Predict.fun review

Predict.fun is a non-custodial order-book market for binary event contracts on BNB Chain, collateralised in USDT. You keep your own keys; the venue never holds your balance, which removes the single largest failure mode of a custodial exchange and replaces it with the ordinary one of looking after a wallet.

Two commercial features stand out. The maker fee is zero, so posting liquidity costs nothing but the spread you are offering. And the venue pays yield on USDT collateral while it sits against an open position — money that would be idle on most venues here earns while a contract runs, which changes the arithmetic on long-dated markets more than it sounds.

The operating entity is named as PDF Labs Limited, which is more than most on-chain venues disclose. It does not state a jurisdiction of incorporation anywhere we could read, holds no licence, and lists no regulator.

Six of the seven categories we track are covered — the widest verified breadth of any venue we built from primary sources, taken from the site’s own category navigation: sports, esports, politics, crypto, finance, economy, culture, plus weather and commodities.

That is genuine range rather than a marketing list, and it is the main reason to look here rather than at a single-theme venue. Read it alongside the regulation rating on this page rather than on its own, though: on an unlicensed venue, breadth means more places to be exposed to the same counterparty, not more protection.

We have not measured active market counts or listing speed, so the rating reflects verified category coverage only. A category with one stale market in it counts the same as one with fifty in that figure, which is a limitation of the measurement rather than a claim about this venue.

This is the finding that decides the review. Predict.fun’s user-facing knowledge base describes settlement through UMA’s optimistic oracle, with bonds, a challenge window and escalation — “If that is also disputed, the question escalates to a full UMA token holder vote”. Read on its own, that is a serious, well-understood mechanism with a public dispute record behind it.

The developer documentation describing the oracle actually deployed says something else. In a bullet list it states “Whitelisted proposers for price proposals” and “No bonding or dispute mechanism (simplified trust model)”. A whitelisted proposer with no bond and no dispute route is the opposite arrangement: there is no challenge to make and no stake behind the answer.

Both pages were live on the same day, 9 September 2026, and we could not reconcile them from any primary source. The practical consequence for a trader is precise: you cannot establish from the venue’s own documents whether a settlement you disagree with can be challenged at all, and that uncertainty exists before anything goes wrong.

The same developer page records that the operator layer “Supports flagging/unflagging questions for emergency resolution by admins”. A second override exists on the Chainlink price markets, where an account holding EMERGENCY_CLOSE_ROUND_ROLE “can call emergencyCloseRounds to settle rounds manually”.

Neither of these is unusual in on-chain markets, and an emergency stop has legitimate uses — a broken data feed has to be stoppable by someone. What matters is that they exist, that they are documented in the developer pages rather than in the terms a user agrees to, and that they sit above whichever of the two resolution mechanisms is the real one.

Our advice is narrow and practical. If you trade here, treat the settlement of any individual market as an operator decision rather than an oracle outcome, size positions on that basis, and read the market’s own rules before committing — not the general documentation, which is the part that disagrees with itself.