September 22, 2026

1. Prediction market

In the outright winner contract on Polymarket, Flávio Bolsonaro stands at 59.55% (USD 11.50M) and Lula at 40.50% (USD 11.87M), in the reading confirmed on September 22 at 14:40 BRT. The gap is 19.05pp, against 22.40pp in the reading confirmed on September 21 at 17:01 BRT. Accumulated volume in the presidential book adds up to USD 153.62M across 19 contracts.

The two moves are not the same size, and the reason is arithmetic. The sum of the pair went from 99.40 to 100.05, so raw Lula rises 2.00pp and Flávio Bolsonaro gives up 1.35pp, and normalized each side moves exactly 1.75pp. The gap closes 3.35pp by the raw count and 3.49pp by the normalized one, and a move that survives normalization describes the race, not the book's spread.

Neither price is an extreme: Flávio Bolsonaro's stands 2.15pp below the series high, 61.70% recorded on September 21. And the two books moved in opposite directions on the same day: in the first-round second-place contract, Flávio Bolsonaro rises 1.00pp, to 69.50% (USD 1.06M), and Lula rises 0.40pp, to 29.80% (USD 0.89M).

In the third-place book, Renan Santos gave up 3.00pp, to 42.50% (USD 0.39M), the largest fall of any contract on the panel this round, and Augusto Cury gave up 0.80pp, to 39.45% (USD 0.27M). The distance between the two shrank for the second round running: 19.95pp, then 5.25pp, now 3.05pp. Ronaldo Caiado got no new price, at 15.50% (USD 0.16M), and keeps third place.

The contract on a Supreme Court justice leaving by impeachment before 2027 remains at 4.30% (USD 0.09M), with no new price, 21.80pp below the series high. In the Senate book this round publishes no new price.

2. What the polling firms registered

Quaest released at 18:06 BRT on September 21 the round registered as BR-06004/2026: 2,004 voters interviewed face to face between September 17 and 20, a margin of 2.0 points, commissioned by Grupo Globo. In the first round, Lula has 37% and Flávio Bolsonaro 33%, with Augusto Cury at 6%, Ronaldo Caiado at 4%, Renan Santos at 3% and Romeu Zema at 1%. In the main runoff pairing, Flávio Bolsonaro has 42% and Lula 41%, and the firm has Lula winning the other four pairings tested.

With it there are three national polls with fieldwork closing on the same September 20, by three methods: face to face at Quaest, telephone at BTG/Nexus and online at Palver. The distance between Lula and Flávio Bolsonaro in the first round comes out at 4 points in Lula's favor, 3 points in his favor and 1 point in Flávio Bolsonaro's favor, in that order.

The same round brought the government ruler, and its two questions moved in opposite directions. Quaest measured 44% approval against 50% disapproval, against 43% to 50% in the previous round, with the point inside the margin. On the rating question, measured in the same interview, the negative reading rose from 38% to 42% and the fair reading fell from 27% to 23%.

Polling calendar, next 7 days

Polls registered with the TSE scheduled for publication between September 23 and September 27. Being in the table does not mean publication is confirmed, because firms can delay or cancel. Filters: national scope and sample of at least 1,000. Each registration number links to the TSE public search.

Source: public TSE registry via the AFOS API. Samples of 3,000 or more in bold. The table lists the 9 live national polls in the window; 114 registrations were left out by scope or sample, and 1 national poll left the TSE registry, so it is not a commitment. The status is registered and not published. Real Time's national label for September 24 rests on a sampling plan the firm repeats in 12 of 12 of its state polls, so that scope is checked at release.

3. What the press covered

At the Supreme Court the day was about case assignment and petitions. Luiz Fux was drawn as rapporteur of a writ of mandamus asking the Senate for entry and exit records of Viviane Barci de Moraes and Daniel Vorcaro. The case was reassigned after Alexandre de Moraes, drawn earlier, declared himself disqualified because it involves his own wife.

André Mendonça's brief became public on September 22 and the act is from the 21st: on the last day of the deadline set by Edson Fachin, he filed 76 pages asking for an investigation into Moraes for abuse of authority and false accusation, and for summary dismissal of the inquiry against him, denying the charges. The request to revoke Vorcaro's arrest has been with Fachin since September 18, with no decision.

Outside the Supreme Court there was an electoral act. The Superior Electoral Court accepted the Lula campaign's petition pointing to a network of Instagram profiles boosting content about Flávio Bolsonaro through co-authored posts. The rapporteur only ordered the case to proceed and did not conclude that there was any irregularity, with five days for the defense to respond.

4. Divergences of the day

In summary

The gap gave back in one day part of what it had opened the day before, and the move is real. It was 3.35pp raw and 3.49pp normalized, and surviving normalization is the test that separates price from spread.

The third national poll from the same fieldwork did not resolve the disagreement, it widened the picture of it. Two firms out of three put Lula ahead in the first round, and the one that disagrees at the top is the same one that disagrees on the third way.

Nothing was decided at the Supreme Court on September 22. What happened was a rapporteur drawing and a petition filed the day before, and the impeachment contract got no new price.

---

Fontes consultadas

matérias com link direto para a notícia (veículos âncora):

• Metrópoles · Fux é sorteado relator de ação em caso que envolve mulher de Moraes
• Gazeta do Povo · Mendonça pede investigação de Moraes por abuso de autoridade e denunciação caluniosa
• Estadão · Mendonça responde Moraes, pede a Fachin arquivamento sumário de denúncia
• O Globo · Fux é sorteado relator de ação que pede registros de visitas ao Senado

matérias secundárias (URL Google News redirect, clique resolve à matéria):

• CNN Brasil · Quaest: Lula tem 37% no 1º turno; Flávio, 33%
• CNN Brasil · Quaest: Flávio tem 42% no 2º turno; Lula, 41%
• Brasil 247 · Quaest: governo Lula é aprovado por 44%
• Terra · Quaest: 42% avaliam governo Lula como negativo; 33% como positivo; 23% regular
• CartaCapital · TSE investiga rede suspeita de impulsionar conteúdos de Flávio Bolsonaro
• Poder360 · TSE investiga rede no Instagram suspeita de impulsionar Flávio Bolsonaro

Fontes técnicas: Polymarket (live quotes via the AFOS proxy, fetched Sep 22, 14:40 BRT), TSE registry (official electoral polls).