Brazil presidential election, second round 2026: the market before the vote, scored

What Polymarket priced before the Brazil presidential election, second round on 25 October 2026, frozen at instants fixed in advance and scored against the official result under the World Election Record rules.

By Vultax Research Team3 min read
Article details

AI-assisted drafting, research and editing. Published by OmniOS OÜ. How we use AI

MethodologyRead as Markdown
Read article
The low white Palácio do Planalto in Brasília under a deep blue sky, its curved columns lifting the roof over a lawn
The Palácio do Planalto in Brasília, the office of the President of Brazil, July 2020. Photo: Senado Federal / Wikimedia Commons, CC BY 2.0, cropped.

Price on the winner at F1

020¢

Luiz Inácio Lula da Silva, 24 hours before polls opened.

Brier score at F1
0.0001.282
0 is perfect; full confidence in a loser scores 2.
Favourite's price at F1
0.080.1¢
Flávio Bolsonaro. The favourite did not hold.
Live in VultaxOpen the Brazil election roomThe official result by state and city, and a replay of the count night. The record's frozen readings are separate.

The score

Luiz Inácio Lula da Silva won, by the provisional official count read from the Superior Electoral Court (TSE) at Mon 26 Oct, 02:36 UTC. At F1, 24 hours before polls opened, the market's favourite was Flávio Bolsonaro at 80.1¢; the favourite did not hold.

The Brier score at F1 is 1.282 and the price on the winner was 20¢. A perfect forecast scores 0; full confidence in a loser scores 2.

Published on 19 November 2026, later than the 72 hours after the provisional official count that rule 7 of the record sets. The freezes, the result source and the scores follow the hashed rules unchanged.

FreezePrice sourceFavouritePrice on winnerFavourite heldBrierLog loss
F7reconstructedFlávio Bolsonaro 80.9¢19.5¢no1.3021.635
F1capturedFlávio Bolsonaro 80.1¢20¢no1.2821.609

Brier score: sum over priced outcomes of (price minus result) squared, prices not rescaled. Log loss, the secondary measure: minus the natural log of the price on the winner, clipped at 1 and 99 cents. Rules v1, SHA-256 beginning df834706, hashed at 22:45 UTC on 19 October 2026.

Download table figures:CSVJSON with sources

What the market priced before the vote

The market scored is Polymarket's "brazil presidential election" event, which settles on the winner of the presidential election, second round included. It had traded $249 million by 19 November.

Each price is the midpoint of the best bid and the best offer on an outcome's Yes order book. (r) marks a price read from Polymarket's price history because the instant passed before the record existed. Outcomes that never reached one cent are left out of the table but stay in the dataset and the score.

OutcomeF7 · 18 Oct 11:00 UTCF1 · 24 Oct 11:00 UTC
Flávio Bolsonaro80.9¢ (r)80.1¢
Luiz Inácio Lula da Silva19.5¢ (r)20¢

Yes-book midpoints, Polymarket. Rules v1, SHA-256 beginning df834706, hashed at 22:45 UTC on 19 October 2026.

Download table figures:CSVJSON with sources

The freezes, by the clock

Polls open at 08:00 local time on 25 October in Brasilia, which is 11:00 UTC on 25 October. F7 is 168 hours before that and F1, the primary freeze, 24 hours before.

No pre-election poll blackout; polls may be published up to election day. Voting runs 08:00 to 17:00 Brasilia time nationwide.

The clock for Brazil presidential election, second round

  1. F7, seven days out

    Sun 18 Oct, 11:00 UTC

    08:00 in Brasilia

    Not recorded

    Passed before the record existed; read from price history and marked reconstructed.

  2. F1, one day out

    Sat 24 Oct, 11:00 UTC

    08:00 in Brasilia

    Recorded

    Order books captured at the instant.

  3. Polls open

    Sun 25 Oct, 11:00 UTC

    08:00 in Brasilia

    The event

    The instant every freeze is counted back from.

Times in UTC with the capital's local time. A freeze shown as missed has a reconstructed price; a scheduled one is captured by the server timer at that minute. The marks are not to scale.

The polls

Each polling firm's last estimate published before F1, 11:00 UTC on 24 October is scored against the official result: the error on the lead of the top two, and the mean error on every party or candidate above 5 percent. Firms count if their fieldwork ended in the 30 days before that cutoff.

The table of firms and their errors is added to this page as a dated update, read from each firm's own release or the outlet that commissioned it. The market scores above do not depend on it.

Most firms published their final runoff polls on the evening of 24 October, after F1. Under the record's rule those come after the cutoff, so each firm's previous estimate is the one scored here. The Brazil room on app.vultax.com sets those final polls beside the result.

The official result

Luiz Inácio Lula da Silva was elected with 64,256,232 valid votes, 51.79 percent, to 59,826,041, 48.21 percent, for Flávio Bolsonaro, with every section counted. The court marks him elected in its results file of 23:26 Brasília time on 25 October, 02:26 UTC on 26 October. Read from the Superior Electoral Court (TSE) at Mon 26 Oct, 02:36 UTC; the certified result is added when it is published.

Polymarket's payout on the winner market: settled on Luiz Inácio Lula da Silva, the official winner.

Check it

The rules, the schedule of instants and the capture code are hashed and published on the World Election Record page.

The dataset for this election, under CC BY 4.0, lists every outcome's price at each freeze, the receipt of the capture it was read from (file name, SHA-256, start and finish times), the official result and the scores. A price marked reconstructed can be checked against Polymarket's public price history.

Press noteUse Vultax research with an AI assistant

Sources and evidence

A pre-registered research record. Vultax Research enters no forecast of its own; prices are Polymarket order-book readings at stated instants, not advice. No other venue's prices are used.

Watch this measurement live in Vultax.

The official result by state and city, and a replay of the count night. The record's frozen readings are separate.

New studies, by email.

One email when a study publishes. Unsubscribe anytime. Privacy