Pricing

The line that matters most

This is repeated throughout these docs deliberately, because it's the thing most likely to get lost once you see how good the one-click experience feels: the ease comes from a suggestion engine, not from the protocol quietly becoming a counterparty.

Why that boundary exists

A protocol that prices its own options — quoting a volatility surface and automatically taking one side of every trade — is a fundamentally different, and riskier, system. If its pricing model is wrong, sophisticated traders will find that mispricing and extract it systematically, and the loss lands on the protocol's own capital, not on one individual's bounded position.

Plume sidesteps that entire category of risk. If the quote engine suggests a bad price, the consequence is contained to a single listing: a writer who accepts a bad suggestion earns less than they should have, or a listing priced too high simply doesn't sell. Either way, the loss is bounded by one person's collateral, and it never becomes a claim against a shared pool of anyone else's funds.

What the quote engine actually is

A small, deterministic piece of software, entirely separate from the contracts, that computes a suggested premium for a given asset, strike, and expiry. It runs three inputs:

• Realized volatility, measured from the asset's own price history as recorded by the same oracle Plume settles against — a continuously updating, first-party dataset that costs nothing extra to collect.
• Earnings awareness. Options tend to be worth more in the week of a company's earnings report. The engine keeps a calendar of known upcoming earnings dates and adjusts its volatility assumption upward in those weeks.
• A time-of-week adjustment. Tokenized stocks trade continuously, but the stock they represent mostly moves during market hours. The engine accounts for this with a documented, deliberately conservative adjustment — a genuinely unsolved problem industry-wide, and one Plume treats as an open area to keep refining.

The output is a single suggested number, shown wherever a price would otherwise need typing, versioned so every quote ever shown can be traced back to exactly which version of the model produced it.

How it learns

Every listing, every fill, and every expiry outcome is logged against what the engine predicted at the time. Comparing those logs over time — model suggestion versus what the market actually did — is how the engine's assumptions get refined. This dataset only exists because trading is happening; the model gets better precisely by watching its own suggestions succeed or fail in a live market.

What's deliberately not being built

An on-chain volatility surface. Protocol-owned market-making capital. Anything that would make the contracts' solvency depend on a pricing model being correct. These aren't oversights — they're excluded on principle, because the moment protocol solvency depends on a price being right, the entire simplicity argument from the collateral invariant stops being true. If Plume ever explores that direction, it would be a distinctly different, separately-reasoned system — and even then, the honest preference is a market-based mechanism like an auction, where the price comes from real bids rather than a model the protocol has to defend.