Polymarket significantly mispriced the outcome of Brazil’s presidential election, assigning Luiz Inácio Lula da Silva only a 20% chance of victory against Flávio Bolsonaro before the final vote. Despite Bolsonaro leading in the first round, Lula won the second-round runoff to secure his fourth term as president. The discrepancy between market odds and actual results has sparked debate among prediction market experts regarding the causes of the error.
One prominent Polymarket trader, known as FhantomBets, attributed the mispricing to retail investors who allegedly inflated Bolsonaro’s odds by betting heavily against Lula, even as live results showed Lula gaining momentum. This trader noted that professional predictors recognized the market was undervaluing Lula at approximately 30-40% on election day, but retail activity kept prices low until late in the evening. However, other analysts, including Luke Nolan from CoinShares, challenged this view, arguing that broader structural issues were more significant than individual trader behavior.
Nolan and Einar Knobel, CEO of TxODDS, pointed to Brazil’s April ban on prediction markets as a primary factor limiting local participation and liquidity. With domestic traders unable to access these platforms, the market relied heavily on international participants who lacked nuanced understanding of Brazilian political dynamics. Knobel highlighted that the Brazil election market attracted only 4.5% of the volume seen during the 2024 U.S. presidential election, resulting in thin liquidity that failed to correct pricing errors effectively. Additionally, experts suggested that initial overreliance on first-round polling data and assumptions about voter transfer contributed to the miscalculation.
The Polymarket error in Brazil underscores the fragility of prediction markets when regulatory barriers exclude local knowledge. By banning these platforms, Brazil inadvertently created an information vacuum where global traders priced events based on incomplete or biased data, such as early poll headlines rather than deep regional insights. This case illustrates that liquidity is not merely about capital volume but also about the diversity and relevance of participant information. When local actors are silenced, the market loses its self-correcting mechanism, making it vulnerable to systematic biases introduced by external observers.
From a Market Structure perspective, this incident challenges the narrative that prediction markets are inherently superior forecasting tools. While they often outperform traditional polls, their accuracy depends heavily on efficient price discovery driven by informed participants. The reliance on retail sentiment, combined with low liquidity, distorted the signal. For future geopolitical events, especially those in jurisdictions with restrictive financial regulations, prediction markets may struggle to provide reliable odds unless they can integrate alternative data sources or attract sophisticated local proxies. Investors should treat high-confidence odds in thinly traded, restricted markets with caution, recognizing that price does not always equal truth.
