How to read election markets without confusing odds with vote share

The most common misreading in prediction markets happens on election night. A candidate priced at 60 is not expected to win 60% of the vote. The contract asks whether the candidate wins at all, and 60 is the market's probability of that binary outcome. A candidate expected to win 52% of the vote in a stable race might trade at 85, because 52% is a comfortable margin. A candidate expected to win 50.5% might trade at 55, because that margin is inside the noise.

What you are predicting

Read the rule. Common election contracts ask, in order of how often they are confused:

• Who wins the office. Settles on the certified result, or on a set of news organizations calling the race, depending on the venue.
• Which party controls a chamber. Settles on the majority after the election, sometimes with specific language about independents who caucus with a party, or about what happens if control changes by resignation before the new Congress is seated.
• Who wins a nomination. Settles at the convention or on a formal declaration, and a candidate who leads in polls can trade well below the front-runner in the nomination contract because the party process is a separate question.
• Margin and turnout. Ladder contracts on the popular-vote margin or on a state's result. These are the ones that actually predict vote share, and they are read very differently from win contracts.

Which evidence matters

Polls move prices, but a single poll rarely should. Markets respond to changes in the polling average, to fundraising and ballot-access news, to primary results, and to legal or health events. Late in a race, early-vote data and turnout reports move prices. On election night, the price moves on the count itself, and by then a contract is a bet on the count, not on the electorate.

Verified official developments, such as a candidate withdrawing, a court ruling on ballot access, or a certified result, are the only things that settle anything. Everything else is a price change.

Early presidential markets are a leaderboard

A contract on the 2028 presidential winner priced at 24 two years out reflects name recognition, the plausibility of a nomination, and the willingness of a few traders to hold a position for two years. It is not a forecast in the sense a polling model produces one. Read the field as a ranking of who the money considers viable, with the actual probabilities being far less precise than the decimals suggest. Long-dated contracts also carry a cost-of-capital drag that pushes unlikely names slightly above their true chance.

A worked example

Suppose the Senate control contract shows Democrats at 52.5 and Republicans at 48.5 on Polymarket. The two do not sum to exactly 100 because each is its own contract with its own spread. The market is saying control is close to a coin flip, slightly favoring Democrats. It is saying nothing about how many seats either party wins. A separate ladder contract on the seat count would tell you that, and it may not exist or may be thin.

Now suppose a well-regarded poll shifts one competitive state by three points. The control contract might move two or three points. That is the right size: one state is one seat among several that decide control. A move of ten points on one poll is a market being thin or reacting, not new information arriving.

Beginner mistakes

• Reading a win probability as a vote share.
• Comparing a nomination contract with a general-election contract as if they asked the same question.
• Treating a two-year-old contract's decimals as precision.
• Ignoring that "control" contracts have rules about independents and vacancies.

What to take away

• A win contract prices a binary outcome; only margin ladders predict vote share.
• The size of a price move should match the size of the news.
• Long-dated fields are rankings with a cost-of-capital drag, not forecasts.
• Control contracts settle on rule text about majorities, not on the headline.