AFOS Analytics
Real-time political pricing
Prediction markets × polls × news
no smoothed averages
Week of two reversals, one in price and one in data. On price, the gap Lula × Flávio opened Monday at +39.05pp, 0.50pp from the cycle record, closed Friday at +37.95pp and kept compressing to +37.25pp on Sunday (↓1.80pp in the window), with Lula losing the 61% (61.50% → 60.50%). The give-back that Edition №7 signaled as possible overshoot confirmed. On data, a correction we need to make upfront: Edition №7 stated that the window 06-12/Jul would bring no registered national polls. That was wrong. Two nationals were published on 08/Jul, both registered at TSE since 02/Jul, and remained invisible to us due to a scope classification bug in our TSE ingestion, fixed on 12/Jul. They matter, and greatly: the Meio/Ideia (BR-05628/2026, n=1.500) showed 1º turno Lula 40.4 × Flávio 32 and runoff 45 × 40; the Gerp/AESP (BR-03067/2026, n=2.000, field 3-7/Jul) showed TIE in 1º turno (36 × 36) and Flávio WINNING the runoff (45 × 42). It is the first national of the cycle to place Flávio ahead in the second round. Two polls, same publication date, opposite winners in the runoff. Gerp is a smaller institute and was already an outlier on 24/Jun, so one poll does not make a trend. But the negotiable fact changes: the market sustained a gap of ~+38pp during a week in which there existed, published and not priced, a national that tied the 1º turno. Below the top, Renan made a sharp V in the winner (9.55% → 8.25% on Thursday → 10.05% on Friday) and Caiado rose to 3rd place in 1º turno (12.5% → 15.5%). The presidential contract accumulates USD 112.2M (USD 111.9M at Friday close). The next window is dense: four nationals between 13 and 16/Jul, including Quaest (n=2.004) and PoderData (n=2.400).
1.Executive Summary
Week 06-10/Jul reversed the previous momentum. The presidential market opened Monday with the gap at +39.05pp, 0.50pp from the cycle high, closed Friday at +37.95pp and continued compressing to +37.25pp on Sunday (↓1.80pp in the window), with Lula returning to 61% (61.50% → 60.50%) and Flávio recovering over the weekend (22.45% → 23.25%). The give-back that Edition №7 signaled as possible overshoot materialized, and the gap did not return to test the top. But the most relevant fact of the window was not in the price, and we did not catch it at the time: two national polls were published on July 8, within the covered week, and fell outside our national cut due to a scope classification bug in TSE ingestion (corrected on July 12). The Meio/Ideia kept Lula ahead (2º turno 45 × 40). The Gerp/AESP showed a tie in the 1º turno (36 × 36) and Flávio winning the 2º turno (45 × 42), the first of the cycle to do so. The reader needs both caveats: Gerp is a smaller-scale institute with reliability 3, and was already an outlier in the June 24 round, systematically more favorable to Flávio than the first-tier consensus. One poll does not make a trend. What changes is the framing: the market carried a gap of ~+38pp during a week in which there existed, published, a national poll that tied the 1º turno. Market snapshot July 12, 20:26 BRT.
2.Why AFOS does not smooth
The aggregation industry reduces 'the market versus polls' to a single, smoothed reading. This week, the average would not only erase the tension between market and polls: it would erase the tension BETWEEN THE POLLS THEMSELVES. Two national polls published on the same July 8 point to opposite winners in the 2º turno.
Why it matters: the two national polls from July 8 (Meio/Ideia n=1.500 and Gerp/AESP n=2.000) were published on the same day, measuring the same electorate, and point to OPPOSITE winners in the 2º turno. An average of the two would produce a number that no institute measured and would erase the only snapshot of the cycle that puts Flávio ahead in the runoff. The caveat is mandatory and comes with it: Gerp is a smaller institute (reliability 3) and was already an outlier on June 24, systematically more favorable to Flávio than the first-tier consensus. Reporting the poll is not endorsing it; it is giving the reader the full contradiction, with institute, sample, field, and protocol, so he can assign the weight. Method caveat: market gap is probability of victory, not vote margin, so the comparison with polls is one of direction and conviction, not pp by pp.
Different from Editions №6 and №7, in which the divergence to report was market × survey, this week the most significant divergence is survey × survey. And it only appeared because we corrected a bug of our own. An aggregator that had averaged the two national polls from July 8 would have produced a number that no institute measured, and the reader would never know that one of them puts Flávio winning the 2º turno.
The same framework we apply to Brazil, measuring the distance between the prediction market and polls and checking the signal against the actual election results, has already been validated against nine elections across four continents: South Korea (2025, Asia's first case), Peru and Colombia (2026), Chile, Germany and Canada (2025) and United Kingdom, Mexico and United States (2024). In some cases the signal was convergence, in others divergence, and the method reports both without retroactive adjustment. The complete dataset, with methodology and code, is open under Apache 2.0 and has a permanent DOI in Harvard Dataverse.
3.Weighted scenarios for the week
Scenarios for the July 13-17 window, which brings four national polls in four days (Nexus/BTG Pactual 13/Jul, 100 Cidades 14/Jul, Quaest 15/Jul, PoderData 16/Jul). It is the first real convergence test since 08/Jul, and what resolves the ambiguity opened by Gerp.
The four national polls confirm the first-tier consensus (Lula ahead with a 5 to 9pp margin in the 2º turno) and Gerp remains isolated as an outlier. The market treats the week as resolved noise, the gap stabilizes in the +36 to +39pp range and the LEVEL divergence persists without resolving: implied probability well above the vote margin. Renan remains near 10% in the winner with ~2-4% in polls, the widest divergence on the panel. Net-neutral for BRL.
One or more of the four national polls confirm the tightening that Gerp signaled (runoff within the margin, or Flávio ahead). In that case Gerp stops being an outlier and becomes the first print of a trend, and the gap of ~+38pp becomes the fragile number of the panel, not the robust one. The market would have carried for weeks a cushion that stated intent does not sustain. Watch especially Quaest (July 15, n=2.004, reliability 0.90), the one with the highest methodological quality in the window. Rapid compression of the gap to +30-33pp would be the signal. Risk of repricing of assets sensitive to electoral risk.
The STF decides on the PT's request to revoke Jair Bolsonaro's house arrest, filed on July 11 after the letter read by Flávio. The contract on impeachment of STF justice remains stuck at the floor at 2.75%, meaning the market does not price in institutional rupture. A hardening decision (return to closed regime) at a moment of family succession in open dispute is the type of shock that current pricing completely overlooks. High volatility in the 2º turno and 3rd place contracts.
4.Indicator Grid
| Contract | Current | Δ week | Vol USD acc. | Implied reading |
|---|---|---|---|---|
| Gap Lula × Flávio | +37.25pp | ↓1.80pp semana | 112.2M | The week opened at +39.05pp (0.50pp from the record) and reversed; it closed on Friday at +37.95pp and continued compressing over the weekend. The overshoot signaled in Edition №7 was confirmed. |
| Lula · winner | 60.50% | ↓1.00pp semana | 7.3M | Lost 61% on Friday after four days stuck at 61.50%; stable at 60.50% for three consecutive days since then |
| Flávio · winner | 23.25% | ↑0.80pp semana | 7.3M | Closed the sixth at 22.55% and recovered over the weekend; the narrowing of the gap came first from Lula's decline and only later from his advance |
| Renan Santos · winner | 10.25% | ↑0.70pp semana | 8.1M | V violent in the window (9.55% → 8.25% on Thursday → 10.05% on Friday), with no single trigger; highest individual volume among living contenders |
| Caiado · 3rd place 1º turno | 15.50% | ↑3.00pp semana | 32k | Doubled its level in the contract without depending on momentum in the winner; it is the cleanest structural gain of the 3rd way in the window |
| Renan · 3rd place 1º turno | 66.50% | ↑3.50pp semana | 136k | Dropped to 61.5% in the middle of the window and WIDENED at the end of the weekend; comfortable favorite, with the divergence against the 2-4% of the broader national polls being the widest in the panel |
| Flávio · 2nd place 1º turno | 78.50% | ↑1.50pp semana | 162k | Condition as certain opponent to Lula reinforced; contract rose while the gap at the top compressed |
| STF impeachment < 2027 | 2.75% | 0.00pp semana | 82k | Stuck at the floor all week, not reacting to the Master case or the PT's request regarding Jair's house arrest; institutional risk is not being priced in |
| PL · more Senate seats | 87.50% | 0.00pp semana | 251k | Stable following the +12.0pp jump from the previous week; the party's institutional reach priced as consolidated |
5.Liquidity and market structure
The five largest book volumes remain in probability contracts ≤10.25%, and the top three in names priced at 0.15%, 0.05%, and 0.15%: Tarcísio (USD 13.42M), Carlos Massa (10.35M), and Eduardo Bolsonaro (10.09M), all legacy positions never unwound. It is legacy volume, not current traction. Read 'low price + high volume' as concentrated conviction already priced in and settled, not as live interest. The week's insight is in 5th place in the ranking: Renan Santos is the only top-5 contender with double-digit probability (10.25%, USD 8.05M) and continues to lead accumulated volume among live contenders, ahead of Flávio (USD 7.3M) and Lula (USD 7.3M). He sustains double digits in the winner contract while national polls measure him at 2-4%, the widest divergence in the panel, and he did this with the largest real flow in the active group.
The 5 largest accumulated volumes account for ~USD 50.8M (~45%) of the presidential market (total USD 112.2M). The cross-reading of price × volume for the week is in the reversal of the top: the gap moved from +39.05pp to +37.95pp with Lula (USD 7.3M) and Flávio (USD 7.3M) in real and balanced flow, meaning the give-back was not made on a shallow book. Low volume spike (USD <500k) in individual contract should be treated as noise until confirmation of recurring flow; this is the case with MDB in the Senate (11.55% with ~USD 5k).
6.Calendar of price-relevant prints
| Date | Sample | Why it matters | |
|---|---|---|---|
| Mon Jul 13 | NEXUS (national) | n=2.000 · BR-07981/2026 · conf. 0.70 | Opens the week of the four national polls; first convergence test since July 8 |
| Tue Jul 14 | 100 Cities (national) | n=2.000 · BR-07294/2026 · conf. 0.70 | Second consecutive national reading; confirms or isolates the NEXUS signal |
| Wed Jul 15 | Genial/Quaest (national) | n=2.004 · BR-07181/2026 · conf. 0.90 | The most important snapshot from this window: the highest methodological reliability of the four. This is the one that determines whether Gerp was an outlier or the first signal. |
| Thu Jul 16 | PoderData (national) | n=2.400 · BR-00059/2026 · conf. 0.70 | Largest sample in the window; closes the sequence of four national polls in four days |
| Sat Jul 25 | Event · Milei in Brazil | não é pesquisa | Confirmed to launch Flávio; news flow catalyst, not electoral data catalyst. Monitor reaction in 2º turno contracts. |
Source: TSE registry via AFOS Analytics API. Correction to Edition №7: that edition stated that the window 06-12/Jul would not bring registered national research. It was wrong. Two nationals were published on 08/Jul (Meio/Ideia BR-05628/2026 and Gerp/AESP BR-03067/2026), both registered at TSE since 02/Jul, and fell outside our national cut due to a scope classification bug in TSE ingestion, corrected on 12/Jul. Status 'registered ≠ published' remains valid: confirmation of effective disclosure requires two primary sources before citing numbers.
7.Watch list, week triggers
- Quaest July 15 is the deciding snapshot. Of the four national polls this week, it is the one with the highest methodological reliability (0.90, n=2.004). If it comes with 2º turno within the margin or with Flávio ahead, Gerp ceases to be an outlier and becomes the first sign of a trend, and the gap of ~+38pp becomes the fragile number of the panel. If it comes with Lula comfortably ahead (5 to 9pp), Gerp remains isolated and the market stays anchored.
- The gap lost the lead. The week opened at +39.05pp, 0.50pp away from the cycle high, and closed at +37.95pp, without retesting the maximum. The give-back signaled in Edition №7 was confirmed. Monitor whether it stabilizes above +36pp (high gap regime) or if the nationals force compression toward +30-33pp.
- Renan: price without backing. Made a V from 9.55% to 8.25% and back to 10.05% in the week, with no single identifiable trigger, and continues with the largest individual volume on the book (USD 8.05M). National polls measure it at 2-4%. The four polls of the week test this foundation: without appearing above 4% in any of them, the two-digit price remains uncovered.
- Caiado is the only structural gain for the third way. It rose from 12.5% to 16.0% in third place of the 1º turno while Renan ceded in the same contract. In the Gerp of July 8, he is the best of the third way in the 1º turno (4%) and has the smallest margin of defeat against Lula in the 2º turno (40 × 36). He is the name to watch in a scenario of weakening of Flávio.
- The STF is not priced in. The ministerial impeachment contract remains fixed at 2.75% and has not moved any day this week, despite Dino's asset freeze of Eduardo Cunha, the advancement of the Master case, and the PT's request (11/Jul) to revoke Jair Bolsonaro's house arrest. Real money treats institutional friction as noise. If this reading is wrong, it is the contract with the greatest asymmetry on the panel.
8.Methodology
AFOS Tradeoff aggregates three signals without mediating them into a composite: Polymarket (denominated in USD, latency ~30min), surveys registered with the TSE (stated intent, variable frequency) and 400+ press sources (event flow). When the three diverge, the divergence is the signal, not the consensus. USD volume is reported alongside implied probability to separate conviction from artificial spike. Liquidity (book depth) is not cited inline: low liquidity on Polymarket does not mean the price is wrong, the market is continuously arbitraged.
This edition corrects a false statement from Edition №7, caused by a bug in our own TSE ingestion. The code is open under Apache 2.0 and the dataset has a permanent DOI on Harvard Dataverse; the correction is documented in the repository. Reporting our own error is part of the method, not an exception to it.
9.Additional reading · macro coverage
Relevant articles and sections from the week in reference outlets. AFOS Tradeoff is the primary source (Polymarket + TSE + polls); the references below are supplementary reading for macro context alignment. Attention: some content operates behind a paywall.
- Poder360 · Gerp/AESP: Lula and Flávio tie at 36% in the 1º turno; in the 2º turno Flávio has 45% and Lula 42% (Jul 08), www.poder360.com.br/poder-eleicoes/flavio-tem-45-e-lula-42-em-eventual-2o-turno-diz-pesquisa/
- Gazeta do Povo · Gerp releases electoral survey for president of the Republic: methodology, field, and TSE registration (July 8), www.gazetadopovo.com.br/eleicoes/2026/pesquisa-eleitoral-2026/gerp-presidente-julho-2026/
- G1 · Dino cites nomination of amendments by Eduardo Cunha despite lacking mandate and orders blockage of up to R$ 6 million (July 12), g1.globo.com/politica/noticia/2026/07/12/dino-cita-indicacao-de-emendas-por-eduardo-cunha-mesmo-sem-mandato-e-manda-bloquear-ate-r-6-milhoes.ghtml
- Estadão · Lindbergh requests revocation of Jair Bolsonaro's house arrest through letter read by Flávio (July 11)(paywall), www.estadao.com.br/politica/lindbergh-pede-revogacao-da-prisao-domiciliar-de-jair-bolsonaro-por-carta-lida-por-flavio/
- Estadão · Motta expresses 'nonconformity' against Dino's decision to freeze assets of Valdemar Costa Neto (July 12)(paywall), www.estadao.com.br/politica/motta-manifesta-inconformismo-contra-decisao-de-dino-de-bloquear-bens-de-valdemar-npr/
- G1 · Republicanos denies closing support for Flávio Bolsonaro and signals neutrality in the elections (July 12), g1.globo.com/politica/noticia/2026/07/12/republicanos-nega-que-tenha-fechado-apoio-a-flavio-bolsonaro-e-indica-neutralidade-nas-eleicoes.ghtml
Macro context references. The primary signal of this edition (reversal of the gap from +39.05pp to +37.95pp in the week, and the two national 2º turno contracts with opposite winners, one of them invisible to us until July 12 due to our own bug) is direct observation of Polymarket pricing crossed with TSE records, not derived from the articles above.