Markets connected to real-world data

Tokenized real-world assets can have an external reference. A tokenized equity follows a public market, a Treasury product may track NAV, and a tokenized commodity may follow an external spot market.

• Safety checksDetect significant divergence between an onchain market and an external reference.
• Fair-value displayGive traders and LPs context before they act.
• Fee inputsReflect changing volatility or uncertainty in risk controls.
• Execution guardsRestrict activity when reference data is stale, unavailable, or unreliable.

Those four uses have very different tolerances for a bad value, which is why they are not built on one undifferentiated notion of the price. Showing a stale reference to a trader is a minor problem that a timestamp solves. Feeding the same stale value into a control that restricts execution can halt a functioning market, and feeding it into a fee model can make a market unusable while it appears to be running normally.

Reference-relative liquidity

A range can be defined as a percentage around fair value instead of as fixed prices. An LP choosing a range from 2% below to 2% above the reference is describing a position relative to the asset rather than to a number, and when fair value moves the range moves with it.

The difference matters most for assets whose reference moves steadily in one direction. A fixed range drifts out of relevance as fair value travels away from it, while a reference-relative range holds its position against the asset.

Sources of reference data

Different asset classes have different notions of a correct price. The reference a market uses should match how the underlying actually settles.

The word price is doing different work in each of those rows. An equity price is a live consensus produced by a market that is trading right now. A NAV is a computed valuation published on a schedule, authoritative about the past and silent about the present. A private-credit NAV can be a month old, entirely correct on its own terms, and still a poor guide to what a buyer would pay today. Treating all three as interchangeable inputs is the most common design error in real-world asset market structure.

What makes reference data usable

A reference price is only as good as the guarantees around it. These are the properties a market needs to know about an input before it can build controls on top of it.

• FreshnessHow recently the value was published, and the age past which it should no longer be trusted.
• Update behaviorWhether the source publishes continuously, on a deviation threshold, or on a fixed schedule.
• ProvenanceWho produces the value, from which underlying market, and how it can be verified.
• Failure modeWhat the source returns when it cannot produce a value, and whether that state is distinguishable from a real price.

The last property matters most. A feed that repeats its previous value when it fails is more dangerous than one that returns nothing, because a market cannot tell the difference between a stable price and a stopped one.

Provenance deserves the same scrutiny as freshness, because a value can be perfectly current and still be the wrong value. A reference derived from a thin venue, or from an aggregate that includes one, imports that venue's problems into every market built on it. Knowing which underlying market a number came from is what lets anyone judge whether it should be trusted for the asset in question.

When data is unavailable

Reference data fails in ordinary ways: a feed is delayed, a market is closed, a NAV has not been published yet. The design question is not how to avoid those states but what a market does while it is in one.

• Stale dataWiden risk controls or restrict execution rather than trading against a price that is no longer true.
• Closed marketAccount for the gap risk of an underlying that can reopen at a different price.
• No referenceShow the absence explicitly instead of presenting the onchain price as validated.

What these responses share is a preference for doing less over guessing. A market that keeps trading against an unusable reference is not more available in any useful sense. It is available for exactly the trades that will cost its LPs the most. Restricting activity is least popular at the moment it is most valuable, which is why the behavior belongs in the protocol rather than in an operator's discretion.

Divergence is information, not an error

An onchain market and its external reference will not agree exactly, and they are not meant to. The gap between them carries meaning that both traders and LPs should be able to read.

Presenting divergence honestly means showing the number and its age rather than hiding it. A market that displays only the onchain price is easier to read and harder to evaluate.