Prediction markets are, in essence, a technology for turning opinions into prices, and the midterm race is their most public test. The tracker that puts polls and prediction markets side by side for every competitive Senate and governor race, refreshed through the day, shows how the two systems work. Most of the time they agree. The rows marked as split are the races where they do not, and those rows are where the technology is most interesting, because they reveal what traders know, or think they know, that polls have not yet captured, or vice versa.
How a price is made
Each price is what a contract paying one dollar if a party wins the chamber costs at the update time shown, so 63 cents means traders put the chance at about 63 percent. In the latest update, on October 26 at 1:50 a.m. Pacific time, with the newest poll dated October 23, Senate Democrats were priced at 63 cents on Kalshi and 65 cents on Polymarket, while House Democrats were priced at 90 cents and 92 cents respectively. Change is measured against 24 hours earlier, and the Senate price moved down by one or two cents while the House price held or rose by one cent. Those small moves are the market digesting new polls, new fundraising reports and new campaign news in real time, in a way that a weekly polling average, by design, cannot.
Why traders and pollsters disagree
Traders and pollsters disagree for structural reasons, not only because one group is wrong. Polls sample voters and ask what they intend to do, with all the limits of sampling, wording and timing that implies. Markets sample people willing to risk money, who may be better informed about a race but who are also a self selected group with their own biases. When a race is marked as split, it usually means one of three things: the polls are lagging a shift the market has already priced, the market is overreacting to a single event the polls will correct, or the race is genuinely uncertain and both systems are guessing. For technologists, the lesson is that neither system replaces the other. The useful product is the comparison itself, a dashboard that shows where opinion and money agree, where they diverge, and how that divergence moves as election day approaches.
