How to read the prediction markets on election night 2026

On November 3, 2026, the two contracts that will move most are the ones on which party controls the Senate and the House when the next Congress is seated. They settle on control, not on seat counts, and on the source the venue names in its rule text, not on a network's projection. The price at 11 p.m. Eastern is a probability built from partial counts, and partial counts have a known way of misleading.

What the control contracts settle on

Read the rule on the tracker before the night starts. The Senate contract asks which party holds the majority, and the rule says how independents who caucus with a party are counted and what happens at 50-50, where the vice president's vote decides organization. The House contract asks the same about 218 of 435. Neither pays on the number of seats: a party that gains seats and misses control loses the contract, and a party that loses seats and keeps control wins it. A seat-count ladder is a separate contract, if the venue lists one.

Each venue also names its source and its moment; what follows is how they read as of September 2026. Kalshi determines the result itself against the source its rule names, and its political contracts have settled within a day of the networks' calls when the rule allowed it and the outcome was beyond doubt. Polymarket's rule text names its own criteria, often a date in January when the new Congress convenes, with its resolution process able to run earlier when the answer is settled. The difference is not academic on a night when a runoff or a recount is possible, and how settlement works explains why "proposed" and "final" are different states. Thirty-five Senate seats are on the ballot, including two special elections, and a state that requires a majority to win can push one of them to a December runoff. The control contract may not have an answer on November 3 even if every network does.

How the price moves as the count comes in

Polls close in waves, starting at 6 p.m. Eastern and ending with Alaska after midnight. Early in the evening the market is reacting to two things: the first returns from Eastern states, and the gap between those returns and what the pre-election price implied. A competitive Eastern race that comes in a few points better for one party than expected moves the control contract before any Western polls close, because traders read it as information about turnout everywhere.

Then the counting order starts to matter more than the votes. States differ in when they count mail ballots. Where mail ballots cannot be processed until election day, in-person votes report first and the early tally leans toward the party whose voters show up in person, then shifts as mail ballots are added. Where mail ballots are counted first, the early tally leans the other way and shifts back. The 2020 vocabulary for this was "red mirage" and "blue shift"; the mechanism is procedural, not partisan, and it runs in whichever direction the state's counting order sends it. A contract that moves 15 points on the first 20% of a slow-counting state's vote is pricing the order of the count, not the result.

Overnight, books thin out. Spreads that were a point wide at 8 p.m. can be five wide at 3 a.m., and a single order can print a price no one else would pay. A print is not a consensus. Reading the odds at that hour means reading the bid and the ask, not the last trade.

Why 95% is not a call

A contract at 95% has not resolved. It is priced as near-certain, on the evidence of the moment. Networks call races on models of the outstanding vote; markets price them on the same public information plus a guess about what other traders know. The market is also allowed to sit at 95% for hours and then settle NO, because a price is a probability and 5% events happen one time in twenty. When the count is slow and the outstanding vote is from a known place, a 95% price can be well founded. When the count is slow and the outstanding vote is from an unknown place, the same price is a guess with a confident face. The rule text pays; the tape does not, and odds are not vote share.

What to watch state by state, without listing every state

Think in types of state rather than names.

• Early-closing, fast-counting states. These are the calibration set. The question is not who won them but whether the margins match what the market expected.
• The three or four races that decide control. Identify them the week before from the individual race contracts.
• Slow-counting states with large mail votes. Do not read the first hours of the count as a trend. Wait for the point where a network says what share of the vote is outstanding and where it is from.
• Runoff and recount states. Note the margin at which an automatic recount triggers and whether a majority is required. These are where the control contract can stay unresolved for weeks.
• The House. The House contract moves on the models and the networks' aggregate calls, which arrive late and change. A House price at midnight is a price on a projection.

When contracts actually resolve

Media calls are inputs; the rule text is the output. A contract that resolves on certified results waits for the states to certify, which runs into December, and for any runoff. A contract that resolves when the outcome is beyond doubt can pay earlier, on the venue's determination. On Polymarket, a proposed resolution enters a challenge window before it is final, and a disputed one can sit for days. Nothing on this site reports a control result until the venue records it. Check the venue's own page before relying on it. The 2028 presidential field will reprice on the midterm results the same night, and it will still be a leaderboard two years out.

A worked example

Suppose that the week before, control of the Senate comes down to four toss-up races, each priced at 50%, and one party needs three of the four. The chance of at least three of four independent coin flips is 5 in 16, so a fair control price for that party is about 31%. Now watch what a single call should do.

Race called	Wins needed from remaining	Fair control price
None yet	3 of 4	31%
First race won	2 of 3	50%
First race lost	3 of 3	12.5%
Two won, none lost	1 of 2	75%
Two won, one lost	1 of 1	50%

A call on one 50-50 race moves the control contract by roughly 19 points up or 19 points down. A move of that size on a single call is the right size. A move of that size on 8% of the vote in that race, before any call, is the market pricing the counting order; the table says how much it should give back. The races are not really independent, since a turnout surprise runs across states, which is why the real move on the first call is often larger than the arithmetic. It is still the arithmetic to check against, and price is probability is the whole method.

What to watch

• The rule text on both venues for both chambers, read before 6 p.m. Eastern, including the 50-50 and independents language.
• The counting order for each decisive state: mail first, or mail last.
• The bid and the ask, not the last trade, after midnight.
• The share of outstanding vote and where it is from, before believing any price above 90%.
• Whether a runoff or recount is possible in a decisive race, which means the contract will not resolve tonight.
• Our catalyst calendar for certification dates, and the markets page for which contracts are trading real volume rather than printing.

Nothing on this page is investment advice, and the venue's current rule text and fee page are the only things that pay.