Binary pricing for prediction markets.
Price yes/no markets as the digital options they are, off a calibrated volatility surface, then settle them on a signed reference delivered on-chain.
Block Scholes2 min read
- Listings catalogued
- 1.8M+
- Settlement TWAP, or any window
- 30 min
- Signed on-chain
- EIP-712
- History since
- 2020
/01 The problem
A yes/no market is a binary option
A contract on whether BTC will be above $[strike] on Friday is a binary option, and it has a fair price well before it resolves. Priced from a spot feed and a rule of thumb, such a market is either too wide to trade or exposed to better-informed traders.
Pricing it requires a view of the probability of each outcome, which an implied volatility surface describes. Resolving it then requires an indisputable reference value.
/02 Binaries
A binary bet is a digital option.
"Will BTC be above $[strike] by Friday?" resolves exactly the way a digital option settles: it pays if the strike clears and nothing if it doesn’t. The calibrated surface prices that boundary at any strike and any expiry, so a market opens at a fair price.
/03 Settlement
Settlement on a signed on-chain reference.
Settle against a 30-minute TWAP of the composite, any TWAP window the contract specifies, or a single point-in-time value, signed and delivered on-chain through the oracle so the contract reads the value directly.
/04 History since 2020
Historical replay before listing
Point-in-time history back to 2020 shows how a binary would have priced and resolved through past shocks, such as sharp sell-offs, volatility spikes and sudden rallies, on the surfaces that existed at the time, with no lookahead.
This allows strikes, expiries, limits and fees to be set with confidence before launch, and a market’s behaviour in the next shock to be explained in advance.
/05 Market fit
From a fair mid to a tradeable market
The surface gives a fair mid price for every strike and expiry. A live market needs more than a mid: most pools apply their own fee surface on top, for example wider in the wings than near the money and shaped by tenor and liquidity, to set the prices traders see.
We work with teams on that layer as part of launching a product: which assets, strikes and expiries to list, how to shape the fee surface for the audience the market is aimed at, and how to stay competitive without leaving the pool exposed. The aim is a product that fits its market as well as a correct mid.
/06 The method
How the reference is made
Filter. Every quote is screened before calibration: one-sided, stale and arbitrageable prices are removed.
Repair. The remaining quotes are checked for static arbitrage across strikes and expiries. A violation small enough to repair is repaired; one too large to repair is discarded.
Weight. Each source is weighted by confidence, and the weights sum to one. The result is fitted across every listed expiry onto one continuous surface, so any strike or tenor has a calibrated value.
Questions
What teams ask before they build.
Why does a binary market need implied volatility?
A binary pays if a level clears by a date, which is how a digital option settles. The calibrated surface prices that boundary at any strike and expiry.Can we launch binary markets on Block Scholes data?
Yes. Prediction-market and event-contract teams price binaries off the calibrated surface and settle them on signed values through the oracle. Talk to us about the assets, expiries and chains you need, and how to shape fees on top of the mid.What can a market settle on?
A 30-minute TWAP of the composite, any TWAP window the contract specifies, or a single point-in-time value, signed and delivered on-chain and over REST and WebSocket.
Read next
The data and tools behind this.
- /01
Infrastructure · Oracle
Signed pricing and volatility data delivered through push or pull infrastructure.
- /02
Data & Analytics · Volatility
Implied and realised volatility, skew, term structure and calibrated surfaces across strike and maturity.
- /03
Data & Analytics · Historical Data
Point-in-time history for backtesting and model validation, to three years of depth.