AFOS Analytics
Real-time political pricing
Prediction markets × polls × news
no smoothed averages
United StatesThe Senate control contract went from 46.50% for the Democrats at Monday's close, Aug 10, to 52.50% at Friday's close, Aug 14. That is 6.00pp over five sessions, and the entire move fit into two days: Aug 11 and Aug 12. After that the price did not budge, and it reaches this issue flat for four closes, at the high of the series that began on Jul 18, with no day above it in thirty.
In the same week the AFOS generic ballot average went from D+5.66 to D+6.20. That is 0.54pp, and this is where the week gets interesting: no new poll entered the base. The most recent field date across the whole set is still Aug 04, as it has been since then. What changed was composition: the 30-day window rolled and the poll count fell from 29 to 20. The average rose because nine polls dropped off the edge, not because anyone measured a voter.
📌 Both numbers moved, and only one of them measured anything. The price reacted to results that arrived on primary Tuesday. The poll average moved by window arithmetic, with zero new information about vote intention. For anyone reading a series, that distinction is the difference between signal and noise wearing the look of signal.
The House did not follow. The Democratic control contract rose 1.00pp on the week, from 86.50% to 87.50%, and it also stopped. It sits one point below the high of the series that began on Jul 28, with 1 day above it in twenty. The same trigger moved the Senate six times as much as the House.
1.Executive Summary
The week had one large move and one apparent move, and they are not the same thing. Beyond separating the two, this issue shows that the Senate seats distribution agrees with the binary contract to within about a point, and it records that the previous issue's base scenario lost six points in seven days.
2.Why this brief does not subtract the market from the poll
This brief never subtracts the market price from the poll average, and last week is the best demonstration of why. The two quantities responded to different things, at different speeds, for different reasons. A subtraction between them would produce a number that moves when nothing electoral happens.
📐 The shape of the dispersion, which is more informative than the average: the 20 polls in the window run from D+2 to D+11, that is 9.00pp of spread, with a median of D+6 against a mean of D+6.20. A house with Republican identification appears at both extremes, which means the pollster's partisan label does not predict which way it errs. ⚠️ And the caveat that weighs most in this issue: with the most recent field date in the base at Aug 04, the ruler has been frozen for thirteen days. An average that moves while the ruler is frozen is composition, not measurement, and that is how it should be read until a new round enters.
3.Weighted scenarios for the week
The scenarios below are not narrative. They come from the Republican Senate seats book, which prices each band separately and whose bands summed to 100.30% at capture, inside the 95% to 105% gate, and inside it as well in every reading of the last 24 hours. It is the only distribution book in this issue that clears both tests, alongside the turnout book.
Sum of the bands in which the Republicans finish with 49 seats or fewer: 25.50% for 47 or fewer, 12.50% for exactly 48 and 13.50% for exactly 49. 📌 The binary control contract pays 52.50% for the Democrats, which means the two books agree to within about a point. That coherence is checkable and worth recording, because on the same day the House seats book summed to 108% and does not clear the gate.
Sum of the bands of 51 seats or more, pulled by the 15.00% on the band of exactly 51. ⚠️ This was the base scenario of Issue №3, when the binary contract paid 54.50% on this side. It lost six points in seven days and stopped being the most likely outcome in the book. The 54.50% price carried no information about its own fragility, and that is precisely what a single number hides.
The band for exactly 50 Republican seats pays 12.50% and is the only one in the book where the election does not decide who controls the chamber: the tiebreak belongs to the vice president. ⛔ AFOS does not assign that band to either side, because doing so would require a political premise that is not measurement. It stands declared separately, and it is the reason Democratic control (51.50%), tie (12.50%) and Republican control (36.30%) add up to the book's 100.30%.
4.Indicator Grid
| Contract | Current | Δ week | Vol USD acc. | Implied reading |
|---|---|---|---|---|
| House · Democratic control | 87.50% | +1.00pp | USD 5.29M | Flat for five closes. One day above it in twenty since Jul 28. |
| House · Republican control | 12.50% | 0.00pp | USD 4.14M | Did not move a hundredth all week, and sits at the low of the series. |
| Senate · Democratic control | 52.50% | +6.00pp | USD 2.27M | High of the series that began on Jul 18, zero days above it in thirty. Low of the series: 43.50%. |
| Senate · Republican control | 48.50% | -6.00pp | USD 1.60M | Close to the low of the series, which is 47.50%. High of the series: 56.50%. |
| Senate · Republicans with 47 seats or fewer | 25.50% | n/a | USD 0.17M | Largest single band in the Senate seats book. |
| Senate · Republicans with exactly 50 | 12.50% | n/a | USD 0.12M | A tie. What decides control in that band is the tiebreak, not the ballot box. |
| Turnout between 120M and 125M votes | 27.00% | n/a | USD 0.01M | Modal band. The book summed to 97.05% and clears the gate, but the money in it is small. |
| Calendar · election on Nov 03 as scheduled | 96.05% | n/a | USD 0.34M | No stored series in our base to compute a Δ. At the Issue №3 capture, on Aug 10, it marked 95.95%. |
5.Liquidity and market structure
6.Electoral calendar through November
| Date | Sample | Why it matters | |
|---|---|---|---|
| Aug 18, Tuesday | Alaska and Florida primaries | 2 states | Alaska uses a single nonpartisan primary, with the top four advancing to a ranked-choice general. It is a design that separates partisan preference poorly. |
| Aug 25, Tuesday | Republican runoff in South Carolina | 1 Senate seat | Darline Graham against Ralph Norman, for the nomination to the seat of the late Lindsey Graham. It is the only race on the near calendar that moves Senate arithmetic directly. |
| Sep 15 | Delaware closes the primary calendar | last state | From then on the candidate field is set and the market shifts to pricing the general race, not the nomination. |
| Nov 03 | General election | 435 House seats and 1/3 of the Senate | The date on which every contract in this grid resolves. Until then AFOS measures divergence and cannot declare anyone right. |
⚠️ There is no mandatory poll registry in the United States, so this calendar is one of elections, not of releases. There is no print date to anticipate, and that is why the average's ruler can sit frozen for thirteen days without anything flagging it.
7.Watch list, week triggers
- If the Senate contract leaves 52.50% in either direction. It sits at the high of the series and has been flat for four closes. A move up is territory with no precedent in the series; a move down, after a 6.00pp jump in two days, is a giveback and deserves a different reading.
- The first new poll to enter the base. The most recent field date is Aug 04. When a new round enters, the average will move by measurement rather than by the edge, and only then does its variation become comparable with earlier ones.
- The South Carolina runoff, on Aug 25. It is the only race on the near calendar that directly alters Senate arithmetic, and the control contract sits at the most sensitive point of the series.
- The sum of the House seats book. It stood at 107.40% at capture and 108% in the 24-hour series, so it remains outside the gate with no publishable number. Until it closes between 95% and 105%, it enters as shape only.
- The state governors book, which swings within the day. It varies ten points in 24 hours and therefore publishes no number. If it stabilises, the panel gains one more distribution.
8.Methodology
AFOS does not average market and poll together, and does not subtract one from the other. It publishes both side by side, with each unit declared, and shows the distance. Prices come from Polymarket through our own proxy, with a cache-bypassing read confirmed by a second capture eight minutes later, at a 0.20pp tolerance. The generic ballot average is a simple arithmetic mean, unweighted and with no pollster excluded, over a 30-day window, and when the same pollster publishes more than one cut of the same round only one enters. A distribution book publishes a number only when its bands sum to between 95% and 105% in every reading of the last 24 hours, and not merely at the capture of the moment. In this issue that left out the House seats book, the state governors book and the popular vote margin book, which summed to 153.60%. Code and method open under Apache 2.0.
9.Additional reading · macro coverage
Complementary reading from the week, all of it drawn from the panel's own press collection. AFOS is the primary source for the numbers above; the items below are context and support no measurement in this issue.
- The New York Times, Aug 12 · Why the polls were so wrong in the Wisconsin governor's primary. It is the reporting on the episode that opened the week in which the Senate contract moved., www.nytimes.com/2026/08/12/us/politics/crowley-hong-wisconsin-governor-primary-polls.html
- The Hill, Aug 13 · How Crowley turned the result around against Hong, with a read on turnout and on electorate profile., thehill.com/homenews/campaign/data-nerds/6026856-david-crowley-francesca-hong-wisconsin-upset-polling-turnout-midterms-discussion/
- The New York Times, Aug 12 · The Republican runoff in South Carolina begins, in the race for the seat of the late Lindsey Graham., www.nytimes.com/2026/08/12/us/politics/south-carolina-republican-senate-primary-runoff.html
- The Washington Post, Aug 14 · The first poll of the Michigan race after the primary shows a close contest for the general election., www.washingtonpost.com/politics/2026/08/14/michigans-marquee-senate-race-close-first-poll-after-el-sayed-primary-win/
- NPR, Aug 10 · In the competitive Maine Senate race, women voters as the deciding bloc., www.npr.org/2026/08/10/nx-s1-5925349/maine-senate-race-susan-collins-troy-jackson-women-voters
- The Guardian, Aug 10 · Profile of the candidate leading the Wisconsin primary, published on the eve of the vote that reversed the result., www.theguardian.com/us-news/2026/aug/10/francesca-hong-democratic-socialist-wisconsin-governor-race
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.