Dug into some @Polymarket data over the weekend to understand how much wash trading there is and what that actually means for prediction markets. Key finding: Polymarket has rampant low-value wash trading, but the trades that truly matter i.e. high-value trade volumes, are taking genuine directional bets. Using @Dune, I pulled all trade data from its 3 main 2028 political markets: Presidential Election Winner, Democratic Nominee, Republican Nominee (~$500M USDC, 8.8 million trades, Jul 2025-Apr 2026). To estimate wash trading activity, I used a basic “round-tripping” filter i.e. wallets that simply bought and sold equal number of shares — netting out to zero. In other words: did this wallet take a directional position, or just bounced in and out? The blue bars in chart below shows an alarmingly high 60-65% of volumes on Polymarket’s top political markets are wash-traded. But there is noise in that signal because legitimate market makers on Polymarket also quote tight spreads and get filled on both sides, similar to wash traders. So I applied another filter to exclude transactions below $500 and $1000, which is where market makers typically trade at. The outcome is wash trading vols dramatically falls to 1–4% (see orange and grey bars). If you believe large trades = “smart money”, then I think Polymarket as a “global truth machine” is intact. Would love to hear different takes.