Opening

A maker posts an offer: a market, a term, a strike, a band, a size, and which side they want. The offer rests on the contract with the maker's escrow already inside it. A taker accepts it, posts the same escrow, and the pair becomes a position with a fixed expiry. Nobody borrows anything.

Escrow is size multiplied by band, in USDG. That deposit is the whole exposure. Nothing accrues against a position while it runs and there is no margin ratio to watch, because every dollar the contract could ever owe is already in it.

Accepting an offer also records that hour's pool price in the same transaction. If the pool is too thin to mark at that moment, the accept fails rather than opening a term against a stale price.

Marking the pool

Once an hour, anyone can ask the contract to record the price of a market's Uniswap v4 pool. The call is permissionless and cheap, and the contract writes at most one mark per hour, so calling it a hundred times in a minute does nothing but spend gas.

Two rules keep the series honest:

• A pool under the depth floor is not marked. The previous mark carries forward, and a carried mark contributes no return rather than an invented one.
• A missed hour carries forward the same way. The number of returns actually observed travels with the series, so annualisation divides by what was seen and not by the calendar.

Carrying forward is deliberately biased against the long side. An hour nobody bothered to record cannot create volatility for the person who wanted some.

Realized volatility

Take the return between consecutive marks, square it, average, annualise, take the root. Squaring is the step that removes direction: a two percent fall and a two percent rise are the same input.

rᵢ = ( Pᵢ − Pᵢ₋₁ ) / Pᵢ₋₁

RV = 100 · sqrt( 8760 / n · Σ rᵢ² )      // vol points

The mean return is not subtracted, which is the usual convention for variance products. It is why a stock that grinds steadily upward all week still settles at a low volatility: what counts is how jagged the path was, not where it ended. The contract uses simple returns rather than log returns; for hourly moves the two agree to well inside a vol point.

Vol points are annualised percentages. Divide by the square root of 365 to read one as an average day: 52 points is a typical day of about 2.7%.

Settlement

Once the expiry hour has fully passed, anyone can settle. The contract reads its own checkpoints at open and at expiry, and makes one transfer.

payoff = N · clamp( RV − K, −W, +W )

Both escrows are then released, adjusted by that transfer. The contract holds 2 · N · W and pays out 2 · N · W, so the books close to the cent whatever the market did. A term with fewer than 24 observed returns does not settle at zero volatility: it voids and returns both escrows untouched.

Why a band

Volatility has no ceiling. An uncapped long against an uncapped short is a promise the short cannot always keep, and enforcing that promise means margin, liquidations, and a keeper that has to win a race in exactly the week when everything is moving at once.

The band replaces all of that. Past the cap the payoff stops, and since both sides put the cap in escrow before the term begins, nobody has to be chased afterwards. The cost is the tail: a truly wild week pays the cap and no more.

Pushing the price

The obvious attack is to shove the pool just before a mark. It works badly here.

• Moving the price costs slippage, and the depth floor means the size needed to move a mark is not small.
• A push and its unwind are two large returns in the series, both squared, both raising RV. A manipulator hoping for a low settlement raises the very number they want to suppress.
• Marks are hourly and a term holds hundreds of them. One doctored hour is one term in a sum of hundreds, which is not enough to buy an outcome.

None of this makes a thin market safe. It makes moving one expensive and visible, and that is as far as the claim goes.

What can go wrong

• The pool is the price. If a market's liquidity leaves, its marks stop meaning anything. Tremo only lists names with real depth, and a listing is not a promise the depth stays.
• You need the other side. Every position is somebody else's opposite position. A strike nobody wants is a strike that never trades.
• The cap cuts both ways. The week you were most right about is the week you are most likely to be capped.
• Quiet is not safe. Short volatility loses in exactly the conditions where everyone wants out at once, and the loss lands all at the end rather than a little at a time.
• Unaudited, and new. The contract has a test suite, which is not an audit and is nowhere near a track record.