AFOS Analytics
Real-time political pricing
Prediction markets × polls × news
no smoothed averages
BrazilThe week of 12 to 16 October gave back everything the gap in the winner contract had opened the week before, and more. It went from 74.10pp at Monday's close to 55.20pp at Friday's, with Flávio Bolsonaro at 77.70% and Lula at 22.50%. That is -18.90pp in five sessions, the largest weekly drop among the 13 weeks that closed toward Lula since 14 April and the second-largest move in absolute value among 27, behind only the +25.30pp of the week of 7 September. Counted from the 78.50pp close of 9 October, published in №21, it is 23.30pp in seven calendar days.
The path went almost entirely one way: 74.10, 72.70, 69.00, 59.00 and 55.20pp. The largest step was Thursday's, 15 October, -10.00pp, and the intraday floor, 49.80pp, came at 16:30 UTC that day, with Lula at 25.50%. On Thursday Datafolha, Palver and Indexa were released, and Lula gave an interview to Jornal Nacional that night; AFOS measures price and fact with each one's timestamp and does not attribute the contract's move to either of them.
The polls moved much less. The eleven national polls of the week that published total votes gave Flávio Bolsonaro 0.9 to 9 points, with a simple average of 2.9, against 3.2 the week before. What changed was within each pollster: the three that measured twice during the week, Gerp, Futura/100% Cidades and PoderData/Aya, all moved toward Lula.
This time the contract that received the most new money was the one that gained price: USD 2.85M in Lula and USD 1.66M in Flávio Bolsonaro, in a book that received USD 4.51M during the week and reached USD 184.64M accumulated. In the 2026 inflation book, the 5.00% to 5.49% band rose 16.60pp, to 62.60%, with USD 2,071 of new money, in the week in which the Senate approved and Congress promulgated the end of the 6x1 work schedule.
1.Executive Summary
The theme of the week is the gap between the speed of the two instruments. The contract took 18.90pp off the gap in five sessions and ended the week paying about three and a half times for Flávio Bolsonaro what it pays for Lula, against more than eight times the week before. The polls barely changed on average, from 3.2 to 2.9 points, but they changed within each pollster: those that measured twice saw the distance shrink. None of the eleven national polls with total votes put Lula ahead during the week; the next one to do so was Vox Brasil, on 18 October. Price and poll answer different questions, the probability of winning on 25 October and the week's voting intention, and this issue measures the two side by side.
2.Why AFOS does not smooth
The twelve national polls released from 12 to 16 October measured the second round with samples of 2,000 to 5,000 interviews and fieldwork between 7 and 16 October; eleven published total votes. A simple average of each one's lead returns 2.9 points for Flávio Bolsonaro, almost the same figure as the week before, which reads as a stable race.
Why it matters: the average of 2.9 points suggests a flat week, and the contract took 18.90pp off the gap over the same period. What moved was each pollster's reading against itself, in the same methodology and with consecutive fieldwork, and that comparison is what the average dilutes by mixing pollsters with distances from 0.9 to 9 points. If the week of 19 to 23 October repeats the move within pollsters, the average will follow; if it does not, one week's move did not make a trend.
Issue №21 assigned about 55% to the base scenario, in which the national polls of 13 to 15 October would repeat the shape of the first week, with most putting Flávio Bolsonaro ahead by 4 to 6 points, and the contract would stay between 85% and 92% for him, with the gap between 70pp and 80pp. It did not materialize: of the eleven national polls of the week that published total votes, only Gerp of 16 October fell in that range, with 5 points, and nine measured up to 3.4 points; the contract closed the week at 77.70% and the gap at 55.20pp. About 30% went to the contrarian scenario, in which Quaest, AtlasIntel and Datafolha would measure the distance falling inside the margin and the gap would give back part of what it had opened. It largely materialized: Quaest and Datafolha measured 2 points each, 46% to 44% and 47% to 45%; AtlasIntel did not publish during the week, and its registration for 15 October was withdrawn from the TSE; and the gap gave back more than it had opened, from 78.50pp on 9 October to 55.20pp. About 15% went to the tail, the 15 October debate or an institutional act. The debate, which №21 gave as scheduled on SBT, RedeTV! and CNN Brasil based on the calendars of 10 October, came off the schedule and did not take place; the two candidates' first encounter moved to Band, on 18 October. The week's acts, the TSE rules for the campaigns' WhatsApp channels and the removal of campaign videos, did not reach either candidacy.
3.Weighted scenarios for the week
The week of 19 to 23 October, the last before the vote on 25 October, has twenty national polls registered with the TSE with scheduled release, among them AtlasIntel, Quaest, Palver and Datafolha, and the Globo debate on 23 October. After the window, Ipespe/Abrapel (48% to 46%) and Veritá (51.1% to 48.9% in valid votes) put Flávio Bolsonaro ahead on 17 October, and Vox Brasil put Lula ahead, 46.4% to 44.6%, on 18 October. In the confirmed reading of 18 October at 17:44 BRT, the contract paid 79.25% for Flávio Bolsonaro and 20.50% for Lula, and the two candidates' first debate, on Band, was scheduled for that night.
The national polls of 19 to 23 October repeat the shape of the week of 12 to 16: most put Flávio Bolsonaro ahead by up to 5 points in total votes, and one pollster measures Lula ahead within the margin. The contract stays between 70% and 85% for Flávio Bolsonaro, and the gap oscillates between 40pp and 70pp without repeating the 10.00pp step of Thursday, 15 October.
Quaest and Datafolha, with fieldwork after the Band debate, and AtlasIntel measure a tie or Lula ahead, as Vox Brasil did on 18 October. In that case the contract, which pays almost four times for Flávio Bolsonaro what it pays for Lula, stands against the reading of three reference pollsters in the last week before the vote, and the gap gives back more than it already has.
The Globo debate, on 23 October, or an institutional act on one of the two candidates changes the race in the final two days, after the last round of polls. The Supreme Court impeachment contract, at 2.70% at Friday's close, is the panel's closest thermometer of institutional risk, and spent the week between 2.40% and 3.40%.
4.Indicator Grid
| Contract | Current | Δ week | Vol USD acc. | Implied reading |
|---|---|---|---|---|
| Presidential gap, Flávio minus Lula | +55.20pp | -18.90pp | USD 184.64M | Path: +74.10, +72.70, +69.00, +59.00, +55.20. Top of 75.10pp at 02:30 UTC on Tuesday and floor of 49.80pp at 16:30 UTC on Thursday. Largest weekly drop among the 13 weeks that closed toward Lula, second-largest in absolute value of 27. |
| Flávio Bolsonaro, presidential | 77.70% | -8.90pp | USD 21.32M | Top of 87.60% at 02:30 UTC on Tuesday and floor of 75.30% at 16:30 UTC on Thursday. Received USD 1.66M during the week, and the move normalized by the pair's sum is -9.84pp. |
| Lula, presidential | 22.50% | +10.00pp | USD 25.16M | Floor of 12.50% at 02:30 UTC on Monday and top of 25.50% at 16:30 UTC on Thursday. Received USD 2.85M, more than Flávio Bolsonaro, and still has the largest accumulated volume in the book. |
| Supreme Court impeachment before 2027 | 2.70% | -0.40pp | USD 0.10M | Top of 3.40% at 03:00 UTC on Monday and floor of 2.40% at 11:00 UTC on Tuesday, with USD 426 of new money during the week. A thin book, where a few trades move the price. |
| 2026 inflation, 5.00% to 5.49% | 62.60% | +16.60pp | USD 27 thousand | Floor of 46.00% at 01:00 UTC on Monday and top of 63.40% at 21:30 UTC on Friday. Path: 46.00, 47.20, 50.20, 51.10, 62.60, with USD 2,071 of new money. |
| 2026 inflation, 4.50% to 4.99% | 18.60% | -15.40pp | USD 32 thousand | Top of 37.80% at 08:00 UTC on Monday and floor of 15.20% at 11:00 UTC on Thursday. Still the second most likely band, 1.60pp above the 5.50% to 5.99% band, with USD 396 of new money. |
| 2026 inflation, 5.50% to 5.99% | 17.00% | +2.80pp | USD 21 thousand | Floor of 11.40% at 15:30 UTC on Monday and top of 31.30% at 03:00 UTC on Wednesday, with USD 445 of new money over the whole week. |
5.Liquidity and market structure
The presidential book received USD 4.51M of new money between Monday's and Friday's closes, and all of it went to the two contracts still open: USD 2.85M in Lula and USD 1.66M in Flávio Bolsonaro. The week before, Lula's contract received USD 3.70M and lost 5 points; this week it received USD 2.85M and gained 10. Volume counts trades in both directions, buying and selling, and says nothing about anyone's direction: the same larger flow to one side accompanied the price falling one week and rising the next. The other seventeen contracts in the book have been closed since the first round, resolved at zero, and their volume is historical.
Volume is accumulated since each contract opened. The week's new money is the difference between the volumes recorded at the closes of 12 and 16 October, and the prices in the table are those of Friday's close, at 22:30 UTC. AFOS does not cite book depth in editorial text, because low liquidity in a continuously arbitraged market does not mean the price is wrong.
6.Calendar of price-relevant prints
| Date | Sample | Why it matters | |
|---|---|---|---|
| 19/Oct | AtlasIntel | n=5,000 | The pollster's first since 9 October, when it measured 51.1% to 45.7% for Flávio Bolsonaro. Fieldwork from 13 to 18 October. |
| 19/Oct | PoderData | n=3,000 | On 16 October it measured a tie at 50% of valid votes, without publishing total votes. Fieldwork from 16 to 18 October, ending on the day of the Band debate. |
| 21/Oct | Quaest | n=2,154 | On 14 October it measured 46% to 44% for Flávio Bolsonaro. Fieldwork from 17 to 20 October. |
| 21/Oct | Palver | n=5,000 | On 15 October, in the pollster's first national poll after the first round, it measured 50% to 49% for Flávio Bolsonaro. Fieldwork from 16 to 20 October. |
| 22/Oct | Datafolha | n=2,520 | On 15 October it measured 47% to 45%, at the limit of the margin of 2 points. Fieldwork from 20 to 22 October. |
| 23/Oct | Vox Brasil | n=2,100 | On 18 October it measured 46.4% to 44.6% for Lula, the second time after the vote with him ahead. Fieldwork from 20 to 22 October. |
Registration with the TSE is not a confirmed release. These are six of the twenty national polls with a sample of 1,000 or more and release registered between 19 and 23 October; the other fourteen include Veritá, Gerp, Nexus, 100 Cidades, DataTrends, Alfa, Ideia, Indexa, Ipespe and the PoderData poll of 22 October.
7.Watch list, week triggers
- Five reference pollsters in four days, from 19 to 22 October AtlasIntel, PoderData, Quaest, Palver and Datafolha publish with fieldwork that touches or passes the Band debate. In the week of 12 to 16 October, the national polls with total votes gave Flávio Bolsonaro 0.9 to 9 points.
- The two debates, on 18 and 23 October Band's is the two candidates' first encounter in this election, and Globo's is the last before the vote on 25 October. The 15 October debate, which was on the calendars of 10 October, did not take place.
- Vox Brasil It is the only pollster that put Lula ahead after the vote, on 9 and 18 October, by 1.5 and 1.8 points, and it has a new release registered for 23 October.
- The inflation book The 5.00% to 5.49% band went from 46.00% to 62.60% in the week in which the Senate approved, on 14 October, and Congress promulgated, on 16 October, the end of the 6x1 work schedule. The move ran in a book with USD 27 thousand accumulated in that band, with USD 2,071 of new money.
- The pair at almost four times With the other contracts closed, the sum of Flávio Bolsonaro and Lula went from 99.10% to 100.20% between Monday's and Friday's closes, and the normalized move was 9.84pp for each side, against 8.90pp and 10.00pp raw.
8.Methodology
AFOS does not produce a weighted average and does not arbitrate between sources. It publishes the three signals side by side, contract price, registered poll and news flow, and reports the distance between them when it exists. A market price goes live only after two independent readings eight minutes apart that agree within 0.20pp, contract by contract, and the reference series are measured in the daily backup of the database and not in the public interface. The week runs from Monday to Friday: the published value is Friday's close, the path is the five daily closes, the Δ is Friday minus Monday, and the gap is measured at paired instants, both sides with the same timestamp.
The polls cited are those registered with the Superior Electoral Court, with protocol, sample, margin and field dates declared, and in the second round AFOS compares the price with each pollster's total vote, not with valid votes. The code and the method are open under the Apache 2.0 licence. Every factual claim in this issue links to its primary source.
9.Additional reading · macro coverage
Complementary macro context. AFOS is a primary source on divergence between market and polls; the links below cover the economic and institutional backdrop of the week.
- Valor Econômico · Reaction of inflation premiums in the interest-rate market to the end of the 6x1 work schedule, on 17 October.(paywall), valor.globo.com/financas/intraday/post/2026/10/fim-da-escala-6x1-assusta-mercado-e-provoca-disparada-de-premios-de-inflacao.ghtml
- TSE, official results · Official first-round tally, by office and by state., resultados.tse.jus.br/oficial/app/index.html
- Reuters Brazil · English-language flow on politics and currency, useful for the non-resident investor., www.reuters.com/world/americas/
- Poder360 · Registry of electoral polls with protocol and history by pollster., www.poder360.com.br/pesquisas/
Get it by email
AFOS Tradeoff is the weekly read on Brazilian political risk, for people who decide on data rather than headlines.
No spam. Unsubscribe anytime.
After signing up you choose whether to receive it in English, Portuguese or Spanish.