Loading...

Free

StableCoin Depegging Risk Prediction Studied with Machine Learning

StableCoin Depegging Risk Prediction Studied with Machine Learning

Stablecoins
Stablecoins are designed in such a way that they are "pegged" to another asset in terms of its relative price. In the cryptocurrency space, the top 4 stablecoins that trade are the USDT, USDC, BUSD and DAI. USDT, USDC, BUSD, and DAI are all "pegged" to the US Dollar, essentially its equivalent exchange in value. For example and in theory, one USDC could be exchanged for one USD. However, in practice, the pegged currency is subjected to supply and demand in the marketplace that can create dislocations from the "peg" and also a depegging event, in which price becomes complete disassociated with the peg. One classic example is Terra, (UST), which infamously was "algorithmically pegged" by a system, but lost its peg and collapsed.

Stablecoin Depegging Risk Prediction
Researches did a study in which they used machine learning to assess and predict depegging events. The study done by Lee, Chiu, and Hsieh was conducted over the Jan 2022 to Dec 2023 time period, and so therefore included some market moving events like the collapse of Silicon Valley Bank (USDC lost its peg over this time horizon briefly). With some mixed findings and results in the end, the study did continue to establish the importance of price, volume, and market capitalization changes for a stablecoin and that random forests and XGB boost algorithms outperformed logistic regression models. Other predictive features include total supply percentage changes, realized daily volatility, and price deviations over a specified time window (ie. 5 days or 10 days etc). Sentiment indicators did not appear to add value to the prediction process. Since XGB book and random forest models performed the best, it implies some non-linear relationships between the variables included. Volatility in major cryptocurrency assets, like Bitcoin and Ethereum, can also add depegging uncertainty to the stablecoin infrastructure.

Authors
You can read the full academic article and dive deeper into the methods and feature sets used here.

Lee, Earl and Chiu, Yu-Fen and Hsieh, Ming-Hua, Stablecoin Depegging Risk Prediction.
Available at SSRN: https://ssrn.com/abstract=4700764