Markets Are Not Polls

A prediction-market price and a poll percentage can be written with the same numeral and still be different objects. The market figure is a price on a claim that usually pays a dollar if named rules say an event happened, and nothing if they say it did not. A poll percentage, as a U.S. regulator uses the word "polling," is a different forecasting form. It is not that claim. Treating the two as interchangeable is a category error.

What this page does not claim. It does not claim that markets beat polls. It does not claim that live venues are calibrated. It does not invent a poll number or a live book. It does not tell you to buy, sell, or fade either figure. The 65-cent and 70-cent figures used below are documentation examples published by a venue and by the CFTC, retrieved 2026-08-15T00:40Z, not live quotes.

The market object

On the usual yes-no event contract, a YES position pays about a dollar if the contract rules say the event happened, and zero if they say it did not. The CFTC describes that structure as a fixed payout, "usually $1," with an expiration. Its own illustration, retrieved 2026-08-15T00:40Z, is a "Will it rain tomorrow?" contract at 70 cents for yes and 30 cents for no. Traders who are correct receive the payout. Profit is payout minus purchase price. Taxes and fees can change the return.

One venue's public docs, retrieved the same hour, put the same object in tokens. Each binary market has two ERC-1155 outcome tokens. Yes redeems $1.00 if the event occurs. No redeems $1.00 if it does not. Every Yes/No pair in existence is backed by exactly $1 of collateral. Prices sit between $0.00 and $1.00 and, the operator writes, "emerge from supply and demand as users trade with each other" on a central limit order book. The operator's worked example: if Yes shares are trading at $0.65, "the market believes there's approximately a 65% chance the event will happen." That is a documentation example, not a live fill. The arithmetic is straightforward because the documented payoff is a dollar. The interpretation is a separate guide.

You can sell the position before settlement. The CFTC states that as a feature of the product: customers can trade in and out at the current market price. A poll percentage does not have an exit.

The poll object — only as far as opened sources go

The CFTC page retrieved 2026-08-15T00:40Z names "polling" once, as another form of forecasting. It does not define a poll. It does not publish a poll percentage, a sample, a field date, or a question. This page opened no polling-methodology primer. So this page will not invent how a poll was sampled, weighted, or asked.

What can be said without invention: a poll percentage is not a token. It is not a bid or an ask. It is not minted from a dollar of collateral. It is not redeemed by an oracle against a resolution source, an end date, and a list of edge cases. It is not a price someone is offering to trade right now. When a headline writes both figures as "65%," it has used one display convention for two measurements.

The numeral is not the object

Vendor pages map $0.25 to 25%, $0.50 to 50%, $0.75 to 75%. They map $0.65 to "approximately a 65% chance." The CFTC maps 70 cents to the market's expectation on its rain illustration. Those mappings are about a $1 claim. A poll percentage, when one exists, is not produced by that mapping. Identity of the numeral is not identity of the object. This page stores no poll that reads 65%, 70%, or any other figure.

The number on the screen may not be the number that trades

On the venue that published the prices-and-orderbook page retrieved 2026-08-15T00:40Z, the displayed price is the midpoint of the best bid and the best ask, unless the spread is wider than $0.10, in which case the last trade is shown. Their own example is a $0.34 bid and a $0.40 ask, which displays as $0.37. A buyer pays the ask. A seller receives the bid. That display rule is that venue's. The CFTC says that in most cases the order book shows bid and ask; it does not say the headline number is the midpoint.

A poll percentage is not a mid, a last, a bid, or an ask. If you cannot see bid, ask, and time on the market side, you do not know which market number you are lining up next to a poll — and you still would not have the same object.

The contract pays on the rules, not the title

A market price is a price on the contract that will be resolved. The venue's resolution page, retrieved 2026-08-15T00:40Z, says the title describes the question and the rules define how it resolves — source, end date, edge cases. Resolution on that venue runs through the UMA Optimistic Oracle: anyone may propose, anyone may dispute. Undisputed resolution is described as about two hours after proposal. Disputed cases are described as days. The CFTC tells customers they are entitled to know who decides settlement, and how, and to review market- and contract-specific rules.

A poll percentage, when one exists, is an answer to the question that was asked of a sample. This page retrieved no paired poll question and contract rule, so it invents no pair. The point does not need a pair: price proximity is not equivalence, and a poll wording is not a resolution source.

Fees sit inside the market number

As of the fees page retrieved 2026-08-15T00:40Z, takers on most categories pay fee = C × feeRate × p × (1 − p). Makers are not charged. The USDC fee peaks at a 50¢ price. Geopolitics is listed as fee-free. Documented taker rates on that page: Crypto 0.07; Sports, Economics, Culture, Weather, and Other 0.05; Finance, Politics, Mentions, and Tech 0.04. The same page's 100-share table at the $0.65 documentation row lists $1.59, $1.14, or $0.91 depending on category. Those are documentation figures, not an invoice for a market we have not named.

The CFTC tells customers that taxes and fees can change the return, and to understand how costs affect returns over time. A poll percentage does not have a taker fee. That is another object difference. It is not a reason to prefer one number.

"Markets beat polls" is not established here

The CFTC page says prediction markets "can sometimes forecast event outcomes better than polling or other forms of forecasting." The sentence has no study, no sample, no date, and no event class. The venue docs opened for this packet do not compare markets to polls. No accuracy study was opened. Whether markets beat polls is UNKNOWN. This page will not say they do. It will not say they do not.

The CFTC also writes that customers may have little or no protections if they trade with unregistered entities that operate outside the United States. This page does not apply that sentence to a named venue. Registration status of the venue in the docs above is UNKNOWN here.

What to do with two 65s in one headline

Read which object each number is. One may be a displayed midpoint on a $1 claim. One may be a poll percentage this page has not sourced. Check whether the contract rules match the question you thought was being asked. Remember that spread and fees sit inside the market number. Do not average the two figures. Do not substitute one for the other. Do not treat a documentation example as tonight's book.

Whether one class of number has been more accurate than the other, on the events you care about, in 2026, is UNKNOWN. The 65% guide owns when a single price can be read as a probability. The complementary-prices guide owns why Yes and No should add to about a dollar. This page owns only the category error.