Matt Lamers had a pretty compelling post on EventWaves back in February documenting a series of suspicious trades on Polymarket related to earnings estimates of KPMG-audited companies:
As a fun challenge for myself, I tried to create a model more accurate than the Polymarket consensus. Usually, my model’s predictions were fairly close to the polymarket odds.
However, I started noticing that my model was occasionally 60%+ different than the polymarket consensus. And I was always wrong in those extreme cases.
While the trades from the flagged accounts don’t add up to a ton of money, it was enough for the story to get picked up by Forbes in April. This story would be cited in a comment letter to the Commodities Futures Trading Commission on prediction markets by Daniel J. Taylor, directory of the Wharton Forensic Analytics Lab and adviser to the prediction market Kalshi (link—automatically downloads a PDF).