Opinion vs Predict.fun: two BNB Chain order books, one USDT, fees that cross at 71 cents

Same chain, same token, same shape

On the cards the two look like one product twice. Both are central limit order books on BNB Chain with outcome tokens that redeem for one unit of collateral; both run offshore, under a company in Panama or the British Virgin Islands; neither checks identity at signup; both list sports and esports fixtures beside crypto, macro and politics; both have a WebSocket and a Python SDK. A reader who has narrowed the offshore field to BNB Chain is stuck between exactly these two, and the question is where they actually differ.

The dollar is not one of the places. Opinion's public API and Predict.fun's SDK name the same collateral contract, 0x55d398326f99059fF775485246999027B3197955. It reports itself as "Tether USD"; it is the Binance-pegged BSC-USD token, and Tether's own list of supported chains gives no USD₮ contract on BNB Smart Chain. Whatever that token's backing means for a holder — set out in what backs an on-chain venue's dollar — it means the same thing on both venues. Choosing one over the other does not change which dollar you hold.

What happens to it while you wait does differ. In markets marked Yield, Predict.fun supplies the USDT behind open positions to Venus Protocol's lending market, pays the interest out weekly and keeps 10% of it. That adds a second dependency to the token: the yields page warns that in a liquidity crunch on Venus's USDT pool, winnings in yield-bearing markets may be unclaimable for a while, and that Predict.fun may pull its funds out of Venus and convert those markets until it passes. Opinion's card records no yield and no lending. The rate on Predict.fun is not fixed, and this page does not value it; what it changes is that a winning position in a Yield market can be waiting on a lending pool as well as on the oracle.

Both cards were read on 23 October 2026. Opinion's terms are dated 19 October 2025, Predict.fun's 18 November 2025, and Predict.fun's card draws partly on Wayback captures of its pages.

Who decides, and how you argue

This is the difference that does not show in a fee field.

• Predict.fun hands resolution to someone outside the company. Every market created since March 2026 resolves through UMA's optimistic oracle: anyone may propose an outcome with a bond, an undisputed proposal settles after a challenge window of about two hours, a dispute posts a counter-bond and starts a second round, and a second dispute goes to a vote of UMA token holders that can take up to a week. The recurring BTC Up/Down rounds settle instead on Chainlink Data Streams reports, with an emergency role that can close rounds by hand while the contract is paused. The daily stock and index Up/Down markets have no source named in any document the card found.
• Opinion keeps it in-house. The method is named on each market page, and the documentation says most markets are resolved by Opinion AI, the operator's own multi-agent AI oracle. The public API flag for it was set on 27 of 172 activated markets on 23 October 2026; the rules text on the others, mostly sports and esports fixtures, names a source of its own, such as HLTV for a CS2 match. Opinion's own recurring Up or Down markets name theirs: a Chainlink data stream for the hourly BTC, ETH and BNB series, Yahoo Finance's official close for the daily stock series. A proposed result can be challenged once, within a window of typically two hours, by staking OPN — the greater of 1% of the market's open interest and 1,000 OPN. An accepted dispute returns the stake and an equal reward, a rejected one forfeits it, and a dispute the review does not answer in time succeeds automatically.

So the recourse is different in kind. On Predict.fun a wrong proposal can be fought by anyone who posts a bond, and the last word is a vote the company does not control. On Opinion the last word is a review the cards do not attribute to anyone outside the operator, and the right to ask for it is one dispute per resolution, first come, paid in a token you have to hold. Both venues' terms say the company is "not involved in nor responsible for the resolution" of any market, which on Opinion sits beside documentation that calls the company's own AI the primary oracle. Who decides the outcome covers the general shapes.

Two fee curves that cross

Both publish a taker fee that peaks at 50 cents. Both cards normalise it to the effective fee at fifty cents — Opinion 0.5, Predict.fun 1 — and that pair of numbers is the least useful thing to compare, because the two formulas are different shapes.

• Predict.fun charges 2% × min(price, 1 − price) per share. At any price up to 50 cents that is exactly 2% of what the shares cost; above 50 it falls with the other side's price, to 0.2 cents a share at 90 cents.
• Opinion charges topic_rate × price × (1 − price) on the notional of the fill, with the top of the published range a 4% base rate — at most 1% of the fill at 50 cents and 0.36% at 10 or 90 cents. Because it is a share of the notional, it is not symmetric per share: at 10 cents it is 0.036 cents a share, at 90 cents 0.32.

Put per share, Opinion's top rate is 0.04 × price² × (1 − price) against Predict.fun's 0.02 × min(price, 1 − price). Below 50 cents Opinion's curve is at most half of Predict.fun's. Above 50 the ratio climbs, the two meet at about 71 cents (where price squared is a half), and above that Opinion's curve is the dearer one. At 90 cents it is 0.32 cents a share against 0.2. Individual Opinion markets can carry a lower topic_rate, which only moves this in Opinion's favour.

Then Opinion's floor reverses the bottom of the range. Every Opinion taker order pays at least $0.25, charged on its first fill whenever the curve fee comes to less. Predict.fun has no fee floor, only a 1 USDT minimum order. Since Predict.fun's fee is 2% of the stake at any price up to 50 cents, it stays under $0.25 on any order under $12.50. So, at or below 50 cents:

• under $12.50 a taker order, Predict.fun is cheaper, because Opinion charges the full $0.25 — on Opinion's own $5 minimum order that is 5% of the stake, against 10 cents on Predict.fun;
• above $12.50, Opinion is cheaper or level, first at the $0.25 floor and then, from $25 at 50 cents, at a curve that is at most half of Predict.fun's.

Above 50 cents the break-even order grows, and from about 71 cents Predict.fun is cheaper at every size. Discounts on both venues come from how the account was opened — a referral binding of up to 10% on Opinion, an invite discount of 10% or a Binance Wallet rebate of 5% on Predict.fun — and none of them can take Opinion below its floor.

The resting side is where Opinion pulls clearly ahead. Opinion pays a maker 50% of the curve fee the taker generated, per fill, daily, with no sign-up: 0.25 cents a share at 50 cents, the −0.25 in its maker field. Predict.fun charges a maker nothing and pays nothing. Its 25% maker rebate was a trial on crypto Up/Down markets dated to end on 16 September 2026, and the page describing it still said only that on 23 October; the Predict Points it gives for resting liquidity are points, not a rebate. Who is paid to quote sets these programmes beside the rest of the category.

Where each is closed

Both bar the United States, the United Kingdom, France, Ontario, Singapore, Poland, Thailand and Taiwan, and the sanctioned territories. Both reach citizens as well as residents — not in the notice at the top of their terms, but in the clause further down that defines the restricted jurisdictions — both prohibit VPNs, and both can put a wallet into close-only mode. Neither checks identity at signup; both reserve the right to ask later.

The lists then part in two places, and for a reader there these decide the page:

• South Korea, mainland China and Hong Kong are barred by Opinion and not by Predict.fun.
• Australia and Belgium are barred by Predict.fun — in the notice at the top of its terms, though the detailed clause further down omits them — and not by Opinion.

Why a venue is unavailable where you are covers what these restrictions consist of.

Wallets, gas and code

• Wallets. Opinion connects an existing wallet (Binance Wallet, OKX Wallet, MetaMask, Backpack, Coinbase Wallet) or a social login, pays the gas for matching and settling, and leaves you the gas for disputes and claims. Predict.fun creates a smart-contract wallet through Privy and ZeroDev from an email, Google, X or Web3 login, pays no per-trade gas, lets you export the key, and documents how to withdraw by calling the ZeroDev contract directly if the site is down.
• Code. Opinion's market-data REST endpoints need no key, at 5 requests a second (15 with a key made by wallet signature); orders go through its Python and TypeScript CLOB SDKs. Predict.fun's REST API is in beta and its mainnet key is requested by support ticket on Discord, for 240 requests a minute; its order-signing SDKs are MIT-licensed. For reading prices without asking anyone first, Opinion is the open one.
• Books. Both cards found liquidity concentrated in a few markets: empty books on the day's new esports fixtures on Opinion, and $20 and $707 traded on two League of Legends Worlds matches on Predict.fun. Quote your own order on the market you want before any fee figure above matters.

Which to use, and when to switch

Start with where you are. In Australia or Belgium only Opinion of the two is open. In South Korea, mainland China or Hong Kong Opinion is closed, and Predict.fun is the one whose terms, as its card records them, do not name you.

Where both are open, use Opinion for taker orders above $12.50 at prices up to 50 cents, for larger ones up to about 71 cents, and for any order you intend to rest. Its curve is at most half of Predict.fun's below 50 cents, its makers are paid on every fill, and its market data is the one you can read without a ticket. The cost is resolution by the operator's own process: one staked dispute, and a review.

Switch to Predict.fun the moment any of these is true: your typical taker order is under $12.50, where Opinion's $0.25 floor is the whole price; you are buying above about 71 cents; you want the 5- and 15-minute BTC Up/Down rounds, where Opinion's shortest crypto Up or Down market is an hour; or the market is one you would want to contest, because on Predict.fun the outcome goes to an outside oracle and, at the end, a vote the company does not control. If you take a Yield market there, count on the second dependency as well: a winning claim can wait on Venus.