Polymarket Gave Lula 20% Before Brazil Election Win

Before Brazil’s second-round presidential vote on Sunday, Polymarket priced Luiz Inácio Lula da Silva at roughly 20%, or one chance in five, and Flávio Bolsonaro at 80%; Lula won and secured a fourth term. The 80-percentage-point gap between the market’s pre-vote call and the result turned one forecast failure into a test of market structure, regulation and the reliability of retail trading.

Bolsonaro, the right-wing challenger, led after the first round in early October but did not reach a majority, forcing a runoff. Lula won the second round with a majority. Prediction markets settle at $1, so the reported $0.60 price for Lula “Yes” implied a 60% chance before the outcome.

What did Polymarket misprice?

FhantomBets, a pseudonymous Polymarket whale, said professional “sharps” judged the election-day probability closer to 30%–40%. The trader said Lula was nearly a lock by 6:15 p.m. ET, when the “Yes” contract traded at $0.60, and described the market as slow to reflect incoming results.

“Retail traders were so against Lula that you could get unlimited size at 60-80c when Lula had [essentially] already won.” FhantomBets said.

FhantomBets also said many whales, including himself, sold long-term positions to raise liquid capital so they could add more size to Brazil. His account placed less-experienced retail traders on the opposite side, offering unlimited size at 60–80 cents after Lula had effectively won. The interviews provide no order-book record to test that claim.

Did retail traders distort the odds?

Luke Nolan, senior researcher at CoinShares, rejected the premise that more users inherently make a market worse. He said Polymarket repriced as polls and results rolled out through the night, and argued that users can add liquidity, different perspectives and local knowledge.

He said the relevant test was whether participation introduced information that traders could act on, not whether those traders held professional credentials. He cautioned against explaining the result simply as professionals being right and retail traders being wrong.

Nolan identified three modeling errors:

• Predictors gave too much weight to Bolsonaro’s first-round performance.
• They underestimated the increase in voter turnout for the second round.
• They assumed how voters would transfer after other candidates were eliminated.

Still, Nolan said a 20% pre-vote price represented a meaningful chance, not a formality. His larger concern was a persistent pattern of overconfidence or failure to incorporate relevant information, rather than an individual upset.

What did Brazil’s April ban change?

Brazil issued a sweeping ban against prediction-market platforms in April, citing investor protection and the rise of gambling addiction. Finance Minister Dario Durigan said the platforms violated betting regulations and lacked proper regulatory oversight. The account provides no statute number, implementing rule or enforcement deadline, so the ban’s practical reach cannot be checked from this material.

Nolan said the ban was likely the larger cause of the mispricing. Einar Knobel, CEO and co-founder of sportsbook and prediction-market firm TxODDS, agreed that people outside Brazil set the price by trading the first-round result, poll headlines and the narrative built around them.

Knobel also said liquidity was relatively thin: the Brazil market attracted just 4.5% of the volume recorded for the 2024 U.S. presidential election. In his view, fewer participants and almost no local knowledge left little in the bid-and-offer book to correct the price. His account identifies a thin market as the practical constraint, although the comparison provides no raw volume or participant breakdown.

What does one election prove?

Nolan cautioned against treating the upset as proof that prediction markets had failed. He said they continued to price events better than traditional betting platforms, citing Betfair’s 83% chance for Hillary Clinton on the day before the 2016 U.S. election. That is a single benchmark, not a systematic calibration study in the account.

“Across enough events, outcomes given a 20% chance should happen roughly 20% of the time. We should also compare their performance with other forecasts available at the same time,” Nolan told Decrypt.

His point addresses calibration rather than one result: an event assigned an 80% chance should not fail 80% of the time across a large sample, though any individual forecast can fail. The interviews leave causation unsettled because no local-user share, order-book depth, polling-error analysis or execution record appears in the material.

FhantomBets attributes the error partly to retail traders; Nolan and Knobel emphasize the ban, offshore participation, thin liquidity and modeling errors. The evidence therefore supports a material mispricing, but not a definitive causal ranking of participant groups.

What changes after the ban?

Any assessment now depends on how authorities implement the April ban. Stricter enforcement could push more activity offshore and deepen the local-information deficit, while a workable framework could restore participants and liquidity. No implementation date is available, so the next verifiable test is the market structure after enforcement, not a calendar deadline.