A cluster of 19 interconnected Polymarket accounts won 41 of 42 bets tied to earnings of companies audited by KPMG, adding another potential insider-information episode to a prediction-market industry already facing two major U.S. criminal cases in the past five months.
Blockchain analytics firm Bubblemaps identified fund flows connecting the accounts, which collectively earned about $22,000 from wagers covering 18 KPMG clients, including Wells Fargo, Home Depot, DoorDash and General Mills. The bets stretched from November 2025 into 2026 and produced a success rate of roughly 98%.
The Wall Street Journal reported the findings on September 11. However, neither the analysis nor the unusual win rate proves insider trading, and the identities of the traders controlling the wallets have not been established.
The findings are particularly sensitive because federal authorities are separately investigating a KPMG employee suspected of using confidential information to wager on whether companies would beat quarterly earnings expectations. No connection between that employee and the newly identified account cluster has been established.
A Performance Pattern Concentrated Around KPMG
The concentration of successful trades around one accounting firm’s clients is what has attracted scrutiny.
A separate Bitquery reconstruction, published after the initial report, produced somewhat different figures because it grouped the transactions by prediction-market positions rather than individual wagers. It identified 21 KPMG-client markets traded by the linked accounts, with 19 winners and two losers, generating approximately $21,519 in profit.
Bitquery found 43 of 44 positions were opened before the relevant company filed its results. It estimated that prices paid by the traders implied roughly 14 expected victories across the 21 KPMG-client markets, compared with the 19 actually achieved. Performance elsewhere was much less unusual. On earnings markets involving clients of other accounting firms, Bitquery found the accounts won 11 of 15 markets, broadly consistent with the probabilities implied by their entry prices.
KPMG has said it has zero tolerance for trading using nonpublic client information and has strengthened its monitoring capabilities.
Another Major U.S. Insider-Information Episode Since April
The scrutiny comes after prosecutors brought major landmark Polymarket cases earlier this year. On April 23, the Justice Department charged U.S. Army soldier Gannon Ken Van Dyke with allegedly using classified information about the operation to capture Nicolás Maduro to place 13 Polymarket bets. Prosecutors say he risked approximately $33,034 and generated about $409,881 in profit.
On May 27, prosecutors charged Google software engineer Michele Spagnuolo with commodities fraud, wire fraud and money laundering. The government alleges he accessed confidential Google data before risking approximately $2.75 million on related Polymarket contracts and earning about $1.2 million.
Both defendants are presumed innocent unless proven guilty. The KPMG-related activity therefore represents the next major major insider-information controversy surrounding Polymarket since April, although unlike the previous two cases, charges have not been announced over the 19-wallet cluster.
The cases are testing how traditional market-integrity laws apply to prediction markets. Polymarket itself prohibits trading using confidential information where doing so breaches a duty of trust or confidence. For an industry increasingly positioning event contracts as financial markets rather than conventional gambling, the repeated appearance of traders with extraordinary informational advantages is becoming a central regulatory challenge.