Reviewed and updated: October 19, 2026.
An algorithm can define how token supply changes. It cannot guarantee that buyers will accept the asset delivered by that rule. To evaluate an algorithmic stablecoin, follow the redemption path and identify where purchasing power comes from when confidence weakens.
Algorithmic does not describe one universal design
Stablecoin systems can use supply adjustments, auctions, market operations, collateral or combinations of these. The presence of software-controlled stabilization does not mean that a token has no reserves. Conversely, a dollar target does not establish a right to receive a dollar from a bank account.
Ask three separate questions: what asset is the price meant to track, what can a holder actually receive on exit, and who is willing or required to provide it? A peg is a target or observed market relationship. Redemption is a mechanism or contractual right. Market liquidity is the ability to trade at a given size.
The historical UST and LUNA example
The BIS's 2022 Annual Economic Report describes TerraUSD's conversion mechanism with LUNA and the May 2022 collapse. UST's intended dollar equivalence depended on conversion into a volatile token and the market's willingness to absorb that token. It was not the same as redeeming a claim against a segregated dollar reserve.
This guide does not repeat the earlier unsourced $60 billion loss estimate. Market capitalization destroyed, realized investor losses and assets leaving a protocol are different measures. A useful historical account must specify its metric and dates.
A numerical model of reflexive pressure
Imagine a fictional stable token S that can be exchanged for one dollar's quoted value of volatile token V. At a V price of $10, redeeming 1,000 S creates 100 V. If V's quoted price falls to $2, the same redemption creates 500 V. At $0.50 it creates 2,000 V.
The arithmetic can remain correct while the economics deteriorate. Newly issued V must find buyers; selling may push its market price below the quote used by the mechanism. More redemptions can require more units, and more units offered for sale can further weaken the exit price. Conversion limits, fees and delayed pricing may change the path, but they do not conjure an outside buyer.
These invented amounts illustrate a feedback loop, not a simulation of Terra's exact contracts. A token's market capitalization is also not a cash reserve that can necessarily be liquidated at its last traded price.
Why historical product labels need dates
The earlier article called FRAX a current example of partial algorithmic backing without specifying a version. The FRAX v3 design document instead describes a target of at least 100% exogenous collateralization, use of market-operation contracts and no guaranteed holder redemption into a particular asset. It also says deployment is iterative. A design target is not verification of today's achieved collateral ratio or every feature's deployment.
Accordingly, this page uses that document to show why a version matters. It does not certify current FRAX reserves, classify every later Frax product or assume a token ticker always denotes the same design.
A stress-test worksheet
- Trace an exit when the stabilizing token falls 50% and trading depth shrinks.
- Identify the price source, its update delay and who can change it.
- Separate external collateral from assets whose demand depends on the same ecosystem.
- Check redemption caps, queues, suspension powers and transaction costs.
- Distinguish a temporary incentive payment from income that can fund ongoing obligations.
The strongest explanation of a peg includes the conditions under which it can fail. Fully automated rules can still leave users exposed to liquidity, governance and confidence shocks.
Sources and scope
- BIS, Annual Economic Report 2022: TerraUSD case
- Frax documentation: FRAX v3 overview and qualifications
Continue reading: Stablecoin architecture · Collateral and liquidation
Editorial update: Removed an unsupported loss total and outdated blanket FRAX classification; separated design targets from live reserve verification; added a reflexive-redemption calculation.