Assay
$ASSAYSoon

Reference

Risks

What Assay does not protect against, and the limits of its design.

For buyers

  • The trigger is strict. A depeg must hold at the conservative edge of the confidence box, at or below −X, for T of regular-session time, with no gap longer than G. A short or noisy depeg pays nothing.
  • Off-hours do not count. A wrapper that falls on a weekend and recovers by the open never triggers.
  • Payouts are capped. A pool pays at most its cap per unit of notional, 20% in production pools, however deep the depeg.
  • The payout is fixed at the trigger. It uses the average over the first T of the streak. A depeg that deepens afterwards does not raise it.
  • A halt is not a trigger. If the stock stops trading, its price goes stale and cannot extend a streak.

For underwriters

  • You can lose your deposit. A trigger at the cap on a pool sold to capacity uses every deposit.
  • Adverse selection. A buyer who sees a depeg off chain can buy cover in the seconds before the first failing observation lands. Sales pause once a streak is open, and observations are posted often, but the window is not closed entirely.
  • Utilization is not weighted by time. Premiums are prorated by the time left in the week, not by how long the cover was held.

For everyone

  • The ratio is published by the issuer. Both multipliers are set by the issuers. An issuer that inflated its ratio would make its own token look depegged. One that deflated it could hide a depeg. The app shows the ratio next to every reading.
  • Snapshot selection. Anyone may post observations, so a poster can choose the worst signed snapshot in each window. The time weighting stops the number of posts from mattering, not the choice of snapshot. The effect is bounded by the depeg the market actually printed.
  • Oracle dependency. Assay trusts Pyth’s signed prices and session flags. The program verifies the signature, not the prices behind it.
  • Smart contract risk. The program and its math are tested, and the on-chain math is checked against a second implementation on 1,500 test cases, but no code is free of bugs.
  • No claim deadline. Unclaimed payouts stay reserved in the vault.