Summary

For 186 days AFOS recorded, in parallel, the price of the election winner contract on Polymarket, the national polls registered with the TSE and press coverage, in a daily edition and on a public dashboard. This report brings these series together and compares them with the official result of both rounds. It describes what each instrument showed; it does not rank the instruments.

1. The market switched favorites several times. Between Apr 22 and Oct 24, Flávio Bolsonaro closed ahead of Lula in the winner contract on 63 of 183 days. Lula's highest close was 66.50, on Aug 1; his lowest, 11.50, on Oct 9.
2. Up to the first round, runoff polls more often had Lula ahead. Of the 111 national polls with the Lula versus Flávio matchup published from Apr 22 to Oct 3, 62 had Lula ahead and 41 Flávio.
3. In the first round, the range of the polls contained the result, and the market got the order wrong. The 9 national polls on the eve ranged from Lula +5.1 to Flávio +4.0; the count gave Flávio +1.87. The 2nd-place contract paid 57.50 on Flávio; Lula came 2nd.
4. In the runoff (second round), market and polls pointed the same way, and the count went the other way. On the eve, the contract paid 83.40 on Flávio, and 9 of the 10 national polls had him ahead, in a range from Lula +0.5 to Flávio +3.0. The count gave Lula +3.58, outside the range. After the first round, 45 of the 49 national polls with the head-to-head in total votes had Flávio ahead or tied.
5. On runoff night, the price moved against the count. From 19:46 to 20:21, Flávio led the TSE count, peaking at 50.49%. In the same interval, Lula's price went from 67.20 to 89.10.
6. The money was contested, not owned. In September, a group of 4 wallets with signs of coordinated activity accounted for 43% of the net aggressor flow against Lula from Sep 12 to 21. In October, the flow switched sides three times and that group fell to 5.4% of the aggressor volume against Lula.

The Polymarket presidential book is of the negRisk type: the candidates are mutually exclusive, and buying a candidate's NO is equivalent to buying all the others. This allows paired positions (for example, Lula YES with Flávio NO) that reduce risk without changing the direction of the bet, which is why, in the flow studies, selling a lot does not mean betting against.

The contract opened the period almost tied and had three phases: Lula ahead from mid-May to September, with the highest close on Aug 1; Flávio ahead from Sep 10; and, after the first round, Flávio comfortably ahead, up to the eve of the runoff.

The largest one-day moves, and what that day's edition recorded

The right-hand column says what was in that day's edition, not the cause of the move. The order in time and the size are measured; the causal relationship is not.

Up to the first round, runoff polls had Lula ahead most of the time, while the market swung. After the first round, the picture turned almost uniformly: 45 of the 49 national polls with the head-to-head in total votes had Flávio ahead or tied, almost all by 1 to 5 points. Six other national polls from the period did not publish the pair in total votes and are left out of the figure.

The two instruments measure different things, and the difference has a known shape: a statistical tie in the polls turns into a high price for whoever is slightly ahead, because the contract pays all or nothing. In 2026, this showed up clearly after the first round. With the national polls showing Flávio 1 to 5 points ahead, the contract reached close to 89 on Flávio, on Oct 9, and ended the eve at 83.40.

Table from the October flow study (cutoff Oct 24, 08:00). Price on a 0 to 100 scale; poll in %. They are different objects; the table shows the distance, not a pricing error.

The reading the data support is that of two instruments that, in the runoff, pointed the same way with very different intensities: the polls, to a tight race with Flávio slightly ahead; the market, to Flávio as the clear favorite. The count gave Lula a margin of 3.58 points of valid votes, a margin that none of the 10 pollsters on the eve showed.

Voting, scheduled to end at 17:00, was extended until 19:00 by an injunction from Justice Cristiano Zanin of the STF, at the request of Lula's coalition, because of a lack of public transport in cities in Bahia. The TSE president, Nunes Marques, limited the extension to the polling stations that were still open; the TSE justices met behind closed doors; the Electoral Public Prosecutor's Office spoke out against it; and the TRE-DF said it had not been notified. The release of results only began after 19:00.

At the first reading, at 19:35, with 8% of polling stations, the contract was already paying 69.50 on Lula. From 19:46 to 20:21, Flávio led the count, peaking at 50.49% of valid votes with 21.57% of polling stations. In that interval Lula's price went from 67.20 to 89.10, with a low of 63.40 at 19:53. Lula moved ahead at 20:24, with 47.62% of polling stations, and the tabulation ended at 23:27, with 51.79%.

The next day, Flávio Bolsonaro did not recognize the result and announced that he will ask the TSE to revoke the winning ticket's mandate; the PL, his party, recognized the result in a statement, with reservations. Governments of more than 25 countries congratulated the winner; the United States government had made no official recognition by the night of Oct 26.

Throughout the cycle, AFOS carried out three studies on who bought and sold in the Lula and Flávio contracts, based on public data from the platform and the Polygon network. The yardstick common to all three: flow is not impact. Whoever trades the most is not necessarily whoever moves the price the most, and similar behavior across wallets is statistical evidence, not proof of coordination.

September: the turn toward Flávio (Aug 25 to Sep 21)

Lula's contract went from 0.625 to 0.385 and Flávio's from 0.351 to 0.610. A group of 4 wallets, with an overlap of markets that would hardly occur by chance (less than 1 in 1 million against a random control) and administrative operations in sequence between the accounts, was the largest identified net buyer: 17.0%, 48.6% and 43.3% of the net aggressor flow in the three windows. In the attributable part of the price move, the group explains 10% to 21%, 16% to 44% and 31% to 65%, depending on the method. The persistent price shocks came 2 to 10 hours after public polls in the same direction. The calculations were redone from scratch with code separate from the original: of the 41 verifiable claims, 28 reproduced exactly, 10 had their wording adjusted and 3 were corrected before the final version.

The largest position on Lula (October 2025 to October 2026)

A single wallet was, from November 2025 to July 2026, the largest holder of Lula YES. In May, a large-position tracker flagged a "whale" of about US$ 328 thousand on Lula, and the article (Bitcoin World) has four figures that match this wallet, without naming it. It bought for a year, almost only with limit orders, reached 1.06 million shares at a cost of US$ 505 thousand and, on its best day, Aug 1, was US$ 159 thousand above cost, without selling. On first-round night the contract fell from 0.37 to about 0.15. Four days later, on Oct 8, it sold everything at market in 41 seconds, at an average price of 0.126, and received US$ 133.6 thousand. With Lula's victory, those shares paid US$ 1 each, about US$ 1.06 million. The measured pattern is that of a high-conviction bettor who also earned rewards as a market maker; the data do not indicate advance information.

October: the flow switched sides three times (Oct 6 to 24)

Gross aggressor volume was 9.67 million shares against Lula and 10.65 million in favor; the balance, small next to the gross, was 978.7 thousand in favor of Lula. The September group stayed on Flávio's side, but fell to 5.4% of the aggressor volume against Lula, with 87% of its flow concentrated in the first six days. The balance in favor of Lula came from new wallets: of the 91 with a detailed position, 17 had their first activity from Oct 1 onward and added up to 1.06 million shares in favor of Lula. The largest buy in the window was 423.9 thousand shares of Lula YES, on Oct 15. The largest sell was the exit of the largest position on Lula, described above, which the study classifies as the unwinding of a paired position, not as a bet against Lula.

The press entered the series as a record of what was published each day, not as a forecast: 5,434 links in 187 editions. AFOS does not measure the accuracy of the press. What the series allows is a reading of sequence: in the flow studies, the persistent price shocks came after public polls in the same direction, and none could be confidently linked to information not yet released.

Two editorial rules stayed fixed during the cycle: a high-impact fact (arrest, court decision, testimony) only enters with two sources from different media groups, and the date of the fact comes from the body of the article, not from the publication timestamp.

• Price. Published only after two readings agree; the series of closes uses the last point recorded each day. Books with few trades can have a wide range within the day.
• Polls. National polls with the head-to-head in total votes; pollsters that published only valid votes are left out of the figures and are in the dataset with a note. Different polls do not form a series: methodology, base and margin vary.
• Result. Valid votes, from the final TSE file. Comparisons between poll totals and ballot-box valid votes are stated at each point.
• Wallets. Public data from the platform and the Polygon network. Coverage of runoff night trades is complete from 20:15 (Lula) and 20:26 (Flávio), partial before.
• Errata. Four figures published in the dataset were corrected during the cycle, all recorded in ERRATA.md, with the old and the new value.

Agreeing is not confirming. From April to September, market and polls frequently disagreed: runoff polls more often had Lula ahead, and the contract switched favorites several times. In the runoff, the two came to point the same way, and that is when they were wrong together. On the eve, 9 of 10 national polls had Flávio ahead and the contract paid 83.40 on him; the count gave Lula a margin of 3.58 points.

The two instruments were not independent. In the flow studies, the price shocks that lasted came 2 to 10 hours after public polls in the same direction. The market read the polls and amplified what they showed: a lead of 1 to 5 points became a price of 80 to 89. When one instrument feeds on the other, agreement between them adds little.

The width of the divergence is information. In the first round, the polls on the eve ranged from Lula +5.1 to Flávio +4.0, and that wide range contained the result. In the runoff, the range narrowed to Lula +0.5 to Flávio +3.0 and left the result outside. A narrow range gives more confidence, not more certainty.

The price is fast, not infallible. On counting night, the contract passed 50 on Lula before the first reading and reached 89.10 while Flávio still led the count. On the eve, the same contract paid 16.50 on Lula. It reacts quickly to what arrives, but it does not know beforehand.

Big money did not decide the price. The largest position on Lula sold everything after the first round, close to the bottom. The most active group of wallets in September lost weight in October. The flow switched sides three times. In none of the three studies did volume alone explain the move.

The data are in the AFOS-Analytics1/brazil-2026-electoral-divergence dataset, on Hugging Face: market and poll divergence, national polls and the TSE registry, official result and retrospective of both rounds, and the counting nights with anonymized trades.

1. TSE, presidential tabulation, elections 6257 (first round) and 6258 (runoff), resultados.tse.jus.br.
2. TSE, ballot box reports from the 20:00 bulletin of Oct 25, 2026.
3. Polymarket, Gamma API, Data API and CLOB; Polygon network.
4. Agência Brasil, "Ministros do TSE se reúnem a portas fechadas após decisão de Zanin", Oct 25, 2026.
5. Gazeta do Povo, "STF concede liminar e prorroga votação até as 19h em todo o Brasil", Oct 25, 2026.
6. Exame, "Por que Zanin, do STF, decidiu prorrogar o horário de votação até as 19h?", Oct 25, 2026.
7. Poder360, full text of Zanin's decision, Oct 25, 2026; "Líderes internacionais dão parabéns a Lula pela vitória", Oct 26, 2026.
8. Gazeta do Povo, "PL contraria Flávio e reconhece vitória de Lula", Oct 26, 2026; Exame, "Flávio Bolsonaro muda discurso e agora fala em pedido de cassação da chapa de Lula", Oct 26, 2026.
9. CNN Brasil, "Trump diz que eleição do Brasil pareceu fraudada, mas que gosta de Lula", Oct 26, 2026.
10. Bitcoin World, "Whale Places $328,000 Bet On Lula Winning Brazil's 2026 Election On Polymarket", around May 27, 2026 (offline).
11. The Intercept Brasil, audio recordings of Flávio Bolsonaro and Daniel Vorcaro, May 13, 2026.
12. AFOS, flow studies of September (data up to Oct 5) and October (cutoff Oct 24, 08:00), and wallet dossier (data up to Oct 15), internal use.

AFOS Analytics · closing report on Brazil's 2026 presidential election · Oct 26, 2026. This document is a descriptive analysis of public data. It is not an electoral poll, it is not investment advice and it does not claim coordination, manipulation or wrongdoing. Corrections: contact@afos-analytics.com.

The days after the count, October 25 to 31

Part I of this report is the text of Oct 26, 2026, unchanged (version 1.0, SHA-256 493fda3b…0ec14c0). This part records the extension of voting and what happened in the following six days: international recognition, the challenge to the result, the defeated candidate's departure from the country, the statements by Donald Trump and Elon Musk, coverage in the international press, what the markets began to price, the first readings by experts and a measurement of the vote by region.

The question of jurisdiction

Organizing the vote is the responsibility of the electoral courts, and the decision came from an STF justice, in an individual injunction, in the middle of election day. The point was raised by observers, the press and the parties, with differing assessments:

• OAS (preliminary report, Oct 27): "unprecedented situation"; the injunction and "the TSE's delay in ruling caused confusion"; the mission urged the parties involved to prevent a repeat and recommended defining when the STF may intervene in electoral matters.
• Carter Center (Oct 27): the last-minute extension created "a challenging operational environment", with contradictory decisions between the TSE and the STF.
• Gazeta do Povo (news report): the decision "increased legal uncertainty in the election", because of the partial closing of polling stations and the difficulty of applying it uniformly.
• Fernando Neisser, FGV Law School (ConJur, Oct 26): the difference in voting hours could perhaps serve as grounds for a legal challenge, but the margin of more than 4 million votes makes annulment unfeasible.
• Flávio Bolsonaro (letter, Oct 28): called the decision an "illegitimate and illegal action by Lula's former lawyer".
• Fox News (Oct 26): treated the extension as a "key irregularity" that fueled uncertainty about the result.

The measured facts are two: the decision changed voting hours in part of the country on election day itself, and the votes recorded after 17:00 added up to less than 1% of the final difference. Judgment on jurisdiction rests with the bodies and experts cited; AFOS records the debate.

The TSE began publishing a dashboard tracking international statements on the result, which scans foreign ministry notes and statements by heads of state every 15 minutes. On the afternoon of Oct 29, according to the court itself, there were 107 countries and 11 multilateral bodies; in the 11:45 scan on Oct 31, 112 countries, according to Poder360. A country with no record appears in gray on the map, which means absence from the database, not refusal. The dashboard warns that the classification is provisional.

The observation missions

The OAS Electoral Observation Mission, headed by José Miguel Insulza, with 88 observers in the runoff, stated in a preliminary report of Oct 27 that it "has no doubts about the integrity of the official data". On the extension of voting, see section 11. The Carter Center expert mission, headed by Jennie Lincoln, also issued a statement on Oct 27, with the sentence "respect for the will of the voters is fundamental".

As of Oct 31, no new lawsuit against the result had been filed with the TSE. The electoral judicial investigation lawsuits pending against the winning ticket predate the runoff.

The "mathematical impossibility" claim rests on the growth in Lula's votes between the rounds. The TSE files, reproduced in the annex to this report, show what changed: turnout rose by 2,960,239 votes, and abstention fell from 21.08% to 19.22% of the electorate. The claim was amplified on social media by Elon Musk (section 16) and, on Oct 29, by Jason Miller, a Trump ally.

On Oct 25, the Federal Police asked the STF for precautionary measures against Flávio Bolsonaro in an investigation into money laundering linked to betting companies (Pet 16.877/DF). Justice Alexandre de Moraes ordered the surrender of his passports within 24 hours and barred him from leaving the country. The deadline expired at 11:00 on Oct 27 without the surrender, and the Federal Police canceled his regular and diplomatic passports. (Agência Brasil, Poder360, Exame, CNN Brasil)

The decision was also contested: according to CNN Brasil, the measure drew criticism among the leadership of the Office of the Prosecutor General, and, according to Diário Carioca, Flávio's defense asked for Moraes to be removed from the case. On the other side, on Oct 27, federal deputies Lindbergh Farias, Rogério Correia and Ana Elisa (PT) asked for Flávio's pretrial detention and an Interpol alert (Poder360). As of Oct 31, none of these requests had been ruled on.

The "rescue"

On social media, supporters called the departure a "rescue". What is confirmed is the departure through Argentina and the entry into the US with the authorization of the American government. Reports that American agents occupied part of the Puerto Iguazú airport and that Santiago Caputo accompanied the operation come from anonymous sources (Clarín, La Nación) and are not confirmed; Rubio denies involvement. A video of military personnel on a plane, captioned as the "rescue", shows behind-the-scenes footage of a television series (Agência Lupa, Oct 30); another, old video was used in the same way (Aos Fatos). A request for or grant of asylum was not confirmed.

In the US Congress, 25 Democratic House members sent a letter to Rubio, led by Sydney Kamlager-Dove and Dina Titus, asking for a reply by Nov 13 on the use of diplomatic and immigration resources in Flávio's departure (La Nación, MS NOW citing Politico). According to Common Dreams, members of the Progressive Caucus demanded on Oct 30 that the US "join the rest of the world in accepting the results". No statements by Republican lawmakers on the subject were found in the period, nor posts by Trump on Truth Social, nor statements by the White House press secretary. As of Oct 31, there was no record of a phone call between Trump and Lula. According to Folha, an internal cable from the US embassy in Brasília described Lula's victory as surprising but definitive; the document was not published.

Owner of X, the platform where the fraud claim circulated the most, Elon Musk commented on the result on Oct 26, in a set of short posts. Between Oct 27 and 31 no posts by him about Flávio's departure from the country were found.

Outlets with different editorial lines redid the arithmetic of the main post and reached the same conclusion: the votes of the eliminated candidates check out, but the calculation leaves out the roughly 4.8 million additional valid votes in the runoff (Poder360, Correio Braziliense, Crusoé, Brasil de Fato). It is the same calculation as in the box in section 13. No fact-checks by agencies nor X community notes on the posts were found, nor any response from the TSE, the STF or the government to Musk.

In the international press, the posts appeared in outlets such as NTD ("Trump, Musk Question Brazil's Election Result", Oct 26), Semana (Colombia) and Heraldo USA; in the major newspapers and agencies searched, Musk was not the subject of a dedicated article in the period.

Context. On Sep 20, Musk reacted with "Wow" to a Polymarket chart that gave 58% to Flávio (Poder360, Revista Oeste). In 2024, X was blocked in Brazil from Aug 31 to Oct 8 by a decision of Moraes, for failing to appoint a legal representative in the country, and returned after paying fines and complying with the orders. On Mar 10, 2026, Moraes closed, at the request of the PGR (Office of the Prosecutor General), the investigation opened against Musk in 2024 (according to Revista Oeste).

Coverage fell into three moments: the result (Oct 25 and 26), the challenge and Flávio's departure (Oct 27 to 29), and the role of the US and Argentine governments (Oct 28 to 31). The table separates news (N) from opinion (OP), and includes outlets of different leanings.

Several of these outlets block automated reading. In those cases the headline and framing come from reproductions (agencies, Brazilian press, the outlets' official social media accounts), and the table does not cite any passage that was not read in the original. The headlines of the Washington Post and Wall Street Journal editorials come from the newspapers' own social media accounts and from reproductions.

The winner contract was resolved for Lula at 00:12 on Oct 26 (Brasília time), automatically and without dispute, with US$ 249.1 million traded in the book. From Oct 26, Polymarket opened contracts on what would come next.

Single reading of Oct 31, 2026, 20:18 (Brasília), through the AFOS route to the public Polymarket API. Small books: with volumes of this order, the price moves with few orders. A contract price is not a probability.

In the international press, the result reopened the debate on the use of these markets as an electoral gauge: CNBC recalled that Kalshi and Polymarket gave more than 80% to Flávio (Oct 26), Futurism called the miss "comically wrong" (Oct 26) and Semafor noted that the markets got three elections wrong in the same week, in Brazil, Israel and Serbia (Oct 30). The AFOS reading on the subject is in section 9 of Part I.

Technology as a target. Jesús Delgado Valery, executive director of Transparencia Electoral, wrote in a LinkedIn post on Oct 30, at 10:28, that the Brazilian case confirms a regional pattern, "from both the left and the right": the challenge exploits the public's limited familiarity with the transmission and tallying of votes, confuses error with fraud and takes advantage of an asymmetry of time: "sowing doubts about an algorithm takes seconds. Refuting it requires weeks of expert reports". For him, "well-designed and auditable technology is not the instrument of fraud, but the best shield against it". The argument picks up his article in Latinoamérica21, of Jun 14, 2026, on Colombia.

The geography of the vote. Federico Bartalucci, a PhD candidate in economic geography at the University of Cambridge, wrote in a LinkedIn post on Oct 29, at 14:56, that "Brazil votes by region first" and that the 2026 change relative to 2022 occurred between cities and outlying areas: according to him, Lula's vote rose 1.7 points in the urban municipalities of the IBGE typology and fell by 1.2 to 3.1 points in the intermediate and rural categories. He places the reading in the literature on the Lula vote (Zucco, 2008) and on the urban and rural divide (Rodríguez-Pose, 2018). The regional figures he cites match the measurement in section 20; the municipal comparison with 2022 is his and was not redone by AFOS.

The judicial route. Fernando Neisser, a professor at FGV Law School, assessed in ConJur, on Oct 26, that the difference in voting hours could perhaps serve as grounds, but that the margin of more than 4 million votes makes annulment unfeasible: "I see no chance of success".

Valid votes in the runoff, final TSE tabulation files by federative unit (election 6258), summed by region. The sum of the 27 states and abroad is exactly the official national total: 64,256,232 votes for Lula and 59,826,041 for Flávio Bolsonaro.

The North ended practically tied: a difference of 1,079 votes out of more than 10 million. Lula won in 13 of the 27 federative units, with his highest percentage in Piauí (75.88%); Flávio had his highest in Roraima (72.46%).

Three points from Part I gain a new source or measurement. The extension of voting, described on Oct 26 as decisions with different scopes at the STF and the TSE, was called an "unprecedented situation" by the OAS mission, which recommended setting limits on the STF's intervention in electoral matters; the votes after 17:00 added up to less than 1% of the final difference. The fraud claim that followed, amplified by Trump and Musk, can be checked against the TSE's public files, whose arithmetic is in section 13, and the defeated campaign's own coordinator said he had no evidence. And the market, which on the eve paid 83.40 on the defeated candidate, came to pay 4.15 on an annulment of the result.

Sources for Part II

Each link below was read for this report. The time of the LinkedIn posts comes from each publication's identifier, which records the moment it was made. In the TSE file by state, replace "sp" with the abbreviation of the federative unit.

1. TSE, "Eleições 2026: painel registra manifestações e saudações de mais de cem países após 2º turno", Oct 29, 2026. www.tse.jus.br/comunicacao/noticias/2026/Outubro/eleicoes-2026-painel-registra-manifestacoes-e-saudacoes-de-mais-de-cem-paises-apos-2o-turno
2. Poder360, "112 países reconheceram eleições no Brasil, diz mapa do TSE", Oct 31, 2026. www.poder360.com.br/poder-eleicoes-2026/112-paises-reconheceram-eleicoes-no-brasil-diz-mapa-do-tse/
3. AP (Boston Globe), Trump's statement, Oct 26, 2026. www.bostonglobe.com/2026/10/26/world/brazil-lula-leftist-denies-trump-ally/
4. Reuters (Irish Times), Oct 27, 2026. www.irishtimes.com/world/americas/2026/10/27/trump-says-brazil-election-looked-rigged-as-bolsonaros-party-recognises-lula-win/
5. Metrópoles, "Trump diz que vai falar com Lula e chama eleição no Brasil de desgraça", Oct 29, 2026. www.metropoles.com/mundo/trump-diz-que-vai-falar-com-lula-e-chama-eleicao-no-brasil-de-desgraca
6. Exame, Milei government "tomou nota", Oct 27, 2026. exame.com/mundo/governo-milei-diz-que-tomou-nota-da-vitoria-de-lula-mas-nao-comentara-resultado-eleitoral/
7. CNN Brasil, Milei government will not comment, Oct 27, 2026. www.cnnbrasil.com.br/blogs/luciana-taddeo/internacional/governo-milei-nao-vai-se-pronunciar-sobre-vitoria-do-lula-diz-porta-voz/
8. Los Andes, Argentina will not congratulate Lula, Oct 27, 2026. www.losandes.com.ar/politica/argentina-no-felicitara-oficialmente-lula-su-reeleccion-brasil-el-presidente-no-es-un-hipocrita-n6011910
9. Gazeta do Povo, Bukele congratulates Lula, Oct 28, 2026. www.gazetadopovo.com.br/mundo/referencia-de-flavio-na-seguranca-publica-bukele-parabeniza-lula-pela-reeleicao-no-brasil/
10. OAS, Electoral Observation Mission, preliminary report on the runoff, Oct 27, 2026. www.oas.org/fpdb/press/2026_MOE_Brasil_Segundo_Turno_Relatorio_Preliminar_POR.pdf
11. Poder360, "OEA valida vitória de Lula, mas alerta para decisões monocráticas do STF", Oct 29, 2026. www.poder360.com.br/poder-eleicoes-2026/oea-valida-vitoria-de-lula-mas-alerta-para-decisoes-monocraticas-do-stf/
12. Carter Center, statement by the expert mission, Oct 27, 2026. www.cartercenter.org/news/carter-center-election-expert-mission-commends-brazils-democratic-presidential-runoff-election/
13. Gazeta do Povo, Flávio drops annulment and will seek revocation of the ticket's mandate, Oct 26, 2026. www.gazetadopovo.com.br/republica/flavio-desiste-anular-eleicao-pedira-cassacao-chapa-lula/
14. CNN Brasil, Flávio drops challenge to the victory at the ballot box, Oct 26, 2026. www.cnnbrasil.com.br/eleicoes/flavio-desiste-de-questionar-vitoria-de-lula-nas-urnas/
15. Metrópoles, Valdemar recognizes Lula's victory, Oct 26, 2026. www.metropoles.com/brasil/presidente-do-pl-valdemar-costa-neto-reconhece-vitoria-de-lula
16. Gazeta do Povo, Tarcísio recognizes the result and defends a vigilant opposition, Oct 27, 2026. www.gazetadopovo.com.br/republica/tarcisio-reconhece-vitoria-de-lula-e-defende-oposicao-vigilante/
17. Poder360, Alcolumbre and Motta recognize Lula's victory. www.poder360.com.br/poder-congresso/alcolumbre-e-motta-reconhecem-vitoria-de-lula-mas-fazem-cobrancas/
18. Metrópoles, "Flávio diz que Zanin deu 10 milhões de votos a Lula", Oct 28, 2026. www.metropoles.com/mundo/flavio-diz-que-zanin-deu-10-milhoes-de-votos-a-lula
19. CNN Brasil, Flávio backs down in letter and calls himself "líder do Brasil", Oct 28, 2026. www.cnnbrasil.com.br/politica/flavio-recua-em-carta-com-teor-golpista-e-se-diz-lider-do-brasil/
20. CNN Brasil, Marinho denies there is evidence of fraud, Oct 28, 2026. www.cnnbrasil.com.br/politica/apos-carta-de-flavio-marinho-nega-que-haja-provas-de-fraude-nas-eleicoes/
21. Metrópoles, opposition leader says PL has no evidence, Oct 28, 2026. www.metropoles.com/brasil/flavio-fala-em-fraude-mas-lider-da-oposicao-diz-que-pl-nao-tem-provas
22. Gazeta do Povo, Nunes Marques: "absoluta normalidade", Oct 28, 2026. www.gazetadopovo.com.br/republica/apos-carta-de-flavio-nunes-marques-diz-que-eleicoes-ocorreram-em-absoluta-normalidade/
23. CNN Brasil, TSE says anyone can audit, Oct 28, 2026. www.cnnbrasil.com.br/eleicoes/apos-flavio-alegar-fraude-tse-diz-que-qualquer-pessoa-pode-auditar-urnas/
24. Agência Lupa, fact-check of Flávio's letter, Oct 28, 2026. www.agencialupa.org/checagem/2026/10/28/apos-fuga-do-brasil-flavio-bolsonaro-mente-sobre-eleicoes-em-carta-com-teor-golpista/
25. Poder360, "Flávio usa teoria imprecisa para chamar vitória de Lula de fraude", Oct 28, 2026. www.poder360.com.br/poder-eleicoes-2026/flavio-usa-teoria-imprecisa-para-chamar-vitoria-de-lula-de-fraude/
26. Band, Flávio's livestream in Miami, Oct 30, 2026. www.band.com.br/politica/sem-provas-flavio-volta-a-falar-em-fraude-e-diz-que-saiu-do-brasil-por-censura
27. Poder360, Michelle: "momento é de união", Oct 30, 2026. www.poder360.com.br/poder-eleicoes-2026/michelle-vira-porta-voz-de-bolsonaro-e-diz-que-momento-e-de-uniao/
28. Exame, Moraes orders surrender of passports within 24 hours, Oct 26, 2026. exame.com/brasil/moraes-determina-que-flavio-entregue-passaportes-em-24-horas-e-proibe-contato-com-willer-tomaz/
29. CNN Brasil, what is known about Flávio's departure from Brazil, Oct 27, 2026. www.cnnbrasil.com.br/politica/o-que-se-sabe-sobre-a-saida-de-flavio-bolsonaro-do-brasil/
30. Poder360, Moraes's decision did not authorize arrest, says Federal Police. www.poder360.com.br/poder-justica/decisao-de-moraes-nao-autorizava-prisao-de-flavio-diz-pf/
31. Poder360, PT asks for Flávio's arrest and an Interpol red notice, Oct 27, 2026. www.poder360.com.br/poder-congresso/pt-pede-prisao-de-flavio-bolsonaro-e-alerta-vermelho-da-interpol/
32. CNN Brasil, Moraes's decision draws criticism among PGR leadership. www.cnnbrasil.com.br/blogs/taina-falcao/politica/decisao-de-moraes-contra-flavio-gera-criticas-na-cupula-da-pgr/
33. Infobae, departure through Puerto Iguazú, Oct 27, 2026. www.infobae.com/america/america-latina/2026/10/27/flavio-bolsonaro-habria-salido-de-brasil-y-viajaria-a-estados-unidos-en-un-vuelo-privado-desde-argentina/
34. Chequeado, what is known about the crossing, Oct 28, 2026. chequeado.com/el-explicador/flavio-bolsonaro-se-fugo-de-brasil-por-iguazu-que-se-sabe-sobre-el-cruce-fronterizo-y-las-dudas-que-persisten/
35. Metrópoles, Trump administration confirms visas, Oct 27, 2026. www.metropoles.com/mundo/governo-trump-confirma-emissao-de-vistos-para-flavio-bolsonaro-e-familia
36. CNN Brasil, arrival in Miami, Oct 28, 2026. www.cnnbrasil.com.br/internacional/flavio-bolsonaro-chega-com-familia-a-miami-apos-fuga-do-brasil/
37. NBC News (Reuters, AP), arrival in Miami, Oct 28, 2026. www.nbcnews.com/world/brazil/flavio-bolsonaro-arrives-miami-election-defeat-rcna606099
38. AP (ABC News), Rubio denies having facilitated the trip, Oct 31, 2026. abcnews.com/Politics/wireStory/rubio-us-government-facilitate-bolsonaros-flight-us-137740656
39. Agência Lupa, video does not show Flávio's rescue, Oct 30, 2026. www.agencialupa.org/verificacao/2026/10/30/video-de-militares-dos-eua-gargalhando-nao-mostra-resgate-de-flavio-bolsonaro/
40. Aos Fatos, old video used to suggest a trip to the US. www.aosfatos.org/noticias/video-antigo-circula-nas-redes-para-sugerir-que-flavio-foi-para-os-eua/
41. Polymarket, contract "Brazil presidential election annulled?". polymarket.com/event/brazil-presidential-election-annulled-by-december-31-20261026114556736
42. CNBC, prediction markets and the election, Oct 26, 2026. www.cnbc.com/2026/10/26/lula-brazil-election-prediction-markets.html
43. Futurism (Yahoo Finance), Oct 26, 2026. finance.yahoo.com/markets/options/articles/prediction-markets-were-comically-wrong-205203851.html
44. Semafor, markets miss recent elections, Oct 30, 2026. semafor.com/article/10/30/2026/prediction-markets-flop-on-recent-elections
45. Latinoamérica21, Jesús Delgado Valery, Jun 14, 2026. latinoamerica21.com/pt-br/a-narrativa-de-fraude-e-seu-bode-expiatorio-a-tecnologia/
46. Zucco (2008), Journal of Latin American Studies 40(1). doi.org/10.1017/S0022216X07003628
47. Rodríguez-Pose (2018), Cambridge Journal of Regions, Economy and Society 11(1). doi.org/10.1093/cjres/rsx024
48. ConJur, annulment of the runoff is unfeasible, Oct 26, 2026. conjur.com.br/2026-out-26/judicializacao-para-anulacao-de-resultado-geral-do-segundo-turno-e-inviavel/
49. TSE, runoff tabulation by federative unit, election 6258. resultados.tse.jus.br/oficial/ele2026/6258/dados/sp/sp-c0001-e006258-u.json
50. Euronews, Trump says election looked rigged, Oct 26, 2026. www.euronews.com/2026/10/26/trump-says-brazil-election-looked-rigged-without-giving-evidence
51. Al Jazeera, PL accepts result after Trump's remarks, Oct 27, 2026. www.aljazeera.com/news/2026/10/27/bolsonaros-party-accepts-result-after-trump-suggests-vote-was
52. Bloomberg, "Trump Says Brazil Election Looked Rigged", Oct 26, 2026. www.bloomberg.com/news/articles/2026-10-26/trump-says-brazil-election-looked-rigged-after-lula-victory
53. CNBC, Trump would approve Flávio's entry, Oct 28, 2026. www.cnbc.com/2026/10/28/trump-bolsonaro-us-brazil-election.html
54. CBS News, visas for Flávio Bolsonaro. www.cbsnews.com/news/brazil-flavio-bolsonaro-u-s-visa-presidential-election-loss/
55. Al Jazeera, Trump calls election a "disgrace", Oct 29, 2026. www.aljazeera.com/news/2026/10/29/trump-says-wont-take-sides-in-brazil-election-but-calls-it-a-disgrace
56. AFP (Manila Times), Trump hardens stance, Oct 30, 2026. www.manilatimes.net/2026/10/30/world/trump-hardens-stance-on-brazil-election-disgrace/2457682
57. Reuters (US News), US does not intend to back Flávio's claim, Oct 30, 2026. www.usnews.com/news/world/articles/2026-10-30/trump-not-planning-to-back-bolsonaros-claim-to-be-brazils-leader-sources-say
58. AP (WJTV), Rubio, Oct 31, 2026. www.wjtv.com/washington/washington-dc/ap-rubio-says-the-us-government-did-not-facilitate-bolsonaros-flight-to-the-us/
59. La Nación, Democrats question Rubio, Oct 30, 2026. lanacion.com.ar/estados-unidos/democratas-indagan-si-funcionarios-de-trump-en-la-argentina-ayudaron-en-la-fuga-de-flavio-bolsonaro-nid30102026
60. Common Dreams, Progressive Caucus, Oct 30, 2026. www.commondreams.org/news/cpc-trump-brazil-election
61. Brasil de Fato (reproducing Folha), US embassy cable, Oct 30, 2026. www.brasildefato.com.br/2026/10/30/embaixada-dos-eua-em-brasilia-classifica-vitoria-de-lula-como-definitiva/
62. Washington Post, "Lula wins reelection in Brazil", Oct 25, 2026. www.washingtonpost.com/world/2026/10/25/lula-wins-reelection-brazil-halting-latin-americas-rightward-shift-now/
63. Washington Post, editorial, Oct 28, 2026. www.washingtonpost.com/opinions/2026/10/28/brazils-treatment-bolsonaro-undermines-rule-of-law-confidence/
64. WSJ Opinion, editorial, Oct 28, 2026 (the newspaper's post). x.com/WSJopinion/status/2115644810539278391
65. Fox News, extension as an irregularity, Oct 26, 2026. www.foxnews.com/politics/brazilian-election-sends-us-progressives-online-fever-key-irregularity-stokes-result-uncertainty
66. Infobae (EFE), the role of the Milei government, Oct 28, 2026. www.infobae.com/america/agencias/2026/10/28/el-papel-del-gobierno-de-milei-en-la-fuga-de-flavio-bolsonaro-a-traves-de-argentina/
67. Agência Brasil, reaction in the international press, Oct 25, 2026. agenciabrasil.ebc.com.br/politica/noticia/2026-10/imprensa-internacional-destaca-vitoria-historica-de-lula
68. Diario de Pernambuco, "Imprensa mundial classifica vitória de Lula como dramática reviravolta e surpreendente", October 2026. www.diariodepernambuco.com.br/politica/2026/10/11727659-imprensa-mundial-classifica-vitoria-de-lula-como-dramatica-reviravolta-e-surpreendente.html
69. Metrópoles, Musk questions the numbers of Lula's reelection, Oct 26, 2026. www.metropoles.com/mundo/elon-musk-questiona-numeros-da-reeleicao-de-lula-matematica-nao-fecha
70. Revista Oeste, Musk questions the election result, Oct 26, 2026. revistaoeste.com/mundo/elon-musk-questiona-resultado-da-eleicao-no-brasil/
71. Correio Braziliense, Musk questions the election with an incomplete post, Oct 26, 2026. www.correiobraziliense.com.br/politica/2026/10/7528018-musk-questiona-eleicao-de-lula-com-post-incompleto-sobre-votacao.html
72. Poder360, Musk: "matemática não fecha", Oct 26, 2026. www.poder360.com.br/poder-eleicoes-2026/musk-questiona-das-eleicoes-no-brasil-matematica-nao-fecha/
73. Jornal Razão, reach of Musk's post, Oct 26, 2026. jornalrazao.com/politica/elon-musk-conta-nao-fecha-10-milhoes-votos-lula-2-turno
74. O Antagonista/Crusoé, Musk's reaction to the numbers, Oct 26, 2026. oantagonista.com.br/mundo/crusoe-a-reacao-de-musk-aos-numeros-da-vitoria-de-lula/
75. ICL Notícias, Musk questions Lula's victory, Oct 26, 2026. iclnoticias.com.br/musk-questiona-vitoria-de-lula-e-casa-branca-mantem-silencio/
76. Brasil de Fato, Bannon and Musk, Oct 27, 2026. www.brasildefato.com.br/2026/10/27/sem-provas-bannon-cobra-de-trump-isolamento-do-brasil-ate-anular-eleicao-musk-ecoa-narrativa-de-fraude/
77. NTD, "Trump, Musk Question Brazil’s Election Result", Oct 26, 2026. www.ntd.com/trump-musk-question-brazils-election-result_1181591.html
78. Semana (Colombia), Musk's reaction to Lula's win. www.semana.com/mundo/articulo/elon-musk-reacciona-sorpresivamente-al-triunfo-de-lula-en-brasil-las-cifras-no-encajan/202608/
79. Poder360, Musk reacts with "Wow" to Flávio's rise on Polymarket, Sep 20, 2026. www.poder360.com.br/poder-eleicoes-2026/musk-reage-com-wow-a-subida-de-flavio-na-polymarket/
80. Revista Oeste, Moraes closes investigation against Musk, Mar 10, 2026. revistaoeste.com/politica/moraes-atende-pgr-e-arquiva-inquerito-contra-elon-musk/
81. Wikipedia, blocking of X in Brazil (2024). en.wikipedia.org/wiki/Blocking_of_X_in_Brazil
82. Migalhas, Zanin extends voting until 19:00 after complaint about transport, Oct 25, 2026. www.migalhas.com.br/quentes/467195/zanin-prorroga-votacao-ate-as-19h-apos-reclamacao-sobre-transporte
83. Poder360, full text of Zanin's decision, Oct 25, 2026. www.poder360.com.br/integras/leia-a-integra-da-decisao-de-zanin-que-prorroga-votacao-ate-as-19h/
84. Gazeta do Povo, STF grants injunction and extends voting until 19:00, Oct 25, 2026. www.gazetadopovo.com.br/eleicoes/2026/stf-concede-liminar-e-prorroga-votacao-ate-as-19h-em-todo-o-brasil/
85. Gazeta do Povo, Zanin's decision increased legal uncertainty. gazetadopovo.com.br/eleicoes/2026/decisao-de-ultima-hora-de-zanin-ampliou-inseguranca-juridica-na-eleicao
86. Agência Brasil, TSE justices meet behind closed doors, Oct 25, 2026. agenciabrasil.ebc.com.br/justica/noticia/2026-10/ministros-do-tse-se-reunem-a-portas-fechadas-apos-decisao-de-zanin
87. Exame, why voting was extended until 19:00, Oct 25, 2026. exame.com/brasil/por-que-o-tse-decidiu-prorrogar-o-horario-de-votacao-ate-as-19h/
88. Jesús Delgado Valery, LinkedIn post, Oct 30, 2026, 10:28. linkedin.com/posts/jesus-delgado-valery
89. Federico Bartalucci, LinkedIn post, Oct 29, 2026, 14:56. linkedin.com/posts/federico-bartalucci-129799130

The official result of both rounds

Figures from the final TSE tabulation file and from TSE documents on ballot boxes, the same as in the AFOS dashboard cards and the dataset (results/ folder). Candidate percentage over valid votes; turnout and abstention over the electorate; blank and null votes over the votes counted.

Who won

Luiz Inácio Lula da Silva (PT), Brasil Pronto pra Mais coalition, was elected president in the runoff, on Oct 25, 2026, with 64,256,232 votes, 51.79% of valid votes. Flávio Bolsonaro (PL) received 59,826,041, 48.21%. Difference: 4,430,191 votes, 3.58 points. In the first round, Flávio Bolsonaro had come out ahead, with 47.03% against 45.16%.

The two rounds side by side

From the first round to the runoff, turnout rose by 2,960,239 votes (1.86 points) and blank and null votes combined fell from 4.77% to 3.24%. Replaced ballot boxes: percentage over the total of ballot boxes, in use plus contingency, the same base as the TSE. The TSE uses the term "replaced" and does not publish a count of "defective" ballot boxes. Runoff: incident report up to 20:00 on Oct 25.

Runoff, Oct 25, 2026

First round, Oct 4, 2026

Sources: TSE, tabulation files https://resultados.tse.jus.br/oficial/ele2026/6257/dados/br/br-c0001-e006257-u.json and https://resultados.tse.jus.br/oficial/ele2026/6258/dados/br/br-c0001-e006258-u.json; TSE, Números gerais, 21:00 press conference, Oct 4, 2026; TSE, Relatório de Ocorrências de Urnas Substituídas e Votação Manual, 5th Bulletin of the runoff, incidents up to 20:00 on Oct 25, 2026.