AFOS Analytics
Real-time political pricing
Prediction markets × polls × news
no smoothed averages
United StatesThe week of 28 September to 2 October produced two books on the same event with opposite relations between price and implied probability. In the House, the Democratic control contract had exactly ONE distinct value for the entire week: 92.50%, across 46 recorded readings, with the floor equal to the top and an intraday amplitude of 0.00pp. The Republican side of the same book went from 7.50% to 8.50% and came back, with the intraday top reached on 28 September at 14:30 UTC. The consequence is arithmetic: in 17 of the 29 instants in which both outcomes were captured in the SAME snapshot, the pair sum sat above 100.00%, peaking at 101.00%. At that peak, the implied probability of Democratic control of the House was 91.58%, which is 0.92pp below its own price of 92.50%. The weekly delta measured against the previous Friday's close, 25 September, is +0.00pp on both sides. A desk reading only the price sees an empty week; a desk reading the pair sees the overprice level breathe almost a point and come back.
In the Senate the shape is the inverse, and it concentrates in one day. There were seven closes at 62.50%, from 25 September to 1 October, and Friday's close, 2 October, came in at 64.50%, a value first reached at 13:00 UTC that day. The pair stood at 100.00% at both borders of the week, so the 2.00pp survive normalisation whole, which is the opposite of what happened in the House. And the two sides did not move together: the Republican side closed at 36.50% on 1 October, a day before the Democratic side left 62.50%, and it touched the intraday floor of 35.50% on 2 October at 10:30 UTC, with a 3.00pp amplitude for the week against 2.00pp on the Democratic side.
For anyone who has to pick an instrument, the concrete measurement in this issue is this: in the House, the price said nothing and the implied probability moved; in the Senate, the price moved and the implied probability moved by exactly the same amount. They are two binary contracts on the same election, in the same week, and the difference between reading one and reading the other is not in what the election did, it is in where the book's spread was.
1.Executive Summary
The three numbers above are of different natures and do not combine. The first two are contract prices, read live and confirmed by a double reading eight minutes apart. The third is an arithmetic average of published polls, on a 30-day rolling window. What this issue measures is the behaviour of each one, and what it refuses to do is the subtraction between them.
2.Why this brief does not subtract the market from the poll
This section exists because the question comes up in every desk conversation: if the market pays 92.50% for a Democratic House and the polling average gives a Democratic lead of 7.92 points, what is the distance between the two readings? The answer is that the question has no numerical answer, and the reason is one of units.
Subtracting 7.92 from 92.50 would produce the number 84.58, which is not a probability, is not a vote point and would change size if the poll were expressed on another scale, for example as a share of the two-party vote. The 2012 case is the short historical example: Democrats won more votes for the House and fewer seats, because the conversion of vote into seat depends on geography and on district design. This brief reports the two readings side by side and never the difference between them.
The practical rule that follows is the one that organises this entire issue: when the two readings diverge, the divergence IS the signal, and it is described, not resolved. Anyone who needs a single figure between the two is asking for a forecast, and AFOS does not produce forecasts.
3.Weighted scenarios for the week
The scenarios below are not narrative: they are read in the prices of the contracts the market already publishes, with the probability of each block declared. In this issue they use the two control binaries and the distribution of state governorships, whose sum of ranges falls inside the 95% to 105% gate and can therefore be cited as it stands.
At the 2 October close, the Democratic control contract for the House paid 92.50% and the Senate one 64.50%, with both pairs summing to 100.00% at that instant, so raw and normalised coincide in both. In the certified reading of 5 October at 03:09 UTC, the Democratic Senate had risen further, to 67.50%, with the pair at 101.00% and therefore 66.83% normalised, while the Democratic House was still at 92.50%. The two contracts pay the same side, and only one of them has moved since Friday.
The Republican Senate control contract closed the week at 35.50%, against 37.50% at the 25 September close, with the pair at 100.00% at both borders. The week's intraday floor, 35.50%, was touched on 2 October at 10:30 UTC, and the intraday top, 38.50%, on 28 September at 20:00 UTC: an amplitude of 3.00pp, larger than that of the Democratic side of the same book, which was 2.00pp. In the certified reading of 5 October at 03:09 UTC the contract stood at 33.50%.
The Republican House control contract closed the week at 7.50%, the same as the 25 September close, and was still at 7.50% in the certified reading of 5 October. Within the week it rose to 8.50% and came back, and it was that round trip that took the pair sum to 101.00%. The state governorships book gives the third dimension, which neither binary touches: in the reading of 5 October at 03:04 UTC, the ranges that give fewer than 24 Republican governorships sum to 55.00% raw, or 57.20% normalised by the book's sum, which is 96.15%. The single most expensive range is 22 or 23 governorships, at 31.00%, and fewer than 22 pays 24.00%.
4.Indicator Grid
| Contract | Current | Δ week | Vol USD acc. | Implied reading |
|---|---|---|---|---|
| House, Democratic control | 92.50% | +0.00pp | USD 7.79M | A single distinct value in the week, across 46 readings, with an intraday amplitude of 0.00pp. In 17 of the 29 instants with both sides in the same snapshot the pair exceeded 100.00%, and at the 101.00% peak the normalised figure fell to 91.58%. |
| House, Republican control | 7.50% | +0.00pp | USD 5.94M | Intraday top of 8.50% on 28 September at 14:30 UTC, with a 1.00pp amplitude in the week. It was this side that moved the pair sum, not the Democratic one. |
| Senate, Democratic control | 64.50% | +2.00pp | USD 3.39M | Seven closes at 62.50% and the entire move in Friday's session, with 64.50% first reached at 13:00 UTC on 2 October. Pair at 100.00% at both borders. |
| Senate, Republican control | 35.50% | -2.00pp | USD 2.40M | It began to give way a day before the Democratic side moved: it closed at 36.50% on 1 October. Intraday amplitude of 3.00pp, from 38.50% to 35.50%. |
| Republican state governorships, sum of the ranges below 24 | 55.00% | 57.20% normalised | USD 0.70M | Reading of 5 October at 03:04 UTC. The sum of the book's seven ranges is 96.15%, inside the gate, and the single most expensive range is 22 or 23 governorships, at 31.00%. |
| Election happens as scheduled | 98.20% | +0.00pp | USD 0.52M | Intraday floor of 97.80% on 29 September at 17:00 UTC and top of 98.30% on 26 September at 10:30 UTC, with a 0.50pp amplitude in the week. |
5.Liquidity and market structure
6.Electoral calendar through November
| Date | Sample | Why it matters | |
|---|---|---|---|
| 3 Nov, Tuesday | Midterm election day | 435 House seats and one third of the Senate | It is the only date on this grid that resolves the contracts. Every price cited in this issue is the implied probability of an outcome known on that day. |
| 28 Sep, Monday | The nine Senate seats most likely to flip party control | 9 state races | The Senate control contract is national and binary; the coverage beneath it is state by state. This brief does not convert one reading into the other. |
| 30 Sep, Wednesday | Congress leaves Washington for the campaign | both chambers | It is a calendar marker and not a price event: in this issue no contract movement is attributed to it, because the time window was not established. |
| 2 Oct, Friday | Republican groups retreat from the North Carolina Senate race and shift spending to Kansas | 2 state races | Two outlets covered the same shift on the same day, and a third treated Kansas as a new battleground. It is also the only day of the week on which the Senate contract moved. |
The last three lines are coverage dates, linking to the outlet and never to an aggregator. They enter as context for the period and not as cause: this brief only attributes repricing to an event when it can establish that the price moved after the event, and this week that ordering was not established for any of the items above.
7.Watch list, week triggers
- The House pair sum It exceeded 100.00% in 17 of the 29 instants measured during the week, peaking at 101.00%. While it sits above 100.00%, the price on the Democratic side understates the fall in its own implied probability, and the difference at the peak was 0.92pp. When it returns to 100.00%, raw and normalised coincide and the price becomes a direct reading again.
- The Senate after the week's close The move did not stop on Friday. In the certified reading of 5 October at 03:09 UTC, Democratic control stood at 67.50%, 3.00pp above the 2 October close, with the pair at 101.00% and therefore 66.83% normalised. The next brief measures the week of 5 to 9 October against the 2 October close.
- The leading fieldwork edge of the polling average It advanced from 22 to 28 September during the week, after seven days stuck, and has stood at 28 September since. While the fieldwork date does not advance, the average can move without any voter having been asked again, because an old round leaves through the border of the 30-day window: in that case the variation is one of composition and is not read as a movement of opinion.
- The distribution of state governorships It is the third question in this issue and the only one without a corresponding binary. The mass is concentrated in 22 or 23 Republican governorships, at 31.00%, and the ranges below 24 sum to 55.00% raw against 41.15% for the ranges of 24 or more. The book's sum is 96.15%, and every comparison here is made after normalising by it.
8.Methodology
Polymarket prices, read through the AFOS proxy with a live and uncached reading, and confirmed by a double reading eight minutes apart with a 0.20pp tolerance before any number goes live: in this issue the two readings came at 03:01 and 03:09 UTC on 5 October, with distinct timestamps, and the published values are those of the second. A daily close is the last point recorded that day in the series backup, and not the API reading, whose query window caps at 90 days and would hide the start of the Senate series, opened on 14 April. The floors and tops quoted as intraday come from the 46 readings per binary outcome recorded during the week. The weekly delta is measured against the PREVIOUS FRIDAY's close, 25 September, which is this series' convention. A binary pair is reported raw and normalised by the sum of the two outcomes, and that sum is declared whenever it leaves 100.00%: in this issue the House pair sum was measured instant by instant, pairing the two outcomes in the SAME snapshot, because a sum assembled from the two sides read at different moments is not the sum of a pair. A distribution is only cited when the ranges sum to between 95% and 105%: in this issue the state governorships book sums to 96.15%, and comparison between ranges is made after normalising by that sum. The generic ballot is a simple arithmetic average, unweighted and with no pollster excluded, on a 30-day rolling window, with one round per pollster per wave under the likely voter, registered voter and adults hierarchy. Decimal point in all three languages.
9.Additional reading · macro coverage
AFOS is the primary source of this analysis. The references below are complementary context reading and do not support any of the measurements above. ⛔ None of them is presented as the cause of a price movement: this brief does not attribute repricing to an event without establishing the time window.
- Politico · Republican groups retreat from the North Carolina Senate race, on 2 October., www.politico.com/news/2026/10/02/republicans-slf-scale-back-north-carolina-senate-spending-01104964
- NBC News · The same spending shift, with Kansas as the destination, on 2 October., www.nbcnews.com/politics/2026-election/gop-super-pac-pauses-spending-north-carolina-senate-race-shift-kansas-rcna601134
- Politico · Kansas treated as a Senate battleground, on 2 October., www.politico.com/news/2026/10/02/kansas-becomes-a-surprise-senate-battleground-01105125
- The Hill · An Alaska Senate race described as tightening, on 2 October., thehill.com/homenews/campaign/6126416-vance-rally-alaska-senate-race/
- The Guardian · The Alaska governor's race, which is the question behind the third book in this issue, on 2 October., www.theguardian.com/us-news/2026/oct/02/alaska-governor-candidate-jonathan-kreiss-tomkins
- The Washington Post · The nine Senate seats most likely to flip party control, on 28 September., www.washingtonpost.com/politics/2026/09/28/9-senate-seats-most-likely-flip-party-control-november/
- Axios · A field reading on Democratic strategy in the House, on 2 October., www.axios.com/2026/10/02/house-democrats-midterm-candidates-jeffries
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.