ELVARA: an intent-based protocol for risk-aware portfolio optimization of a real-world asset proxy basket

Shahin Ramezany, Rachsuda Setthawong, Pisal Setthawong

Abstract

The purpose of this paper is to design and evaluate a reproducible single-chain proof-of-concept protocol for optimizer-driven portfolio intents over a real-world asset (RWA) proxy basket. The system combines an off-chain mean-risk optimization engine with a minimal on-chain intent registry, thereby linking portfolio analytics to a verifiable machine-readable format. In particular, the off-chain module converts optimized weights into a deterministic JSON paper-reproduction artifact and an API response while the on-chain contract validates authorization, vector consistency, epoch monotonicity and expiry freshness before publishing the artifact. The novel contribution of this paper is a reproducible interface between risk-aware portfolio optimization for an RWA-oriented proxy basket and verifiable on-chain publication. The experiment consists of seven exchange-traded fund (ETF) proxies for RWA-related sectors: gold, silver, oil, real estate, U.S. Treasuries, energy infrastructure, and agriculture. Using a fixed walk-forward design from 2021–03–01 to 2026–03–03, with 504 training days, 63-day test windows, 63-day rebalance steps, and a fixed 10 bps transaction cost, several risk-aware allocation rules were compared, including conditional value at risk (CVaR), semivariance, variance, entropic value at risk (EVaR), worst realization, maximum drawdown, and an equal-weight baseline. The result in the final fixed-window baseline experiment showed that CVaR produced the highest point-estimate total return under the chosen specification, yielding a final portfolio value of $1,637,035 from a $1,000,000 start, corresponding to a 63.7% total return, versus 55.4% for equal weight, while also achieving a slightly lower maximum drawdown (−9.7% versus −10.3%) and a higher Sharpe ratio (1.18 versus 1.08). However, the sensitivity analysis also showed that strategy rankings changed when different walk-forward schedules and weight bounds were used. This means the baseline results depend on the chosen settings. The main contributions are a deterministic, machine-readable interface that connects risk-aware portfolio optimization to verifiable on-chain smart contract publication, a validated intent-registry architecture, and a fully reproducible empirical comparison of risk-aware allocation rules against an equal-weight benchmark.

Source: semanticscholar

Read the AI summary and key takeaways for traders on WOBR Quant Research.