Brazil presidential election odds split from the polls

The Brazil presidential election is close enough on the calendar, and messy enough on the screens, that a single percent will lie to you. This is not investment advice. On the Polymarket event page opened early Friday, Oct. 2, 2026 (Moscow time), the winner market listed Flávio Bolsonaro at 60.1% with a Buy Yes price of 60.1 cents, and Luiz Inácio Lula da Silva at 40% with a Buy Yes price of 40 cents. A Reuters story from Sept. 24, quoting a Datafolha poll taken Sept. 22–23, still had Lula ahead of the same senator. The event page and the Reuters poll story are both real. They are not measuring the same object on the same day.

TL;DR - Polymarket's page showed about $159,810,998 in volume, an end date of Oct. 4, 2026, and Flávio as the priced favorite. A FAQ block on that same page rounded him to 60% and Lula to 40%. - Datafolha's simulated runoff, in Reuters, was Lula 47% and Flávio Bolsonaro 45%, inside a 2-point margin of error, after a prior week at 46% to 44%. - In the first-round scenario Reuters reported, Lula led 40% to 36%, with Augusto Cury at 5%, Ronaldo Caiado at 4%, and Renan Santos at 3%.

What actually happened

Reuters, from São Paulo on Sept. 24, says President Luiz Inácio Lula da Silva held a two-point lead over Senator Flávio Bolsonaro in a potential runoff. The Datafolha numbers are 47% for Lula and 45% for Bolsonaro. A previous Datafolha poll the week before had been 46% to 44%. Reuters says the result is a technical tie because it sits inside the poll's margin of error of 2 percentage points. The sample is 2,002 people, interviewed Sept. 22–23.

The first-round cut in that same story is not a tie in the same way. Lula leads with 40%, Bolsonaro 36%, Augusto Cury 5%, Ronaldo Caiado 4%, and Renan Santos 3%. The previous first-round read Reuters mentions was 39% for Lula and 36% for the senator. Reuters also notes that if no candidate wins more than 50% of valid votes in the first round, the top two advance to a runoff. It identifies Senator Bolsonaro as the eldest son of former president Jair Bolsonaro, who Reuters says is serving a 27-year prison sentence for attempting to overturn his 2022 defeat.

That poll is more than a week older than the market snapshot in this article. Do not "update" 47 and 45 by vibes. If a newer Datafolha exists, it is not the page opened here. The Sept. 24 Reuters piece is the poll source. Anything after it needs its own link.

The market page is a multi-outcome winner contract, not a first-round contract and not a runoff contract. The header read Flávio Bolsonaro 60.1%, Lula 40%, Renan Santos under 1%, Geraldo Alckmin under 1%. Other listed names, including Tarcísio de Freitas, Jair Bolsonaro, Fernando Haddad, and Michelle Bolsonaro, also sat under 1%, with Buy Yes prices around 0.1 cent on several of them. Volume on the event was $159,810,998. The page's market context says it opened Sept. 18, 2025, and lists an end date of Oct. 4, 2026. A FAQ on the same page called the leader "Flávio Bolsonaro" at 60%, and the next outcome Lula at 40%. The button said 60.1 cents. The FAQ said 60%. That is rounding on one page. Use 60.1% when you mean the Yes price that was on the button, and say the FAQ rounded it.

What Polymarket thinks

Polymarket's own experimental summary, timestamped on the page as updated Oct. 2, 2026, 12:45 a.m. UTC, says traders have Flávio ahead in a polarized race and points at late-September and early-October polls, including Datafolha, that it describes as a statistical tie or a narrow Lula lead in first-round and runoff scenarios. It offers reasons the market can still favor Flávio: right-wing consolidation, security-focused messaging, and the ineligibility of his father. It also says a runoff on Oct. 25 would decide the race if nobody wins a first-round majority. That Oct. 25 date is in Polymarket's market-context blurb. Reuters, in the story opened here, describes the runoff rule and does not print Oct. 25. Do not treat the blurb as a second official calendar.

The blurb is labeled experimental, not trading advice, and not an input into how the market resolves. Read it as color the site generated, then go back to the prices. The price says 60.1 cents for Flávio's Yes. The Sept. 24 poll says 47% to 45% in a hypothetical runoff and 40% to 36% in a first-round list. A person who averages 60.1 and 45 into "about 52" has built a fake number. The gap is the story. Markets and polls answer different questions: who traders will pay to hold until resolution, versus who respondents named to an interviewer over two days in late September.

Renan Santos is 3% in the Reuters first-round list and under 1% on the winner market. A winner contract crushes long shots a first-round poll still has to count. Do not swap those shares.

An October Fed contract and a Bitcoin level contract are different questions. They are not evidence about Lula or Flávio.

Why a favorite on the screen can lose the poll

A winner price above 50% means traders, in aggregate, will pay more than half a dollar for a share that pays a dollar if that person becomes president, under the market's rules. It does not mean a majority of voters named that person this week. The Brazil presidential election can have a first-round leader who is not the runoff favorite, and a runoff poll that is a technical tie, while a winner market leans to one side because of turnout guesses, coalition guesses, or who is even eligible. Reuters says Jair Bolsonaro is in prison and not the candidate. The market still lists his name under 1%. Eligibility is already inside the price of the people who are actually running.

The poll ended Sept. 23. These prices were on the page early Oct. 2. A 60.1% Yes can move before Oct. 4 without that poll being "wrong," because the contract pays on the eventual winner, not on a first-round plurality. Polymarket's blurb puts a runoff on Oct. 25. Confirm that date outside an experimental summary before you repeat it as a civic fact.

Volume is not wisdom, but it is a reason not to call this a ghost market. $159.8 million, which is how the page's FAQ rounded $159,810,998, is a lot of trading for an election contract. It is still not a sample of Brazilian voters. The Datafolha sample is 2,002 people and a stated margin. The market sample is whoever showed up with money. Both can be informative. Neither is the count on election night.

What to do as a reader (not a trade)

Write two lines and refuse to merge them. The Brazil presidential election does not fit in one percent. Line one: Datafolha, via Reuters on Sept. 24, runoff 47% Lula to 45% Flávio, margin 2 points, first round 40% to 36%, fieldwork Sept. 22–23, n=2,002. Line two: Polymarket winner market, page opened Oct. 2, Flávio Yes 60.1 cents, Lula Yes 40 cents, event volume $159,810,998, end date listed as Oct. 4, 2026. If a post says "Polymarket has Lula at 47," it has swapped the documents.

Say which percent you mean when the page itself rounds. The Brazil presidential election header and the Yes button showed 60.1%. The FAQ said 60%. That is not a fight worth a thread, and it is not permission to invent 61% or 63% because some other write-up used a range.

Do not trade the gap as a personality test. A market can lead a poll because it is early, because it is wrong, or because it is answering "who wins in the end" while the poll answered "who did you name today." The reader move is to keep the question attached to the number. First round, runoff, and overall winner are three questions. Datafolha, in the Reuters piece, answered the first two. The contract answered the third, at a later time. The Brazil presidential election will answer all three in the count, which neither a Sept. 24 poll nor an Oct. 2 price can do for you.