How the tracker works
Every figure on these pages comes from the live order books of Polymarket and Kalshi, read every 5 minutes.
TruthIndex™
The TruthIndex™ is a weighted average of each venue's midpoint, the price halfway between its best bid and its best ask. Each venue counts in proportion to its weight L, the money committed by the orders resting in its book within 5 percentage points either side of its own midpoint.
TruthIndex™ = Σ (Pi × Li) / Σ Li
How to read this chart
The diagram works through the TruthIndex™ for one real outcome. For each venue, the dot is its midpoint, halfway between the best offer to buy (the bid) and the best offer to sell (the ask). The shaded box covers 5 points on either side of the midpoint, and the money in orders inside that box is the venue's weight, L.
The TruthIndex™ (the diamond) is the average of the two midpoints, weighted by L, so it sits closer to the venue with more money behind its price.
(62.5% × $226.1K + 61.5% × $879.5K) ÷ ($226.1K + $879.5K) = 61.7%
Each resting order is valued at the money behind it: an offer to buy at price p commits p per contract, an offer to sell at p commits 1 − p per contract, since every contract pays $1 if the outcome occurs. A venue with more money close to the current price carries more weight.
When one venue lists an outcome, the TruthIndex™ is that venue's midpoint, and the page marks it with the venue's name. When two venues quote an outcome and neither has orders within the band, the TruthIndex™ is the plain average of their midpoints. Each outcome is calculated on its own, so the outcomes of a contest can sum to slightly more or less than 100%.
Race ratings
Each race is rated from its leader's TruthIndex™: Toss-up under 60%, Lean 60% to 75%, Likely 75% to 90%, Safe 90% and above, labelled by the leader's party. Where several listed candidates belong to the leading party, their figures are added, so a race between two Republicans is Safe R. Where the highest-priced candidate has no listed party and a venue also prices the party outcome, the party outcome leads. A race led by a candidate from outside the two main parties is shown as Other leader. The maps are coloured by these ratings. Ballot measures are rated on the probability that the measure passes, with the same cutoffs on either side: Toss-up between 40% and 60%, then Lean, Likely and Safe Yes above it and Lean, Likely and Safe No below it.
Figures are the leader's probability, or for a measure the probability of the side ahead.
Seat projections
Projected seats count, for each party, the races where it is rated Safe, Likely or Lean, plus for the Senate the seats not on the ballot this year: 34 held by Democrats and the independents who caucus with them, and 31 held by Republicans. Toss-ups are counted separately. Expected seats add up each party's probability of winning across every priced race, so a 60% race counts as 0.6 of a seat; a party's probability is its party outcome where a venue prices one, otherwise the sum of its candidates. The share left to independent and third-party candidates, and to races with no price, is shown as the remainder.
Changes and timeline
The 24-hour and 7-day changes compare the current TruthIndex™ with its value 24 hours and 7 days earlier. Rating changes compare each race's rating today with its rating on the first day of the past week, worked out from every outcome's TruthIndex™ that day. The map timeline replays the TruthIndex™ and each venue's daily closing prices over the past 60 days. Kalshi's daily price is the midpoint of its closing bid and ask, kept when the two lie within 5 points of it.
Outlook and margin markets
Seat-count, balance-of-power, popular-vote, wave and margin-of-victory markets are shown venue by venue. The two venues define these questions with different ranges and wording, so they are never combined.
Which venues are counted
A venue is counted for an outcome when its book holds both bids and asks, its best bid sits below its best ask, and both lie within 5 points of its midpoint. A book with bids only or asks only, a bid and ask more than 10 points apart, a best price outside 0% to 100%, or a bid and ask both at 0 (the venue's marker for no quote) is left out of the TruthIndex™ for that outcome. Each contest page states, venue by venue, whether the venue was counted and why.
Where no venue can be counted because the book holds bids only or asks only, the page shows the implied probability instead: the best ask when there is no bid, the best bid when there is no ask.
| Status shown | When |
|---|---|
| Counted | The book holds bids and asks, the best bid is not above the best ask, both lie within 5 points of the midpoint, and orders rest within the band. |
| Not counted: no order book | The venue has no order book for the outcome. |
| Not counted: invalid prices | A best price lies outside 0% to 100%. |
| Not counted: bids only or asks only | The book holds bids only or asks only. Where no venue can be counted, this gives the implied probability. |
| Not counted: bid and ask both at 0, no quote | The best bid and the best ask are both 0, the venue's marker for no quote. |
| Not counted: crossed book | The best bid is above the best ask. |
| Not counted: bid and ask more than 10 pts apart | The best bid or the best ask lies more than 5 points from the midpoint. |
| Not counted: no orders within 5 pts of its midpoint | The quote is usable but no order rests within 5 points of the midpoint. Its midpoint is used only when no venue carries weight: on its own, or in a plain average with the other venue. |
Spread
Spread is the largest absolute gap between the Polymarket and Kalshi midpoints for any outcome in a contest, in percentage points, counting venues that pass the checks above. Tables open sorted by spread, widest first.
24-hour change and history
The TruthIndex™ is recorded at every refresh. The 24-hour change compares the current value with the value recorded 24 hours earlier. On the election night board, the one-hour change compares the current value with the value recorded an hour earlier, from every refresh of the last two hours. The Polymarket and Kalshi lines on each chart show each venue's own daily price history.
Coverage and matching
A contest appears when Polymarket or Kalshi lists a market on it: Senate, governor and House races, chamber control, state ballot measures and mayoral races on the 3 November 2026 ballot. Outcomes are paired across venues only when both venues ask the same question: party with party, candidate with candidate, Yes with Yes. The question is read from each market's resolution rules: a Kalshi market labelled with a nominee's name that resolves on the winning party is a party market. Where one venue prices which party wins and the other prices which candidate wins, the two are shown separately.
Resolution rules
Each venue writes its own settlement rules. They can differ on the source used to call a race, the treatment of runoffs and recounts, and the settlement date. Every contest page reproduces both venues' rules.
Archive
After certification of the last covered contest, the pages are frozen as a permanent record of final pre-election prices and outcomes.
Analytics only, no execution, no custody, no positions.