Venezuelan President Nicolás Maduro; the three accounts together earned more than $630,000 in profit. According to publicly available court records, on April 23, 2026, the CFTC filed a lawsuit in this case, accusing an active-duty U.S. service member of using one of the accounts to engage in insider trading.

The paper also concludes that insider trading activity is too concentrated in isolated events to explain the broader price-discovery mechanism across the entire platform.

Account Composition Distribution

According to the paper, the overall distribution of Polymarket accounts is as follows: 67% of accounts belong to the “unlucky and unskilled” losing group, which bears all of the platform’s cumulative losses; market makers and technically proficient “buyers” together account for less than 3.5% of the total number of accounts, yet they receive more than 30% of total earnings.

Frequently Asked Questions

What are the authors and data scale of this Polymarket study?

According to the SSRN research paper, it was published on April 20, 2026 (with revisions on April 25), and is co-authored by Roberto Gomez-Cram, Yunhan Guo, and Howard Kung of London Business School, as well as Theis Ingerslev Jensen of Yale University. The data covers Polymarket’s 98,906 events, 1.72 million accounts, and total trading volume of $13.76 billion.

What exactly is the performance of the 3.14% informed traders found in the research?

According to the paper, for each one-percentage-point increase in the net buy-in volume of these skill-based traders, the probability of correctly predicting the final market outcome increases by 8 basis points. In the skill persistence test, 44% of accounts maintain their skill rating, far higher than the 10% for mutual funds.

What are the details of the CFTC’s lawsuit regarding the insider trading case in the Maduro contracts?

According to the research paper and publicly available court records, on April 23, 2026, the CFTC filed a lawsuit accusing an active-duty U.S. service member of using Polymarket accounts to conduct insider trading in contracts related to Maduro, in the hours before the start of a U.S. secret military operation on January 3, 2026. The three involved accounts together earned more than $630,000.