Paper: arXiv 2306.10612

Authors: Thomas N. Cintra, Maxwell P. Holloway

Abstract

We consider a liquidity provider’s (LP’s) exposure to stablecoin and liquid staking derivative (LSD) depegs on Curve’s StableSwap pools. We construct a suite of metrics designed to detect potential asset depegs based on price and trading data. Using our metrics, we fine-tune a Bayesian Online Changepoint Detection (BOCD) algorithm to alert LPs of potential depegs before or as they occur. We train and test our changepoint detection algorithm against Curve LP token prices for 13 StableSwap pools throughout 2022 and 2023, focusing on relevant stablecoin and LSD depegs. We show that our model, trained on 2022 UST data, is able to detect the USDC depeg in March of 2023 at 9pm UTC on March 10th, approximately 5 hours before USDC dips below 99 cents, with few false alarms in the 17 months on which it is tested. Finally, we describe how this research may be used by Curve’s liquidity providers, and how it may be extended to dynamically de-risk Curve pools by modifying parameters in anticipation of potential depegs. This research underpins an API developed to alert Curve LPs, in real-time, when their positions might be at risk.

Complexity vs Empirical Score

  • Math Complexity: 4.0/10
  • Empirical Rigor: 7.5/10
  • Quadrant: Street Traders — empirical and applied, lower math complexity

Why this score: The paper applies established Bayesian Online Changepoint Detection (BOCD) algorithms rather than developing novel mathematics, focusing on practical implementation with backtested results on real Curve Finance data from 2022-2023, and mentions a deployed API.

Research Flowchart

  flowchart TD
  A["Research Goal: Detect<br>Stablecoin/LSD Depegs<br>for Curve LPs"] --> B["Methodology: Construct Metrics &<br>Bayesian Online Changepoint Detection BOCD"]
  B --> C{"Data Sources:<br>2022-2023 Curve StableSwap Pools<br>including UST & USDC depegs"}
  C --> D["Computational Process:<br>Train BOCD on 2022 UST Data"]
  D --> E["Validation:<br>Test on 2023 USDC depeg<br>and 17 other months"]
  E --> F["Key Findings:<br>Detected USDC depeg 5 hours early<br>with few false alarms"]
  F --> G["Outcome: Real-time<br>Alert API for LPs"]

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes a Bayesian Online Changepoint Detection (BOCD) model to detect stablecoin and LSD depegs for Curve liquidity providers.

Source sentence
We fine‑tune a Bayesian Online Changepoint Detection (BOCD) algorithm to alert LPs of potential depegs before or as they occur.

Assumes 2

  • All metrics can be transformed to be approximately Gaussian. statistical
    Source sentence
    We make a simplifying assumption in this paper that all our metrics can be transformed into being approximately Gaussian
  • Informed trades arrive as a Poisson process. model
    Source sentence
    The authors posit that informed trades arrive as a Poisson process with intensity 𝜖𝑖

Data

  • Curve LP token prices for 13 StableSwap pools · Curve market-prices unknown
    2022-2023 · 13 StableSwap pools
    Source sentence
    We train and test our changepoint detection algorithm against Curve LP token prices for 13 StableSwap pools throughout 2022 and 2023

Costs

Not applicable The paper reports no trading or execution result.

Finds 3

  • The BOCD model detected the USDC depeg in March 2023 about 5 hours before USDC fell below 99 cents, with few false alarms over 17 months. performance
    Source sentence
    We show that our model, trained on 2022 UST data, is able to detect the USDC depeg in March of 2023 at 9pm UTC on March 10th, approximately 5 hours before USDC dips below 99 cents, with few false alarms in the 17 months on which it is tested.
  • The BOCD model performs well on large‑volume pools (e.g., 3pool, ETH/stETH, FRAX/USDC) and yields relatively few false positives, detecting depegs several hours before price drops. performance
    Source sentence
    Our bocd model performs well on pools with large volume and large TVL, including the 3pool, ETH/stETH, and FRAX/USDC, whereas results are much noisier for smaller pools with less stable metrics.
  • Net swap flow, Shannon’s entropy, and 5‑minute markout metrics are the most predictive across all pools. predictive-accuracy
    Source sentence
    we find that the net swap flow, shannon’s entropy, and 5‑minute markout metrics are the most predictive metrics across all pools

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 9 related papers). Coverage is partial, so this is neither replication nor refutation.

Cited by 2 papers in the archive (2 by independent authors): Leverage Staking with Liquid Staking Derivatives (LSDs): … (uses its method or data)

Code

No repository by the authors found in the paper.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.