Prediction Market Basics

Prediction Markets vs Polls: What Each One Measures

Polls measure stated intention in a sample. Markets measure what people will risk money on. They answer different questions, and each fails in its own way.

ProudGlobeHead Markets DeskSeptember 16, 20264 min read

A clipboard beside a price line, separated by a divider

Every election cycle produces the same argument: the polls said one thing, the markets said another, and afterwards everybody explains why their preferred instrument was really right. The argument is mostly confused because the two things are not measuring the same quantity.

What a poll measures

A poll asks a sample of people what they intend to do or believe, and scales the answers up to a population using a model of who will actually turn out. It measures stated intention at a moment in time, filtered through a turnout model.

That is genuinely valuable. It is direct evidence about a population, it has a known methodology, and it comes with an error range. Its weaknesses are equally well known: sampling frames that miss people, response rates in the low single digits, respondents who say one thing and do another, and turnout models that are assumptions dressed as data.

Crucially, a poll's number is not a probability. "Candidate A leads 51-49" is not a claim that A wins 51% of the time. Converting a polling lead into a win probability requires a model with assumptions about correlated error across regions, and reasonable modellers disagree substantially about those assumptions.

What a market measures

A market asks nobody anything. It observes what price clears when people can buy and sell a claim on the outcome. The number it produces is already a probability, because that is what the price of a dollar-paying contract is.

Its participants can incorporate anything they like: polls, models, fundamentals, a private belief that the polls are biased, a view about turnout, news from twenty minutes ago. Whatever they believe, it costs them money to be wrong, and that cost is the discipline the mechanism relies on.

A funnel turning scattered dots into one line — Both methods compress many opinions into one number, by very different routes.

The honest comparison

Timeliness. Markets update continuously; polls update when somebody runs one. After a major event, a market reprices in minutes and a poll takes days. This is the clearest advantage markets have, and it is not close.

Aggregation. Markets can absorb information from any source, including the polls themselves. A market price is downstream of polling, not a rival to it. A polling average is one of the main inputs into the market's estimate.

Selection. Poll respondents are chosen to represent a population. Market participants choose themselves, and they skew towards people with money, interest and internet access. If the question is "what does the population intend", the poll has the better sample by construction. If the question is "what will happen", self-selection matters less than it sounds, because the mechanism rewards accuracy rather than representativeness.

Liquidity. A thin market with a few thousand dollars of interest is not the wisdom of crowds. It is the opinion of a handful of people with a spread. Before quoting a market probability at anybody, look at the volume. Where the wisdom of crowds breaks down covers this in detail.

Bias. Both have their own. Polls have house effects and turnout assumptions. Markets have the favourite-longshot bias, thin tails that stay overpriced because nobody wants to lock up capital to collect two cents, and occasional participants trading for reasons other than expected value.

Where they disagree, and what to do about it

When a market and a polling average diverge substantially, there are only a few explanations, and it is worth working out which one applies.

The market may know something newer. This is the common case when the divergence appears suddenly after news.

The market may be thin. Check volume and the spread before concluding anything. A 12-point divergence in a market with $3,000 of volume is noise.

The market may be pricing a different question. This happens more often than people expect. The polling average is about vote share; the contract might be about a certified result, a specific threshold, or a date. Read the rule.

The market may be applying a correction the poll cannot. Participants who believe polls have systematically missed a group will price that belief in. Sometimes they are right and sometimes they are not, but it is a real adjustment rather than a mistake.

Using both

The practical answer is to treat them as complements. Polls are evidence; markets are an aggregation of evidence with a price attached. When both point the same way, you have a stronger read than either alone. When they diverge, that divergence is itself information about where the disagreement lives.

This site tracks the market side continuously, including the cases where two venues disagree with each other about the same event — which is a smaller but much sharper version of the same question. The gap board is where that lives.

Prediction markets versus sportsbooks covers the other comparison people reach for, and the one where the structural difference is largest.

Educational content only. Nothing here is financial advice, a solicitation, or an offer to trade. Prediction markets can move against you and a position can lose its entire value. Where this guide describes a difference between venues it describes a theoretical cross-market edge in observed prices, which is not a profit: fees, spreads, capital lockup, transfer time and differing settlement rules all come out of it first.