# A Theoretical Approach to Stablecoin Design via Price Windows

Crypto.world

> Backed stablecoins try to hold a $1 peg by letting anyone mint a coin from the mechanism or redeem it back at prices near $1, which confines the market price to an interval between those mint and…

**arXiv**: 2602.15981

**Published**: 2026-02-17

**Authors**: Katherine Molinet, Aris Filos-Ratsikas

## AI summary

## Overview

**Research area:** Algorithmic game theory and mechanism design (cs.GT), applied to the economics of backed stablecoins.

**Technical level:** Advanced. The paper is written in formal mathematical style, with a model, definitions, a theorem with proof sketch, and stochastic analysis of a speculator's optimal strategy. The prose is unusually readable for theory work, but following the results requires comfort with limits, lim sup/lim inf, and expected-value arguments.

**Scope (one sentence):** The paper analyzes whether stablecoins that mint and redeem at fixed dollar prices for all agents ("price window" designs) can be both short-term and long-term stable when backed by a volatile asset, without relying on secondary stabilization mechanisms.

## What This Paper Is About

Backed stablecoins try to hold a $1 peg by letting anyone mint a coin from the mechanism or redeem it back at prices near $1, which confines the market price to an interval between those mint and redeem prices. The paper asks whether that confinement actually produces lasting stability when the reserves behind the coin are themselves volatile (for example, Ether or BTC). The answer is largely no: the mechanism either bleeds reserves to a patient arbitrageur, or widens its price band so much that the stablecoin simply inherits the volatility of whatever backs it.

## Key Contributions

1. **A taxonomy based on information use, not on collateral type.** The authors split backed stablecoins by whether the mint and redeem functions can depend on the specific transaction (agent identity, coin history) or only on the system state. Designs that treat all agents and all coins interchangeably are called "price window" designs; this class covers USDT and USDC, and also crypto-backed designs such as USDN and Cardano's Djed. DAI is the contrasting non-interchangeable case, because personalized redemption prices let it block an agent from redeeming for more backing coin than they paid to mint.
2. **A formal notion of stability that encodes long-term sustainability.** The paper defines ε-stability over a price sequence and then "weak ε-stability," which requires the mechanism to hold the price within $[1-\varepsilon, 1+\varepsilon]$ over at least one non-convergent (that is, never-settling) sequence of backing-coin prices. Requiring non-convergence rules out designs that only look stable because their input prices eventually stop moving.
3. **A general impossibility-with-tradeoff theorem (Theorem 3.1).** For constant fees, the mechanism is ε-stable under attack by any "sensitive" speculator whenever a quantity $L := (1+\varepsilon_\alpha)\liminf \frac{1}{p_t} - (1-\varepsilon_\beta)\limsup \frac{1}{p_t}$ is positive, and only if $L \geq 0$. This yields a theoretical minimum fee size needed to prevent eventual reserve depletion, stated in terms of the asymptotic spread of the backing-coin price sequence.
4. **An explicit speculator model with a computable expected time to depletion.** The paper replaces the black-box speculator with a utility-maximizing agent characterized by patience δ, beliefs about the price distribution, and risk tolerance Λ, derives a closed-form condition for whether reserves can ever be drained, and then relaxes the independence assumption to simulate the strategy on historical Ether and Bitcoin price data.

## Main Findings

- **Zero-fee price windows are never weakly 0-stable.** With $\varepsilon_\alpha = \varepsilon_\beta = 0$, Theorem 3.1 implies depletion occurs if and only if $\liminf \frac{1}{p_t} = \limsup \frac{1}{p_t}$, which by the squeeze theorem means the price sequence converges. A mechanism that mints and redeems at exactly $1 is therefore only stable over price sequences that are already settling down, such as USDT and USDC where $\frac{1}{p_t} = 1$ for all $t$.
- **The required fee scales with the backing coin's volatility.** Setting $\varepsilon_\alpha = \varepsilon_\beta = \varepsilon$, avoiding depletion requires $\varepsilon \geq \frac{\limsup \frac{1}{p_t} - \liminf \frac{1}{p_t}}{\limsup \frac{1}{p_t} + \liminf \frac{1}{p_t}}$; the threshold and the tail spread of the price sequence are proportional, with the denominator acting only as a normalizing factor.
- **The price band inherits the backing asset's volatility.** Because the sustainable window must be about as wide as the volatility of the underlying price sequence, the market price of the stablecoin is permitted to fluctuate proportionally to the backing coin. The authors describe this as price stability being "largely an inherited property" for designs that rely on a price window as their primary stabilization mechanism.
- **Fees delay but do not fix the problem.** Transaction fees add reserves as a buffer and push back the depletion time, but used in isolation they cannot deliver short-term and long-term stability simultaneously over a volatile backing coin.
- **Only the tail of the price sequence matters.** Since a sensitive speculator's profit can be scaled arbitrarily small, the exhaustion argument hinges on whether total profit becomes infinite, which can only happen in the tail; hence the criteria depend solely on lim sup and lim inf.
- **In the IID setting, a sharp threshold separates survival from ruin.** The optimal strategy is a two-threshold rule: buy when the backing-coin price rises above $y_2$, sell when it falls below $y_1 \leq y_2$. A speculator with $n_0 > 0$ can never deplete reserves if and only if $Y = \frac{\frac{1}{1-F(y_2)}\int_{y_2}^{\infty} x f(x) dx}{\frac{1}{F(y_1)}\int_{-\infty}^{y_1} x f(x) dx} = 1$ (or is so risk averse that $\Delta_t = 0$ for all $t$); larger $Y$ means faster depletion.
- **For a normal distribution, volatility alone determines the outcome.** The ratio reduces to $Y = \frac{\mu + \sigma^2 \frac{f(y_2)}{1-F(y_2)}}{\mu - \sigma^2 \frac{f(y_1)}{F(y_1)}}$, so $Y = 1$ if and only if $\sigma^2 = 0$, and larger $\sigma^2$ gives larger $Y$ — matching the general theorem's link between tail spread and exploitable opportunity.
- **Numerical example: depletion in under ten days.** For prices drawn at random from a normal distribution with $\mu = \sigma^2 = 100$, a speculator with $\Lambda = 0$ and $\delta = 0.1$, and $R_0/n_0 = 100$, the computed value is $k(t) = \log(1 + R_0/n_0)/\log(a_1) \approx 15.78$, corresponding to roughly 227 timesteps; at one interaction per hour, that is less than 10 days to depletion.
- **Historical simulation: three years for Ether- and Bitcoin-backed systems.** Relaxing the independent-draws assumption, the authors simulate the speculator over historical Ether and Bitcoin price data and report that a sufficiently wealthy speculator would have been able to drain the reserves of a stablecoin system backed by each of these currencies within three years.
- **Contrast with prior work.** Unlike Potter et al. (2024), who assume all value is measured in dollars and find undercollateralization only in "bad" economic states, this paper shows reserve depletion can occur in good states through a speculator who values backing coins more than dollars — a reasonable assumption for high-return assets like Bitcoin or Ether.

## Methodology in Plain English

The authors build a deliberately stripped-down model. Time is a discrete infinite sequence of steps; at each step an exogenous, oracle-supplied backing-coin price $p_t$ is revealed. A mechanism mints coins at $(1+\varepsilon_\alpha)$ dollars' worth of backing coin and redeems at $(1-\varepsilon_\beta)$ dollars' worth, provided it holds reserves $R_t$ of at least $(1-\varepsilon_\beta)/p_t$; otherwise it hands back whatever remains. The gap between these two prices is the price window, and it creates a floor and a ceiling on the market price.

The adversary is defined abstractly rather than behaviorally. The authors compute $s_t$, the maximum profit an omniscient speculator could extract from a single unit of backing coin over the whole price sequence, and then call a real speculator "sensitive" if it captures at least a $1/k$ fraction of that optimal profit for some fixed $k$, on every price sequence. This avoids assuming anything about the speculator's internal strategy while still guaranteeing that profitable opportunities get exploited. The main theorem then compares the profit available in the tail of the price sequence against the fee-imposed spread; if the spread can be arbitraged forever, reserves eventually hit zero.

For the applied part, they replace the abstraction with a concrete agent who discounts the future at rate δ, knows the price distribution's density $f(x)$, and has risk tolerance Λ. Solving the per-period utility maximization yields expected portfolio growth, from which expected time to depletion follows; the cleanest closed form requires risk neutrality ($\Lambda = 0$). They then drop the independence assumption and replay the threshold strategy against real Ether and Bitcoin price histories.

## Why This Matters

**Impact on research.** The paper gives the stablecoin-design literature something it has largely lacked: a theorem that separates design families by an information property (interchangeability) rather than by collateral type, and that produces a quantitative, falsifiable prediction — the minimum fee is proportional to asymptotic backing-coin volatility. It also reframes stability as an inherited rather than engineered property for a whole class of designs, and offers an explanation for why crypto-backed projects tend to avoid price windows even though fiat-backed projects overwhelmingly use them.

**Real-world applications.**

- Sizing mint and redeem fees for crypto-backed stablecoins so that arbitrage cannot systematically drain reserves.
- Stress-testing a proposed reserve level against a speculator model with specified patience and risk tolerance, rather than against a purely qualitative risk narrative.
- Explaining observed price behavior in DAI-like systems, where the coin trades above or below peg depending on whether ETH is in a bad or good state, as noted in prior empirical work the paper cites.
- Distinguishing, at the design stage, which stabilization failures require a secondary mechanism (as Bullmann et al.'s primary/secondary distinction suggests) and which can be handled by pricing alone.

**Industry relevance.** Any issuer or auditor of a crypto-backed stablecoin, and any protocol considering a fixed-price mint/redeem module, can use the fee formula as a starting sanity check. The paper is candid that the theoretical fee may be impractically large for a highly volatile backing coin, but argues it may be reasonable for an asset that is close to stable — which is precisely the fiat-backed case that dominates the market.

## Future Directions

- **Extending the theorem to time-varying fees.** The authors state that Theorem 3.1 generalizes to $\varepsilon_\alpha$ and $\varepsilon_\beta$ that change over time, applying to their lim sups instead of constants, but the proof is not included and is left for future work.
- **Making the speculator model more realistic.** The closed-form depletion time requires risk neutrality ($\Lambda = 0$) and independent price draws; solving the general case with risk aversion and dependent prices is left open, and Section 4 is explicitly presented as a high-level summary with details deferred to the appendices.
- **Answering the field's two open questions.** The paper positions itself as a small piece of a larger program: whether a decentralized, provably stable coin can exist at all, and what the full space of possible stablecoin designs looks like. The authors hope their minimal-assumption, taxonomy-by-emergent-properties approach can be built on.
- **Exploring hybrid designs.** Since price windows alone cannot deliver both short- and long-term stability, a natural next question is which secondary stabilization mechanisms could be layered on top, and how the interchangeability property interacts with them.

## Target Audience

Theorists and researchers working on stablecoin design taxonomies and mechanism design will get the most from Sections 2 and 3, which formalize the interchangeability distinction and prove the central tradeoff. System designers and mechanism engineers will find the fee-sizing formula and the depletion-time calculations practically usable, particularly for near-stable backing assets. Economists studying arbitrage and peg deviations in crypto markets will find the speculator model and its volatility-to-depletion-time relationship relevant. Readers without a mathematical background should expect to work through formal definitions, though the interpretive discussion in Section 3.3 is largely accessible prose.

## Authors’ abstract

In this paper, we explore the short- and long-term stability of backed stablecoins offering constant mint and redeem prices to all agents. We refer to such designs as price window-based, since the mint and redeem prices constrain the stablecoin's market equilibrium. We show that, without secondary stabilization mechanisms, price window designs cannot achieve both short- and long-term stability unless they are backed by already-stable reserves. In particular, the mechanism…

[The full abstract on arXiv](https://arxiv.org/abs/2602.15981)

[Read the original paper](https://arxiv.org/abs/2602.15981)

---

Canonical: https://crypto.world/paper/a-theoretical-approach-to-stablecoin-design-via-price-windows-2602.15981
