Think of a fantasy football league where someone forgot to close the waiver wire after the Super Bowl, and bots kept trading players from eliminated teams for 123 days straight. That’s roughly what happened inside Polymarket’s Ethiopian election market, where $285 million changed hands after the winner was already obvious, with a median trade holding period of about nine seconds.
Prediction markets have spent the last two years auditioning for the role of truth machine. Kalshi’s performance during the 2024 U.S. presidential election gave the sector a credibility lift, reinforcing the idea that market activity can outperform traditional polling. When volume is inflated by accounts cycling contracts between themselves in under a minute, that “wisdom of crowds” framing starts looking thin. Volume can just as easily reflect bots, platform incentives, or unresolved contracts as it can reflect informed opinion.
The Ethiopian numbers are difficult to explain away. Abiy Ahmed’s Prosperity Party won 438 parliamentary seats; its closest competitor took 13. Yet only $173,000 traded on Ahmed winning, while more than $41 million moved on a long-shot candidate who lost.
MSR Decode founder Chris Park, speaking to WIRED, said approximately $284 million of the $285 million total came from pairs of accounts repeatedly buying and selling contracts with each other, according to his analysis.
Barnard College economist Rajiv Sethi examined more than 38,000 post-election trade loops in that market. More than 98% produced zero gain or loss for either side.
Sethi called it “obviously some form of volume manipulation” in WIRED. He was careful to add, however, that suspicious patterns do not conclusively prove wash trading, since trader identities and intentions remain unknown. He described the phenomenon as a “zombie market,” active long after the underlying event stopped mattering.
When the Platform Design Is the Problem
Romania tells the same story with a different setup, and the numbers are, if anything, worse.
Romania’s original 2024 presidential election was canceled amid foreign-interference concerns. Polymarket left that market open alongside a newer one covering the rescheduled vote. Sethi’s detection algorithm flagged an estimated 99.6% of volume during certain weeks as potential wash trading.
One possible motive was speculation about a future Polymarket cryptocurrency token. CMO Matthew Modabber reportedly told WIRED in 2025 that “there would be a token,” without specifics. Andre Guettler of Ulm University described the trading patterns as characteristic of airdrop farming, where users run circular, low-risk transactions to qualify for future token distributions. That explanation remains possible, not confirmed.
Polymarket’s head of global intelligence and investigations, Shana Bautista, told WIRED the company flagged, investigated, and blocked the relevant accounts. Polymarket did not answer follow-up questions about why markets with more than $600 million in reported volume accumulated that activity before intervention arrived.
This pattern echoes the streaming-count gaming era in music, when engineered play counts manufactured chart legitimacy that artists and labels then cited as proof of cultural relevance. Inflated numbers become their own argument, until someone examines who is actually listening.
Polymarket reportedly raised more than $1 billion at an estimated $21 billion valuation, with an annualized revenue run rate above $1 billion and trading volume as a central growth metric. At that scale, faster market resolution after outcomes are known, cleaner separation of organic volume from incentive-driven activity, and transparent surveillance disclosures are the minimum the platform owes its users. Researchers who keep citing its numbers as signal deserve the same transparency. The CFTC declined to comment on whether it was examining the trades, which means that accountability, for now, falls entirely on Polymarket itself.
