PillarLab AI study finds blind copy-trading of Polymarket whales fails
PillarLab AI published a 30-day study of Polymarket that tracked 214 arbitrage setups and identified only 31 as viable. The analysis concluded that blind copy-trading of whale positions does not reliably produce trading edge. The published figures did not include methodology details, specific whale identities, or profitability results beyond the success rate of identified setups. PillarLab AI argued that structured analysis is needed instead.
The 31-of-214 success rate resets expectations for retail traders who treat large-wallet positions as actionable signals. It means that whale tracking without structural analysis is closer to noise than alpha, which undermines a common retail strategy on Polymarket. For the platform, this study feeds a credibility problem: if the most visible trading edge is illusory, sophisticated traders may demand more transparency on market-making flow from DRW, Wintermute, and IMC before committing size.
Retail volume could stall as copy-trading shrinks. PillarLab AI's conclusion that structured analysis closes the gap is an opening for data vendors, but only if they can prove their own signals beat the base rate. The immediate consequence is a harder growth path for Polymarket's retail business just as it courts institutional liquidity.
The finding that whale-behavior signals do not guarantee trading edge squares with the documented $40 million arbitrage economy on Polymarket, where sophisticated traders profit from structural pricing gaps rather than from mimicking large positions.