DeepBook Predict is an on-chain binary options protocol on Sui. What that means in plain English — the contract, the order book, the mint-to-redeem lifecycle, and where Predict Bay fits.
Predict Bay ·
DeepBook Predict is an on-chain protocol on the Sui network for binary options — contracts that pay a fixed amount if an asset settles on one side of a strike price at expiry, and nothing otherwise. It sits alongside DeepBook, Sui's native on-chain central limit order book. Positions are minted, held and redeemed entirely on-chain, which means no off-chain matching engine holds the book, and the outcome is decided by a published rule against an on-chain oracle price rather than by an operator's judgement.
ROUND CLOSES IN
--:--
LAST PRICE
—
STRIKE
—
NO POSITIONS YET THIS ROUND — BE THE FIRST
Practice mode · no wallet, no deposit, free to start
That sentence packs in three ideas worth unpacking separately: what a binary option actually is, what changes when the order book lives on a public chain instead of a company's servers, and how a position travels from mint to payout. This page takes them in order, then explains where Predict Bay sits — we integrate the protocol's Sui testnet deployment, and we are precise below about what that does and does not let you do today.
An option is a contract whose value depends on where a price ends up by a certain time. A binary option strips that down to a single question asked once, at expiry: did the underlying settle above the strike, or below it? One answer pays a fixed amount. The other pays nothing. There is no third case and no partial credit.
The consequence is that you are exposed to direction, not magnitude. If you hold a contract that pays out above $70,000 and BTC expires at $70,010, you get the full payout — exactly the same payout you would get at $95,000. Being spectacularly right is worth the same as being barely right. In exchange, the most you can lose is what you paid for the contract. There is no margin, no liquidation price, no funding rate, and nothing that can close your position early because the path got rough before you were proved correct.
| Instrument | Payoff shape | What you have to get right | Worst case |
|---|---|---|---|
| Spot | Linear with the price | Direction and magnitude | The asset falls to zero |
| Perpetual future | Linear, amplified by leverage, adjusted by funding | Direction, magnitude and the path in between | Liquidated before your view plays out |
| Vanilla option (long) | Scales with how far past the strike you finish | Direction, magnitude and timing | The premium you paid |
| Binary option (long) | Fixed payout on one side of the strike, zero on the other | Direction at one specific moment | The premium you paid |
That last row is why binaries read like odds rather than like leverage. A contract paying $1 if BTC is above $70,000 at 4pm is worth something between $0 and $1 right now, and that number is effectively the market's implied probability of the event. Buy at 35 cents and a winning contract returns the full dollar — a 65-cent profit on a 35-cent stake, roughly 2.9x your money back. (The numbers here are illustrative arithmetic, not quoted prices.) The price *is* the probability, which is the property that makes binary options and prediction markets look so similar from the outside. They are not identical, and the differences are covered in binary options vs prediction markets.
In DeepBook Predict a position is defined by two things: an expiry and a strike expressed as a price tick. A tick is the smallest price increment the market recognises, so a strike is stored as a whole number of ticks rather than as a floating-point price — an on-chain system needs prices that are exact integers, not values that drift by a fraction of a cent depending on who does the arithmetic.
Sentinel tick values express one-sided contracts. A position bounded on the low side by a strike and open-ended above it is an "above X" contract; the mirror image is "below X". A bounded pair of ticks — a floor and a ceiling — expresses a range contract that pays if the asset finishes between them. Same machinery, three different shapes of bet.
A central limit order book, or CLOB, is the familiar market structure: resting bids and offers, matched by price-time priority. Almost every venue you have used runs one in memory on its own servers. The book you see through the API is a rendering of a private data structure, the match happens inside the operator's process, and your balance is a row in the operator's database until you withdraw.
An on-chain order book puts that structure into the chain's own state. Orders are objects the network stores, matching is executed by the protocol's published code, and the resulting positions and balances are readable by anyone without asking the operator for an API key. For a spot swap this is a nice property. For options it is closer to a requirement, for two reasons.
First, options have a lifetime. A spot trade is over the instant it fills. A binary option has to survive from mint until expiry — sometimes an hour, sometimes days — and for that whole period something has to hold the record of who owns it and the collateral that will fund the payout. On-chain, that record is chain state and the collateral sits in contracts whose rules anyone can read before committing a dollar. Off-chain, both are entries in a company's ledger, and your only guarantee is that the company keeps its word.
Second, binary settlement is discontinuous. With a linear instrument, a one-cent error in the settlement price costs you a cent. With a binary, one cent either side of the strike is the difference between the entire payout and nothing at all. That makes the settlement price the single most contentious number in the product, and the argument for deciding it with a published on-chain rule rather than an operator's judgement gets very strong. The mechanics of that — oracle reads, settlement windows, why the price source is chosen in advance — are covered in oracle resolution and on-chain settlement. The liquidity side, and why an order book behaves differently from an automated market maker for this kind of contract, is in DeepBook's CLOB and prediction market liquidity.
DeepBook is Sui's native on-chain order book — shared trading infrastructure deployed as Move packages on the chain itself rather than inside one application, so any app on Sui can route through the same liquidity instead of bootstrapping its own. DeepBook Predict is the options protocol built on that stack. The relationship is roughly the one between an exchange's spot market and its derivatives market: same venue, same plumbing, different instrument.
What carries over is the discipline. Prices live on a tick grid. Sizes live on a lot grid. Positions and balances are Move objects, and ownership of a Move object is chain state anyone can verify — which is what lets a position be transferred, read by another contract, or redeemed by whoever holds it at expiry without a custodian in the middle. What changes is the thing being traded: on DeepBook you swap one asset for another, and on DeepBook Predict you mint a contract that resolves to a fixed payout depending on where a price lands at a set time.
Every DeepBook Predict position moves through the same four stages, and all four happen on-chain. Nothing in this sequence requires you to trust an intermediary to pass a message along.
Mint
You choose an expiry and a strike, then pay a premium to mint the position. The premium is derived from pricing inputs published on-chain rather than from a number an interface makes up privately — every input is a value on Sui that anyone can read. What you receive is an on-chain object in your own account.
Hold
The position sits in your account until expiry. There is no margin to top up, nothing to roll, and no funding to pay. Its value moves as the underlying moves relative to the strike and as time runs down — a contract five minutes from expiry with the price sitting far above the strike is worth close to its full payout; the same contract far below it is worth close to nothing.
Settle
At expiry the protocol reads its on-chain oracle price and compares it to the strike. That comparison determines the outcome. It is not a call anyone makes; it is a rule the contract executes with an input that was published on-chain.
Redeem
If the position finished in the money, the holder redeems it for the payout. On the testnet deployment the settlement asset is dUSDC. If it finished out of the money it expires worthless, and the premium you paid is the whole of your loss.
One practical detail newcomers hit early: on-chain markets are built on grids, not on arbitrary numbers. Prices snap to a tick and quantities snap to a lot, and venues of this kind generally enforce a minimum order size on top of that, so a dust position worth a fraction of a cent is not something you can express. The grids are what let the chain compare and match orders as exact integers, and the minimum keeps the book from filling with positions too small to be worth the state they occupy. In practice it means a very small test trade has to meet the market's minimum before it will go through — check the size the interface shows you rather than assuming any amount is valid.
A binary's fair price is the probability of the event, discounted a little for time. Working out that probability needs more than a spot price: it depends on how far the strike is from where the asset trades now, how long is left, and how much the asset tends to move over that horizon. That third input is volatility, and it is the reason an options protocol needs richer data than a swap venue does.
The deployment Predict Bay integrates against reads feeds published on Sui — spot, forward and volatility-surface — alongside Pyth price feeds for the underlying asset. Pyth is an oracle network that aggregates prices contributed by trading firms, exchanges and other first-party publishers and makes them available on-chain for contracts to read. The point is that every input to pricing and settlement is a value on a public chain you can inspect, rather than an internal number produced by whoever is running the interface.
The short-duration markets on Predict Bay work on the same principle. A market's baseline strike is fixed from the oracle price at the round's open and is identical for every participant — nobody gets a friendlier strike for arriving later — and the number the round resolves against comes from the same Pyth feed that set it.
Predict Bay is an interface, not the venue. Every market page carries a mode switch; with EVM trading paused it renders two options — Practice for free and Sui for DeepBook Predict. The Sui mode is designed to route the trade to the protocol rather than to an internal order book: the app builds the mint transaction, submits it, and later reads your positions and balances back from the chain. That is the path the testnet deployment exercises, and it is switched off for real funds today. The distinction still matters, because it tells you who you would actually be trusting with an outcome — in that mode, the answer is the protocol and its oracle.
The markets themselves are short-duration price markets on BTC, ETH, SOL, SUI, DOGE and XRP, in hourly and daily rounds. A strike is set at the open, you pick whether the asset finishes above or below it, and the market resolves against an on-chain price feed. Practice mode mirrors those markets exactly — same strikes, same rounds, same resolution source, no money at risk — which makes it a genuine rehearsal rather than a toy.
Two things about access are worth calling out, because they remove the usual first hurdle. You can sign in with Google through zkLogin, so there is no seed phrase to write down and lose; the cryptography that derives a Sui address from an ordinary login is explained in prediction markets without a seed phrase. And transactions can be sponsored, meaning you are not required to hold SUI for gas before you can do anything. Connecting a Sui wallet directly is supported too, for anyone who would rather bring their own keys.
On the token: $PBAY is a concept at the design stage. If and when it launches, the intended role is to align the people who use the platform with the platform itself. Anything more specific than that — supply, distribution, timing, venue — does not exist yet in any form we would put in writing, and the token page is where any of it would appear first.
If this is your first prediction market, the useful order is: trade a few practice rounds until the strike-and-expiry mechanic is second nature, then read how settlement is decided, then form a view on whether on-chain custody is worth anything to you. That last question is the one this whole protocol exists to answer, and it is easier to answer after you have watched a few contracts resolve.
Prediction markets carry risk and you can lose what you commit to a position. Nothing here is financial advice.