Karshi and Polimarket.
Gabby Jones Bloomberg | Martin Lelièvre | Getty Images
Prediction market platform Polymarket has seen trading volumes explode since its launch, with notable spikes in the 2024 election and last fall. However, the majority of closed individual markets on the site never had a reported volume of more than $10,000.
According to Polymarket’s Gamma API, approximately 70% of Polymarket’s total closed markets had a reported value of less than $10,000 from 2021 to the end of May this year, according to a CNBC analysis. Gamma API records notional volume on both sides of a trade.
Less than 10% of the total closed markets raised between $100,000 and $1 million in reported trading volume.
More than 45,000 markets, or nearly 5% of all closed markets, reported no trading volume at all.
According to an analysis conducted on the on-chain platform Dune, Polymarket’s competitor Kalshi also had a large number of shallow markets. Unlike Polymarket’s Gamma API, Kalshi’s notional trading volume on Dune only counts towards one side of the trade.
A market with low trading volume is not ideal for prediction market traders. First of all, prices can fluctuate widely, said Konstantin Bürgi, professor of economics at University College Dublin.
“A thin market essentially means that small investments can cause large market fluctuations and are usually more volatile,” Bürgi told CNBC.
Eric Zitzewitz, an economics professor at Dartmouth College, said new traders could also be left vulnerable in low-volume markets because buy-sell spreads can widen and trade prices may be higher.
not very attractive
Thin markets are less attractive even to experienced traders.
“I like high-volume, short-term (markets),” said Logan Sudays, a 26-year-old Atlanta-based former financial risk analyst who started trading prediction markets full-time last fall. “It’s more capital efficient.”
Polymarket reported the highest number of contracts with a volume of at least $1 million in markets lasting up to one week. These week-long markets included deals related to the war against Iran, US President Donald Trump, or Elon Musk.
Short-term markets can be a sweet spot for traders, as can markets with a large number of participants.
“People like to trade things that are closer to resolution,” Zitzewitz, a professor at Dartmouth College, said last month, noting that participants “are more likely to trade when there are a lot of people in the market.”
Bots dominate shallow markets
Within Polymarket, bots account for more than 80% of trading volume in markets under $10,000, said Joshua Dela Vedova, a business professor at the University of San Diego.
Della Vedova identified wallets or digital accounts that made more than 50 transactions per day or more than 1,000 transactions in total as bots.
Using this definition, Della Vedova found that from November 2022 to February 2026, bots earned approximately $1.2 million in shallow markets and approximately $35.1 million in markets with volume of $10 million or more.
“They are making money in every market,” Dela Vedova said, using Polymarket’s on-chain data. This is in contrast to retail traders who face losses in shallow or heavy markets.
Bots dominate trading volume in shallow markets, but they do not push prices away from fair value due to the risk of large losses. The University of California, San Diego professor added that polymarket bots prefer markets with heavy trading over markets with low trading because their ultimate goal is to make a profit.
“[Bots]prefer to trade in these larger markets because they make money on every trade, but they will trade across the entire range,” Della Vedova said.
Accuracy in thin markets
Experts are divided on whether it is accurate to say that the market is thin.
Evercore ISI Strategists analyzed five years of completed markets for both Polymarket and Calci and found that high volume markets have more reliable odds than thin markets.
“Most of the market probability lies in the thin tail of trades, the weakest part of the correction,” the strategists said after finding that only 8% of the market reached $1 million in volume on both platforms.
Other researchers stated that the relationship between market size and accuracy is nonlinear. For Tice Ingerslev Jensen, a finance professor at Yale University, accuracy is determined by who is trading, not how much is being traded in a particular market.
Jensen and researchers at the London Business School found that skilled or knowledgeable traders drive much of polymarket accuracy.
“A thin market doesn’t automatically make it inaccurate, but it does make it less reliable,” Jensen told CNBC. “The key question is whether experienced traders still have enough motivation and ability to trade.”
influence remains
Harry Crane, a statistics professor at Rutgers University, said the abundance of shallow markets on both platforms is unlikely to affect the operation of prediction markets to the general public or Wall Street.
“The volume of trading in these markets should be taken into account,” Crane said, but “lack of liquidity in and of itself does not discredit market signals or make markets economically useless.”
Polymarket declined to comment, and Carsi did not respond to CNBC’s request for information on the findings.
As prediction market volume continues to grow at a breakneck pace, Crane expects lower volume markets to remain shallow while larger markets expand. The important thing is that traders are aware of the risks.
“Always protect yourself,” Crane added. “Each company will have to deal with it on its own.”
Methodology:
CNBC obtained all closed market data from 2021 through the end of May 2026 via Polymarket’s Gamma API. Gamma API counts notional amounts on both sides of a trade. For this reason, CNBC wrote about the amount “reported” from the Gamma API.
This analysis on Polymarket was reviewed by Joshua Della Vedova, Professor of Business Administration at the University of San Diego. He cross-checked our findings with an independent on-chain transaction dataset consisting of 222 million resolved on-chain polymarket transactions. Our findings were consistent with those of Della Vedova.
