Prediction‑market advocates often present platforms as gauges of public sentiment, sometimes calling them “truth machines” that can out‑perform polls. In 2024, for example, Kalshi out‑predicted many pollsters on the U.S. presidential race. Yet the trading data can also reveal puzzling patterns.
Two global election markets on Polymarket’s offshore site generated more than $600 million in volume, far exceeding the usual activity for political markets. The bulk of the trades involved contracts tied to long‑shot candidates who ultimately lost.
In Ethiopia, Prime Minister Abiy Ahmed’s Prosperity Party swept the federal election, securing 438 seats in parliament and positioning Ahmed for another five‑year term. The nearest rival won only 13 seats. Normally, trading volume on prediction markets mirrors the odds of each outcome, with frontrunners drawing the most activity. Here the pattern was starkly different: only $173,000 was wagered on Ahmed’s re‑election, while over $41 million changed hands on a contract that his chief of staff, Shimelis Abdisa, would become the next leader. The total volume in the Ethiopian market topped $285 million, almost entirely on candidates who did not win.
Even stranger, about $284 million of that activity came from pairs of accounts that repeatedly bought and sold with each other, according to former Polymarket trader Chris Park, who now leads the forensic analysis firm MSR Decode. Experts have called the behavior a form of volume manipulation. “It’s obviously some form of volume manipulation,” said Rajiv Sethi, an economics professor at Barnard College who studies prediction markets.
When WIRED asked Polymarket about the activity, Shana Bautista, the firm’s head of global intelligence and investigations, said the team “immediately flagged this activity, investigated, and blocked the accounts from trading.” Polymarket did not explain why the accounts were able to generate such high volume before being shut down.
The timing added to the oddity. Polymarket’s rules stipulated that the Ethiopia market would settle in October after the prime minister’s formal appointment, meaning the market stayed open for months after Ahmed’s victory was essentially certain. For 123 days after the election, bursts of seemingly irrational trading continued, prompting users to post comments like “THIS NOT RESOLVED?” and “fuckkkk.” Sethi describes the situation as a “zombie market,” where participants trade on events that have already occurred.
A second market, covering Romania’s 2024 election, displayed even more dramatic anomalies. The original election was halted over concerns of foreign interference, and a new vote was held months later. Polymarket kept the original market open alongside a new one. During that overlap, an algorithm designed by Sethi’s team to detect wash trading flagged nearly all the volume as suspicious. “Ninety‑nine point six percent of volume in this market for certain weeks has been wash trading,” he said.
Wash trading involves accounts repeatedly buying and selling the same asset to create the illusion of active trading, while in reality money merely circulates between the same parties. The practice is prohibited on regulated U.S. exchanges and is explicitly banned in Polymarket’s terms of service. Earlier research by Sethi and colleagues uncovered widespread wash trading on Polymarket’s offshore platform.
Although wash trading could explain the Ethiopian and Romanian cases, the unknown identities of the traders leave some uncertainty. “What appears to be wash trading isn’t always wash trading,” Sethi cautioned.
MSR Decode’s analysis of the Ethiopian market suggested the most plausible explanation was “activity farming,” where participants inflate their trading volume in hopes of benefiting from a future Polymarket token. In 2025, Polymarket’s chief marketing officer Matthew Modabber told a crypto‑focused podcast that a token was planned, though details were sparse. Sethi agreed that traders might have been motivated by potential token allocations to high‑volume accounts.
Such behavior has been observed elsewhere in the crypto space. “These are textbook examples of how market participants farm a potential airdrop,” said Andre Guettler, director of Ulm University’s Institute of Strategic Management and Finance.
MSR Decode examined over 38,000 “trade loops” that occurred after the Ethiopian election—instances where individual wallets bought and sold to each other in rapid succession. More than 98 percent of these loops resulted in neither side gaining nor losing money, and the median holding time for contracts was just nine seconds. The firm concluded that sophisticated algorithmic traders were unlikely to be responsible for the anomalous volume.
These unusual trading patterns are not new on prediction‑market platforms, but the Ethiopian and Romanian episodes highlight how easily volume can be inflated when markets remain open after outcomes are effectively decided.