Brazil Election: Prediction Markets Outperform Polls in the First Round

Flávio Bolsonaro outpaced incumbent president Lula in the first round of the Brazilian presidential election, with 47.03% of the vote against 45.16%, sending the outcome to a runoff on October 25. Polls showed Lula winning; predictive platforms, however, had seen the senator’s lead since September. A contradiction that revives the debate on their reliability, although they are blocked in Brazil.

A first round against the polls

Already under fire in the United Kingdom, predictive markets closely followed the campaign in Brazil. On Sunday, however, results showed a gap with the main opinion polls.

Flávio Bolsonaro, senator and son of former president Jair Bolsonaro, gathered 47.03% of the votes, against 45.16% for Lula. Neither candidate obtained an absolute majority, so the runoff will take place on October 25.

However, Datafolha placed Lula in the lead with 42% of voting intentions, against 38% for Flávio Bolsonaro. Other polls also predicted a very close election.

After the results were announced, Lula acknowledged thinking he could win outright in the first round and called on his supporters to continue the campaign.

I was sure I would win the election in the first round. Bolsonaro represents uncertainty for the country.

Odds shifted as early as September

On Polymarket, the contract dedicated to the Brazilian election, launched in September 2025, has exceeded 171 million dollars in cumulative trades.

In early July, Flávio Bolsonaro had about 22% chances according to platform odds, against 60% for Lula. Then from September, the trend reversed. On September 20, the senator reached about 58%, while Lula dropped to 42%. Kalshi showed a comparable trend.

After the first round, Flávio Bolsonaro’s odds climbed to around 82%, against 17.5% for Lula, according to data collected on Monday.

The 2024 U.S. election had already drawn attention to this gap between polls and predictive markets. For the midterm elections, the two approaches converge this time toward a Democratic scenario.

Markets without Brazilian voters

This comparison deserves an important nuance. The Resolution 5,298 of the National Monetary Council, effective from May 4, prohibits derivatives related to political, electoral, sports, or cultural events.

The Ministry of Finance then ordered the blocking of 27 platforms, including Polymarket. In July, the latter limited Brazilian accounts to closing their positions. In other words, participants influencing these markets were not representative of Brazilian voters.

The phenomenon extends beyond Brazil. In France, Polymarket was blocked before a 2027 presidential race already closely followed on the platform. Elon Musk’s reactions on X at the end of September also helped draw attention to these markets.

The October 25 runoff will therefore shed new light on the gap between polls and predictive markets. But their particular audience remains an essential element to consider when analyzing their results. Their role continues to spark debate, moreover.