Source-linked AI summary
Redefining Stablecoins from Nominal to Real Value: A Maximum Likelihood Approach
Tomonori Kanno, Kensuke Ito, Yushi Yoshimura, Kyohei Shibano
TL;DR
Conventional stablecoins inherit fluctuations in fiat or other nominal units of account, motivating a real-value alternative for stablecoin stability. The paper introduces MLV, inferred from global asset-price data, and evaluates its real-time computation and portfolio-optimization use. It reports real-time computation from 500 asset price series and improved portfolio returns, Sharpe ratios, and turnover under MLV-based optimization.
Problem
Conventional stablecoins peg to units such as fiat currencies and gold that themselves fluctuate, while purchasing-power measures depend on locality- and time-varying market baskets.
Method
The paper defines MLV by estimating latent real-value movements from observed asset prices through maximum likelihood and applies it to collateral management and portfolio optimization.
Results
500 asset price series support real-time MLV computation, while MLV-based optimization raises annualized return and Sharpe ratio and substantially lowers turnover versus USD-based optimization.
Takeaways & Limitations
The results support redefining stablecoins as assets pegged to real rather than nominal values.
Takeaways & Limitations
MLV derivation depends on assumptions about stochastic distributions and covariance estimation, while practical deployment faces asynchronous trading, time-zone, data-acquisition, rebalancing, and return-allocation challenges.
Abstract
from arXiv · showhide
Stablecoins, typically pegged to fiat currencies, cannot achieve true stability because they inherit fluctuations in the underlying unit of account. To overcome this limitation, we introduce a stablecoin pegged to the Maximum Likelihood Value (MLV), a newly defined unit of account derived as the most probable configuration of latent real-value movements that explains observed nominal-value (price) changes. Grounded in inferential statistics and modern portfolio theory, MLV represents the most stable unit of account, as it enforces a zero real return on the minimum-variance portfolio. Empirical results confirm the operational viability of an MLV-pegged stablecoin: MLV can be computed in real time from 500 asset price series and improves annualized returns and Sharpe ratios while substantially reducing turnover in portfolio optimization.
Disclosure
The paper discloses patent applications concerning the MLV and portfolio optimization, alongside AI assistance in preparing a figure. Figure 1 uses a log-scale vertical axis to show volatility in USD/JPY and USD/XAU series normalized to 1.0 in 1980.
- The concept of MLV and its portfolio-optimization application are covered by patent applications filed by VLUE, Inc. in Japan and internationally.
- The figure was generated with AI assistance and then refined and finalized by the authors.
- Figure 1 presents USD/JPY and USD/XAU time series normalized to 1.0 in 1980, with observed volatility illustrating instability in conventional units of account.The figure’s vertical axis is shown on a log scale.
1 Introduction
The paper argues that fiat- and gold-pegged stablecoins inherit instability from nominal units of account and proposes the Maximum Likelihood Value (MLV) as a real-value alternative. It describes an MLV-based collateral and optimization scheme, supported by theoretical, operational, and portfolio results.
- Conventional stablecoin pegs inherit volatility from fiat currencies and gold, whose values fluctuate relative to other assets.Figure 1 illustrates this instability using USD/JPY and USD/XAU time series.
- Purchasing-power-based units are unsuitable for cross-border transfer because their market baskets vary by jurisdiction and over time.The paper therefore seeks a unit free from locality and arbitrariness.
- MLV infers the most probable latent real-value configuration explaining observed price changes using global financial data rather than a market basket.For an observed 1.5× USD/JPY increase, the paper gives scenario 3,626 as one selected configuration: 1.0× for USD and 0.67× for JPY.
- MLV-pegged stablecoins exchange deposited assets for equivalent MLV-denominated value and periodically rebalance collateral portfolios using MLV and modern portfolio theory.The resulting returns are distributed among the portfolio, users, and operator.
- The study contributes a real-value stablecoin definition, the MLV unit of account, and evidence that MLV-based portfolio optimization is effective.Users returning stablecoins receive assets equivalent in MLV terms, which may differ in composition from their original deposits.
- The paper presents the work as the first study to estimate unobservable real values and demonstrate the feasibility of a truly stable stablecoin.
2 Related Work
Prior stablecoin units of account use nominal currency baskets or inflation indices, leaving exposure to inflation, deflation, locality, or publication delays. The paper positions MLV as a real-time, high-stability alternative that also supports real-value portfolio optimization.
- Units of Account: Prior unit-of-account designs follow basket or indexed approaches, grouping existing currencies or linking value to inflation indices.
- Units of Account: Currency baskets cannot guarantee real-value stability because they remain exposed to inflation and deflation in their constituent currencies.They reflect real value only when aggregate constituent-currency value remains constant.
- Units of Account: Indexed units approximate purchasing power but are neither computable in real time nor free from locality and arbitrariness.Their underlying inflation indices are published ex post and depend on region-specific baskets.
- Units of Account: The paper presents MLV as combining real-time computability and high stability while avoiding nominal-basket and inflation-index limitations.
- Modern Portfolio Theory: Modern portfolio theory optimizes portfolios through mean-variance tradeoffs, with the Sharpe ratio measuring risk-adjusted performance.
- Modern Portfolio Theory: MLV-based optimization uses the MLV as benchmark to represent portfolio returns and cross-asset correlations in real-value terms.
3 Model
The MLV estimates latent real-value movements from observed nominal price changes using maximum likelihood, then constructs a volatility-free unit of account. Under modern portfolio theory, it is defined so the minimum-variance portfolio has zero real return.
- 3.1 Definition of the MLV: The MLV resolves the one-degree-of-freedom indeterminacy in real asset values by selecting the real-value movements that maximize their joint probability density.Observable prices provide only N−1 independent values for N assets, so the MLV uses a maximum likelihood framework to estimate the missing dimension.
- 3.1 Definition of the MLV: The MLV estimates each asset’s real log return and uses the resulting sequence to construct a volatility-free unit of account.The reference asset only fixes scale; changing it does not affect relative MLV-denominated values.
- 3.2 Derivation of the MLV: Under an elliptical-distribution assumption, the MLV is computed from observed price log-returns and the covariance structure of real-value movements.The derivation uses the fact that the elliptical density is maximized when the relevant quadratic form is minimized; the resulting estimate can be computed from observable price data.
- 3.3 Stability of the MLV: The minimum-variance portfolio minimizes risk among fully invested portfolios, with its weights determined by the covariance structure of asset returns.The paper notes that the covariance representation may use Tyler’s scatter matrix in the derivation and the correlation matrix in experiments.
- 3.3 Stability of the MLV: The MLV-denominated return of the minimum-variance portfolio equals zero for every time t, making the MLV the most stable unit of account under modern portfolio theory.The paper states that maximum-likelihood estimation of real-value fluctuations is equivalent to defining them so this portfolio’s real return is always zero.
- 3.4 MLV-based portfolio optimization: MLV-based portfolio optimization uses the covariance structure of real returns and incorporates expected dividend returns, rebalancing costs, and a risk-tolerance threshold.The paper presents this optimization as the practical portfolio-rebalancing procedure and evaluates its performance empirically.
4 Experiments
The experiments assess whether MLV can be computed in real time and whether MLV-based portfolio optimization improves performance and rebalancing stability. Results indicate that roughly 500 observed assets support real-time computation, while MLV-based risk optimization improves returns, Sharpe ratios, and turnover under CVaR constraints.
- Experiment design: MLV-pegged stablecoins are evaluated for real-time computability and sufficient portfolio returns through two experiments.The first experiment examines estimation accuracy as the number of observed assets changes; the second compares portfolio optimization across USD- and MLV-based risk and evaluation spaces.
- Real-time computability: The computability experiment estimates MLV correlation matrices from randomly selected assets over rolling windows and evaluates average RMSE against ground-truth matrices across 10 runs.Assets are selected from a universe of N=500, with n varied from 50 to 500.
- Real-time computability: Estimation error rapidly decreases with more observed assets and stabilizes beyond approximately n=200.Accuracy gains diminish with additional observations.
- Real-time computability: n≈500 observed assets is sufficient for real-time MLV computation on standard hardware despite rolling-update costs scaling as O(n^3).The 500-asset level is presented as a practical observation level after estimation accuracy stabilizes.
- Portfolio optimization: 15.6% annualized return, 0.699 Sharpe ratio, and 0.133 turnover result when MLV is used as the risk space in USD evaluation, compared with 11.7%, 0.554, and 0.529.The comparison uses dividend-maximizing portfolios under a weekly CVaR constraint and evaluates returns, volatility, Sharpe ratio, drawdown, and turnover by unit of account.
5 Conclusion
The paper proposes MLV-pegged stablecoins as a real-value alternative to nominal pegs and reports theoretical and experimental support for the approach. It also identifies estimation, implementation, and decentralization challenges for future work.
- Conclusion: The study proposes MLV, a unit of account that infers unobservable real-value movements from financial asset prices.The paper theoretically establishes MLV as the most stable unit of account from inferential-statistical and portfolio-theory perspectives.
- Conclusion: Experiments demonstrate that MLV can be computed in real time and can improve portfolio optimization performance.These findings support the paper’s proposed stablecoin design.
- Future work: Further MLV derivation work must address the stochastic-distribution assumption and improve real-time estimation of Σ_t.Such refinements may improve representation of tail dependence and asymmetric dependence.
- Future work: Practical implementation remains constrained by region-specific trading hours, time zones, and allocation of realized returns among stakeholders.The paper points to asynchronous-trading research and fee and compensation structures as relevant areas.
- Future work: Decentralized issuance could eliminate single points of failure, but sustainable decentralization requires incentive mechanisms grounded in cryptoeconomics and tokenomics.The discussion cites DAI and lessons from failed algorithmic stablecoins as relevant precedents.
rdiv
Expected dividend returns are estimated from historical observations within a fixed lookback window, while covariance rescaling does not affect the resulting optimization because positive scalar factors cancel.
- Dividend-return estimation: At each rebalancing time, the expected dividend return vector is estimated using historical data from a fixed lookback window.The set L_t contains past trading times within that window.
- Dividend-return estimation: The vector bμ_i,t provides an empirical estimate of the unobserved true expected dividend return vector μ_i,t.This estimate is used at each rebalancing time.
- Covariance treatment: Any positive scalar rescaling of Σ_t cancels out in equation (5), leaving the optimization invariant to covariance scale.The passage states the invariance without changing the covariance structure’s role.