AFOS Analytics
Real-time political pricing
Prediction markets × polls × news
no smoothed averages
United StatesThis is the opening issue of our coverage of the United States midterm elections, on November 3, 2026, and it does not report a week. It reports the state of the board, where each number comes from and what this brief refuses to do. Weekly numbering starts with Issue №2. The reason is one of method and is better said in the first line: our own market collection was switched on July 28, so the house series covers three of last week's five trading days, and reporting 'the week' with three days would be selling coverage we do not have. Add to that the fact that the two main contracts did not move: in the House contract Democrats remain at 85.50% and in the Senate one Republicans at 55.50%, both with 0.00pp of variation between July 29 and 31. What did move was the polling: our house average of the generic ballot went from D+6.57 on July 29 to D+5.57 on July 31. We record the difference in behaviour, not a cause: three trading days do not separate repricing from a change in sample composition, and the average itself changed base, from 22 to 21 polls inside the 30-day window. Market reading of July 31, 19:49 BRT.
1.Executive Summary
The board this brief begins to follow has two chambers in play and different behaviour in each. In the House, the market is far from a tie: 85.50% for Democrats against 13.50% for Republicans. In the Senate, it is close: 55.50% for Republicans against 43.50% for Democrats. The Senate seat distribution, a market with no equivalent in Brazil, shows why: the most likely band is Republicans with 47 seats or fewer, at 23.50%, but the bands for 49, 50 and 51 seats add to more than 42%, meaning the market prices a narrow majority as the central outcome. Seat 50 is the one that decides, because a tie is resolved by the vice-president's tie-breaking vote. On the polling side, the house average sits at D+5.57, and the dispersion between recent pollsters is the finding that matters most in this opening: Quinnipiac measured D+7 and Reuters/Ipsos D+2 with fieldwork in practically the same window. Five points apart, between two first-rank houses, in the same week. A single average number hides exactly that, which is why the next section exists.
2.Why this brief does not subtract the market from the poll
The temptation, on a panel showing both sides, is to do the arithmetic: the market gives Democrats 85.50% in the House, the poll gives D+5.57, so one could subtract and announce the divergence. This brief does not do that, and it is not modesty: it is that the result would mean nothing.
What is crossed here, then, is direction and movement, never level. It is fair to say that polling moved a point and the price did not move at all, as happened between July 29 and 31. It is not fair to say that the distance between them is so many points. There is a market that would measure the same quantity as the poll, the popular vote margin, and it would settle the matter for good. Today it does not serve, and that is measured: its bands add to 146.3% when a coherent distribution adds to near 100%, on only USD 113 thousand of volume. It is collected every day and stays off the screen until the arithmetic closes. We also record what weakens this opening's reading: three trading days are few to claim the price ignored the polling, and the average itself changed base over the period.
3.Weighted scenarios for the week
Scenarios anchored in the seat distribution the market itself prices, not in narrative. This is a point where the United States coverage differs from Brazil's: here there is a contract giving the probability of each seat band, so the scenario does not have to be invented, only read. The Senate bands add to 100.3%, inside the house coherence gate.
The bands for 49, 50 and 51 Republican seats concentrate more than 42% of the probability, and the 47 or fewer band alone marks 23.50%. Added together, the bands that hand Republicans the Senate match the 55.50% of the control contract, which is a coherence check between two independent markets and not a house forecast. In the House the picture is different: the most likely band is Republicans below 190 seats, at 22.00%, consistent with the 85.50% Democratic probability in the control contract. What to watch is whether those two readings keep agreeing with each other.
Between July 29 and 31 the generic ballot average gave up 1.00pp and the two main contracts stood still. If that pattern persists for two or three weeks, with the average falling and the House contract anchored above 85%, the reading becomes that the market is pricing geography, not votes: a Democratic majority in votes that does not convert into seats in the same proportion. The signature would be the House seat distribution shifting without the control contract moving. It is the most direct test of the thesis that the two quantities do not subtract.
The popular vote margin contract measures the same quantity as the generic ballot. The day its bands add to between 95% and 105% and the volume leaves the current USD 113 thousand level, the crossing becomes point against point, with no model and no caveat. Today it adds to 146.3%, with almost half the probability parked in a single escape line. It is a market with volume two orders of magnitude below the House contract, so little money is enough to reprice it, for better or worse.
5.Liquidity and market structure
The liquidity distance between first and last is three orders of magnitude. In a market of USD 11.8 thousand, a single participant shifts the price, and the reading is worth less than the number suggests. That is why this brief does not treat a thin-band move as signal.
⚠️ The volume is Polymarket's, and not the entire American market. Kalshi is out of the coverage by declared choice, and that is a real limitation of this panel, not an implementation detail.
6.Electoral calendar through November
| Date | Sample | Why it matters | |
|---|---|---|---|
| Nov 3, 2026 | Election day | whole House + part of the Senate | Resolution date for every contract followed here. It is the only certain date on this calendar |
| Through the year | State primaries | each state has its own calendar | Not in the house dataset. When they enter, they enter with a source, not by estimate |
| Weekly | National generic-ballot trackers | 21 pollsters inside the 30-day window | The house average runs once a day over the index; it is each pollster's publication that is weekly |
| Daily | House market capture | 8 markets, every 30 minutes | With a guaranteed daily point even when the price does not move, so the series does not become a log of changes |
November 3, 2026: election day, with the whole House and part of the Senate in play. It is the resolution date for every contract followed here. State primaries: they run through the year, each state on its own calendar, and they are not in the house dataset; when they enter, they enter with a source. Polling cadence: national trackers publish weekly, and the house average runs once a day. ⚠️ Because there is no prior registry, this brief never announces a poll that has not been released, unlike the Brazilian edition.
7.Watch list, week triggers
- Do the two Senate readings keep agreeing? The control contract gives Republicans 55.50% and the sum of the seat bands that hand them a majority matches it. These are two independent markets measuring the same thing by different routes. If they come apart, one of them is wrong, and the natural suspect is the one with less volume.
- The dispersion between pollsters is larger than the move in the average. Quinnipiac measured D+7 and Reuters/Ipsos D+2 with fieldwork in practically the same window. Five points between two first-rank houses, against 1.00pp of variation in the average over three days. While the spread between houses dominates the move in the average, the aggregate number says less than it appears to.
- The sum of the popular vote margin bands. It is the number deciding whether the direct crossing comes to exist, because that market measures the same quantity as the poll. Today it adds to 146.3%, with almost half the probability parked in a single escape line. The gate to reach the screen is adding to between 95% and 105%.
- Turnout is the thinnest contract in the set. With USD 11.8 thousand accumulated, it is the first to distort when money comes in, and the first place where a move will look like signal without being one. It is on the screen because its bands add up, not because the volume recommends it.
8.Methodology
Market: Polymarket, through our own proxy, captured every 30 minutes with a guaranteed daily point even when the price does not move. A band market is only displayed if its bands add to between 95% and 105%; outside that range it is collected and not displayed, with the measured sum in plain sight. Polls: Wikipedia is used as an INDEX of which polls exist and who published them, and each number is attributed to the pollster with a link to the primary source. AFOS does not redistribute anyone's table. The average is a simple arithmetic mean, unweighted and excluding no pollster, because weighting would require a public quality yardstick, and none has existed since FiveThirtyEight closed in March 2025. When the same pollster publishes more than one cut of the same round, only one enters, by the hierarchy likely voters > registered voters > adults: in this reading, Reuters/Ipsos entered with the registered-voter cut, at D+2, and the adult cut, at D+4, stayed out. From the July 31 reading, 278 rows were published out of 279 read, one discarded for an unreadable shape at the source, and none lack a link to the primary source. American aggregators do not enter: those are third-party models. AFOS reports the crossing. It does not forecast, does not recommend a position, and does not say who will win.
9.Additional reading · macro coverage
Stories from the window in outlets on the AFOS fixed list. The Tradeoff is a primary source (Polymarket + the polling index); the references below are complementary context, and the reading belongs to whoever published it, not to us. The addresses point to the outlet, not to an aggregator. Note: some operate behind a paywall.
- NBC News · Mandela Barnes drops out of the Wisconsin governor's race (Jul 31), www.nbcnews.com/politics/2026-election/democrat-mandela-barnes-drops-wisconsin-governors-race-rcna590096
- NBC News · Democratic Senate candidates pursue a new strategy in rural states: dropping out (Jul 30), www.nbcnews.com/politics/2026-election/democratic-senate-candidates-pursue-new-strategy-rural-states-dropping-rcna589868
- Axios · Divided Democrats face a moment of truth in the Michigan Senate primary (Jul 31), www.axios.com/2026/07/31/democrats-michigan-el-sayed-stevens-senate-primary
- Fox News · Fox News Poll: GOP divisions shape the Texas Senate race (Jul 30), www.foxnews.com/politics/fox-news-poll-gop-divisions-shape-texas-senate-race
- The New York Times · Elon Musk plans at least USD 100 million in midterm spending to help Republicans (Jul 30)(paywall), www.nytimes.com/2026/07/30/us/politics/elon-musk-midterms-republicans.html
- CBS News · Four major Latino organizations join forces for the first time to mobilize voters (Jul 30), www.cbsnews.com/news/latino-organizations-voters-2026-midterm-elections/
- CNN · Bakari Sellers on never getting optimistic about Democrats (Jul 30), edition.cnn.com/2026/07/30/politics/video/bakari-sellers-dont-ever-get-optimistic-about-democrats
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.