Testnet launches Jan 20, 2027. Mock money only.
ETHfirst market
BTCnext
SOLcrypto
GOLDcommodity
CLcrude oil
EURforex
JPYforex
AAPLequity
NVDAequity
TSLAequity

Concepts

Stochastic Volatility (SVI)

SVI, Stochastic Volatility Inspired, is the volatility smile Optionbolt runs on-chain. It gives Black-Scholes a different implied volatility for each strike, so the option bond is priced off a realistic curve instead of one flat number.

The volatility smile

A single implied volatility misprices real options. Strikes away from the money trade richer than one flat number predicts: the further a strike sits from the underlying, the higher its implied volatility, a shape called the smile (or, when it is lopsided, the skew). Pricing the option bond off a flat volatility would under-collateralize the wings, exactly where a written option is most dangerous.

SVI captures that whole curve with five parameters, cheaply enough to evaluate on-chain.

The model

Optionbolt uses raw SVI (Gatheral, 2004). It expresses total implied variance as a function of log-moneyness, the log of strike over forward:

formula
k    = ln(strike / forward)                       log-moneyness
w(k) = a + b·( ρ·(k - m) + √((k - m)² + σ²) )     total variance
IV   = √( w(k) / T )

The forward is the underlying’s current mark. Five parameters shape the curve, set per asset and expiry:

ParamMeaningConstraint
aOverall variance level (vertical shift).keeps variance ≥ 0 everywhere
bWing slope: how fast the smile rises.b ≥ 0
ρSkew; negative tilts it toward richer puts.-1 < ρ < 1
mHorizontal shift that centers the smile.
σCurvature through the middle.σ > 0

The no-arbitrage floor is a + b·σ·√(1 - ρ²) ≥ 0, which keeps total variance non-negative at every strike. The core checks it, and the per-parameter bounds, when a ticker is created, so a malformed smile can never be listed.

raw a, b, ρ, m, σ
0%26%52%78%103%$2k$3k$4k$5k$6kKimplied volmark
IV at mark 40.1%ATM 40.1%skew +9.5pparb-safe
Two ways to drive the same smile: the raw SVI parameters, or the market stats you would read off an option chain, which solve back to those parameters. Fix a strike and tenor, then drag the mark to watch a fixed-strike option trace the smile as the underlying moves. Push the level negative and the arb-safe check trips, the condition the on-chain contract rejects.

How Optionbolt uses it

Governance sets the five parameters per asset-expiry when it lists a ticker. At mint the core computes the strike’s implied volatility with SVI and feeds it to Black-Scholes to size the option bond, and each position snapshots the volatility it was priced at, so later changes never touch an open position. The same smile re-prices an option at resolution to open the recollateralization auction.

The smile is written in log-moneyness, ln(strike / forward), so the core reads the live forward (the perp mark) at each mint and re-evaluates the same fixed parameters. Implied vol tracks the market as spot moves, without the smile being reset.

The parameters are not set once and abandoned. A limited set of admin roles can adjust the live smile, set a failsafe volatility, or halt new mints on a ticker, with any change applying to new mints only. The aim is a system that needs minimal DAO intervention, not one an admin steers day to day.

Calibrating the smile

The market-inputs mode above is how the smile actually gets set: not by guessing the five raw numbers, but by reading the stats a trader already knows, the at-the-money level, the skew, how fast the wings climb, and how round the trough is, then solving back to the parameters. Those stats come from a real options market.

For an asset with a deep, liquid listed-options market (BTC and ETH on Deribit, index and commodity options on the major venues), the five parameters are fit to that observed implied-volatility surface. A liquid chain is the useful benchmark: its quotes are real, tight, and tradeable, so they anchor the smile to prices the market will actually stand behind rather than a number someone picked. The fit runs off-chain (an optimizer); governance or a keeper then pushes the result on-chain, where the core only evaluates the stored parameters and never calibrates.

Because this is a collateral floor and not a tradeable quote, the parameters are then set deliberately conservative, the level biased high and the wings generous, so the bond over-collateralizes relative to that fair surface. The live chain shows where fair sits; the floor sits above it.

Why SVI, not SABR

SVI is the on-chain-tractable smile. Evaluating it is one logarithm plus a couple of square roots, all of which the fixed-point math library already has. SABR’s Hagan formula is a stack of transcendental functions that belongs off-chain in a keeper. SVI is what can run inside the core and price collateral in the same transaction as the mint.

On-chain and validated

The parameters and the math are 18-decimal fixed point. The Solidity library is diff-tested against a JavaScript reference and cross-checked against two independent open-source SVI implementations, matching to within fixed-point rounding.

Note

SVI supplies the volatility; Black-Scholes turns it into a price. Together they size the option bond on-chain from a realistic smile, not a single guessed number.

Related