A poll is a sample. A price is a stake. Only one of them is scored.
Why a number with money behind it is built differently from a number with a pundit behind it, and what to do when the two disagree.
Every election cycle, every rate decision, every awards season produces two kinds of number about the same event. One comes from a poll or a panel: a sample of people asked what they think, published a few days later, with a margin of error that covers the sampling and nothing else. The other comes from a prediction market: a price, updating continuously, produced by people who have put their own money on being right.
Most coverage treats these as competing forecasts and asks which one is correct. That is the wrong question, and it hides the useful one. They are not two guesses at the same thing. They measure different things, under different conditions, and the difference between them is where the information is.
This page explains what each number actually measures. The guide behind the form goes further: where markets are weak, how to read a price as a probability rather than a prediction, and the four explanations for a poll and a price disagreeing, in the order worth checking them.
Get the guide
What The Polls Won't Tell You (But The Markets Will). A reader's guide to the difference between a sample and a price. Sent to your inbox as a PDF.
Check your inbox. The guide is on its way. If it has not arrived in a few minutes, look in your spam folder or email hello@ipredicta.co.
That did not go through. You are not subscribed and the guide has not been sent. Please try again in a moment, or email hello@ipredicta.co and we will send it by hand.
18+ · Not financial advice · Trading involves risk of loss
What a poll measures
A poll is a defined number of respondents, chosen by a method that approximates the wider population, asked a fixed question during a window of days, with the results published some time after the window closes.
Four things limit what it can tell you. The people who answer are the people who could be reached and agreed to talk, and response rates have been falling for twenty years. What they report is stated intention, which is a different quantity from what they will actually do. The fieldwork window closes before publication, so anything that happens after it is absent by construction. And nothing is at stake for the respondent. There is no cost to an answer that is casual, aspirational or untrue.
None of that makes a poll worthless. It makes a poll a measurement of a specific thing at a specific time, and the specific thing is not the outcome.
What a price measures
A price on a prediction market is not a sample. It is a continuously updated number produced by everyone willing to commit capital to a position, weighted by how much they commit, and bound to a written rule that says exactly what the contract pays on.
That gives it four properties a poll cannot have. It updates continuously, so there is no window. It is weighted by conviction rather than by headcount, which is not democratic and is not meant to be. It is bound to a rule, so it is a claim about a precisely specified event rather than a headline. And it is scored. Every position resolves, and every participant learns, in money, whether they were right.
The result is a number that measures something a poll does not: the aggregate belief of people who have accepted a cost for being wrong.
Why the incentive matters more than the intelligence
Consider two people making the same forecast. The first is a commentator. They state it publicly and confidently, because confidence is what the format rewards. If it fails, nothing is deducted. If it succeeds, it may be remembered. The asymmetry favours boldness.
The second holds a position in a market. They have made the same forecast by committing capital at a price. If it fails, the capital is gone. If it succeeds, they are paid in proportion to how unlikely the outcome was. The asymmetry punishes overconfidence directly.
This is not an argument that market participants are cleverer than commentators. Frequently they are the same people. It is an argument about conditions. The same person, forecasting under a cost for error, forecasts more carefully than they do without one. A price is the aggregate of many such careful forecasts, and it inherits their discipline.
The price contains the poll
The framing of polls against markets is wrong, and this page has used it only for clarity. A price is not a competitor to a poll. It is a consumer of one.
When a poll is published, participants read it, judge how much weight it deserves against everything else they know, and adjust. The price after publication contains the poll, discounted by the market's collective view of its reliability. The same is true of the turnout model, the base rate, the morning's news and the commentator's opinion. All of it is input. The price is the output.
That is why a poll and a price disagreeing is informative rather than contradictory. The market has seen the poll. If the price has not moved to match it, participants have judged, with capital, that the poll is measuring something other than the outcome, or measuring it badly, or that something else outweighs it.
They may be wrong. Markets misjudge polls, and thin markets misjudge everything. But the disagreement is a considered one, made by people who paid to make it, which is a different kind of disagreement from the one on a panel.
What this does not mean
None of this establishes that any price is correct. Markets are wrong for reasons the guide catalogues and for reasons nobody has, and capital committed to any market is capital exposed to loss. A reader who understands a price better than the panel does can still be on the wrong side of it, and frequently will be.
What it establishes is that a poll and a price are different instruments, and reading one as a bad version of the other misreads both.
The guide goes further
What The Polls Won't Tell You (But The Markets Will) covers what this page has not: how to read a price as a probability rather than a prediction, the four conditions that make a market unreliable and how to spot each, and the four explanations for a poll and a price disagreeing, in the order that costs least to check.
Illustrative examples throughout, no live prices and nothing that dates.
Check your inbox. The guide is on its way. If it has not arrived in a few minutes, look in your spam folder or email hello@ipredicta.co.
That did not go through. You are not subscribed and the guide has not been sent. Please try again in a moment, or email hello@ipredicta.co and we will send it by hand.
18+ · Not financial advice · Trading involves risk of loss