AFOS Daily · Daily Synthesis
September 21, 2026
Prediction Markets × Polls × News
Synthesis generated from auditable data. Every claim cites its source.
The outright winner contract opened the widest gap in the 158 days of recorded series on a day when two national polls came out with opposite signs in both rounds. Flávio Bolsonaro rose 1.85pp and Lula gave up 3.00pp between the 00:06 and the 17:01 readings, and the gap went from 17.55pp to 22.40pp, in a presidential book holding USD 152.91M. The two national polls closed fieldwork on the same September 20 and disagree on everything: on the main pairing, on the order of the third way and on the sign of the pairings against Ronaldo Caiado and Romeu Zema.
1. Prediction market
The presidential outright winner contract on Polymarket closed the 17:01 certified reading with Flávio Bolsonaro at 60.90% (USD 11.37M accumulated) and Lula at 38.50% (USD 11.66M accumulated), a gap of 22.40pp. Total volume in the presidential book adds up to USD 152.91M. Against the 00:06 certified reading the same day, Flávio rose 1.85pp and Lula gave up 3.00pp, and the gap went from 17.55pp to 22.40pp.
In the readings recorded through the day the move was larger than the certified reading shows, and the series of extremes is in this edition's closing section.
In the first round second-place contract the movement ran the other way: Flávio fell 2.00pp, to 68.50% (USD 1.05M), and Lula gave up 0.15pp, to 29.40% (USD 0.88M). It is the only book in which Flávio lost price on the day.
In the first-round third-place book, Augusto Cury rose 7.70pp, to 40.25% (USD 0.25M) and Renan Santos fell 7.00pp, to 45.50% (USD 0.36M), the largest rise and the largest drop on the panel. Ronaldo Caiado rose 4.00pp, to 15.50% (USD 0.14M), and Romeu Zema gave up 0.55pp, to 0.50%. Renan Santos has already been paid 71.50% there on July 13.
The contract on a Supreme Court justice leaving by impeachment before 2027 stands at 4.30% (USD 0.09M), 0.10pp below the 00:06 reading, in a thin book where a tenth does not separate news from noise. In the Senate, the Liberal Party stays at 83.50% (USD 0.26M) and the MDB gave up 1.25pp, to 12.35%.
2. What the polling firms registered
BTG/Nexus, protocol BR-00485/2026, interviewed 2,006 voters by phone between September 18 and 20, margin of 2.0pp, and puts Lula at 40% and Flávio Bolsonaro at 37% in the first round, with Cury at 6%, Caiado at 5%, Renan Santos at 3% and Zema at 1%. In the runoff it tested five pairings and Lula wins or ties all five, according to UOL and Opera Mundi.
Palver, protocol BR-00860/2026, interviewed 5,000 voters online between September 15 and 20, margin of 2.8pp, and puts Flávio Bolsonaro at 42% and Lula at 41%, with Renan Santos third, at 11%, and Cury fourth, at 2%. In the runoff it also tested five pairings, and CNN Brasil records the result: Lula loses to Flávio 43% to 47%, to Romeu Zema 43% to 44% and to Ronaldo Caiado 43% to 44%, and beats Augusto Cury 43% to 35% and Renan Santos 43% to 34%.
Both closed fieldwork on the same September 20, which the TSE registration confirms. BTG/Nexus is phone with a probability sample; Palver is online through social media ads, with a non-probability sample adjusted by quotas and weights, according to its own sampling plan filed with the TSE.
On government approval, BTG/Nexus measures 46% approval against 50% disapproval, according to Metrópoles.
Polling calendar, next 7 days
Polls registered with the TSE scheduled for publication between Sep 23 and Sep 26. Inclusion in the table does not mean confirmed publication: firms may delay or cancel. Filters: national scope and sample equal to or greater than 1,000, with each protocol linked to the TSE public search.
| Date | Firm | Sample | Scope | TSE protocol | Conf. |
|---|---|---|---|---|---|
| Sep 23 | AtlasIntel | 5,000 | national | BR-04739/2026 | 0.9 |
| Sep 24 | Palver | 5,000 | national | BR-09587/2026 | 0.7 |
| Sep 24 | PoderData | 3,000 | national | BR-01739/2026 | 0.7 |
| Sep 24 | Alfa Inteligência | 2,700 | national | BR-02512/2026 | 0.7 |
| Sep 24 | Datafolha | 2,002 | national | BR-00304/2026 | 0.9 |
| Sep 24 | Real Time Big Data | 2,000 | national not sustained | BR-04202/2026 | 0.9 |
| Sep 24 | 100 Cidades | 2,000 | national | BR-05268/2026 | 0.7 |
| Sep 26 | Veritá | 40,500 | national | BR-04467/2026 | 0.7 |
Source: public TSE registry via the AFOS API. Samples equal to or greater than 3,000 in bold. The table lists the 8 national and live registrations in the window; 104 registrations from the period were left out by scope or by sample, and 1 national left the TSE registry, so it is not a commitment. The status is registered and not published. Real Time Big Data is flagged as scope not sustained: its national label comes from the sampling plan, and at that firm the plan says national in 13 of 13 registrations that the methodology resolved as state-level.
3. What the press covered
On September 21 the Official Gazette of the Union published the dismissal of Eduardo Bolsonaro from his post as a Federal Police clerk for job abandonment. The date matters: Personnel Order nº 135 was signed on September 18 by minister Wellington César Lima e Silva, and the disciplinary administrative process was concluded by the Federal Police in July. What changed on September 21 was the publication, not the decision, as recorded by Gazeta do Povo, Poder360, Metrópoles and CartaCapital.
At the Supreme Court the picture stays where it was: the two requests to investigate justices tied to the Master case remain undated, and the request to revoke Daniel Vorcaro's detention has been with justice Edson Fachin since September 18, undecided. Correio Braziliense records the reaction to a photograph of the prosecutor-general with Vorcaro.
4. Divergences of the day
Market × poll: the outright winner contract prices a 22.40pp distance, and the day's two polls measure a statistical tie in both rounds, with one point in the first round in each. The two instruments do not disagree by a little: they disagree on the order of magnitude of the race.
Poll × poll: the divergence is larger in the third way than at the top, and the two firms measured the same days. What separates them is not time.
The signature of the five pairings: in Palver, Lula scores 43% against all five opponents, and all the variation sits on the opponent's side. In BTG/Nexus he ranges from 44% to 47% depending on whom he faces. Two firms measuring the same pairings in the same week produce different response structures, and that is information about the instrument.
In summary
The gap in the outright winner contract opened to 22.40pp, and the day's extreme is not in the reading the panel publishes. Flávio Bolsonaro rose 1.85pp, to 60.90%, and Lula gave up 3.00pp, to 38.50%, between 00:06 and 17:01. In the recorded readings the gap reached 23.80pp at 15:00 and Flávio 61.70% at 14:30, the two highest of the 158 days of series. Lula is not at his floor, 34.50%, from April 25.
The third-place book moved more than the presidential one and closed almost the entire distance between the top two. Cury rose 7.70pp, to 40.25%, Renan Santos fell 7.00pp, to 45.50%, and the gap between them shrank from 19.95pp to 5.25pp in sixteen hours. Cury has already been paid 60.30% there on September 4.
The two national polls closed fieldwork on the same September 20 and disagree on everything they measured together. BTG/Nexus puts Lula ahead in both rounds and Cury third, 6% against 3% for Renan Santos; Palver puts Flávio ahead in both and Renan Santos third, 11% against 2%. Eight points on the same name, between firms that measured the same days.
Fontes consultadas
matérias com link direto para a notícia (veículos âncora):
- Gazeta do Povo · Governo Lula demite Eduardo Bolsonaro por abandono de cargo da PF
- Gazeta do Povo · O que mostra a pesquisa Palver para presidente
- Gazeta do Povo · O que revela a nova pesquisa Nexus/BTG Pactual para presidente
matérias secundárias (URL Google News redirect, clique resolve à matéria):
- CNN Brasil · Palver: Lula empata com Flávio, Caiado e Zema no 2º turno
- UOL Notícias · BTG/Nexus: Lula e Flávio Bolsonaro empatam na margem de erro nos 2 turnos
- Correio Braziliense · Palver: Flávio e Lula empatados na estimulada e na espontânea no 1º turno
- Opera Mundi · BTG/NEXUS: Lula e Flávio Bolsonaro mantêm empate técnico nos dois turnos
- Metrópoles · BTG/Nexus: presidente Lula é aprovado por 46% e 50% o desaprovam
- CNN Brasil · Nexus/BTG: Lula é desaprovado por 50%; aprovação é de 46%
- Correio Braziliense · Henrique Vieira vê foto de Gonet com Vorcaro como desconfortável
Fontes técnicas: Polymarket (live quotes via the AFOS proxy, fetched Sep 21, 17:01 BRT), TSE registry (official electoral polls).
Sources cited in this text: Polymarket, TSE (registro público), Gazeta do Povo, CNN Brasil, UOL Notícias, Correio Braziliense, Opera Mundi, Metrópoles, Poder360, CartaCapital, Folha de S.Paulo, O Globo, Estadão, JOTA
Method: this synthesis is generated automatically from auditable data on the AFOS Analytics platform, under code-versioned rules in git. All claims can be verified on the platform or in the linked sources. Understand the automated governance.
Integration: for live data and detailed candidate analyses, access the full dashboard. To understand the method in depth, read The Method.
Glossary: Brazilian political terms used in the syntheses (TSE, STF, BolsoMaster, lideranças envelhecidas, etc.), definitions in 3 languages. See the full glossary →
Get it by email
Every day, AFOS Daily cross-references prediction markets, polls registered with the Brazilian electoral court, the press and the day’s coverage.
No spam. Unsubscribe anytime.
After signing up you choose whether to receive it in English, Portuguese or Spanish.