<p>We propose an algorithmic trading framework that addresses two core portfolio management problems simultaneously: principled asset selection and optimal rebalancing scheduling. Asset selection is performed using Ledoit–Wolf shrinkage covariance estimation with hierarchical correlation clustering to select <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(n=10\)</EquationSource> </InlineEquation> maximally uncorrelated stocks from the S&P 500 universe without survivorship bias. Portfolio weights are optimised via an entropy-regularised Genetic Algorithm (GA) accelerated on GPU, alongside closed-form minimum-variance and equal-weight baselines and a three-way ensemble. The central contribution is the formulation of the portfolio rebalancing schedule as a Quadratic Unconstrained Binary Optimisation (QUBO) problem, solved using the Quantum Approximate Optimisation Algorithm (QAOA) — a classical variational circuit simulation — within a walk-forward framework that eliminates lookahead bias. This recasts dynamic rebalancing as a combinatorial optimisation problem amenable to variational methods. We note that the QAOA implementation used here runs on a classical statevector simulator; no quantum hardware is employed, and the formulation is intended to demonstrate the viability of the QUBO structure and walk-forward scheduling methodology as a stepping stone toward future quantum hardware deployment. Backtests on S&P 500 data (training: 2010–2024; test: 2025, <InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(n=249\)</EquationSource> </InlineEquation> trading days) show that the GA + QAOA strategy achieves a Sharpe ratio of 0.588 and total return of 10.1%, compared with 0.575 for the strongest classical baseline (GA with 10-day periodic rebalancing). The QAOA-scheduled approach executes 8 rebalances versus 24 for the classical benchmark, corresponding to a 44.5% reduction in transaction costs. Multi-restart QAOA with 4,096 measurement shots demonstrates concentrated probability mass on high-quality rebalancing schedules, indicating stable convergence of the variational optimisation procedure. These results provide preliminary evidence that QUBO-based rebalancing scheduling can reduce turnover while maintaining competitive risk-adjusted performance relative to classical rules applied to the same portfolio. We acknowledge that a single-year out-of-sample test is an exploratory result, and multi-year rolling evaluation is identified as a primary direction for future work.</p>

Optimal rebalancing with uncorrelated asset selection for algorithmic trading

期刊论文
发表日期：

2026年7月6日

查看全文

Abraham Itzhak Weinberg

摘要

We propose an algorithmic trading framework that addresses two core portfolio management problems simultaneously: principled asset selection and optimal rebalancing scheduling. Asset selection is performed using Ledoit–Wolf shrinkage covariance estimation with hierarchical correlation clustering to select \(n=10\) maximally uncorrelated stocks from the S&P 500 universe without survivorship bias. Portfolio weights are optimised via an entropy-regularised Genetic Algorithm (GA) accelerated on GPU, alongside closed-form minimum-variance and equal-weight baselines and a three-way ensemble. The central contribution is the formulation of the portfolio rebalancing schedule as a Quadratic Unconstrained Binary Optimisation (QUBO) problem, solved using the Quantum Approximate Optimisation Algorithm (QAOA) — a classical variational circuit simulation — within a walk-forward framework that eliminates lookahead bias. This recasts dynamic rebalancing as a combinatorial optimisation problem amenable to variational methods. We note that the QAOA implementation used here runs on a classical statevector simulator; no quantum hardware is employed, and the formulation is intended to demonstrate the viability of the QUBO structure and walk-forward scheduling methodology as a stepping stone toward future quantum hardware deployment. Backtests on S&P 500 data (training: 2010–2024; test: 2025, \(n=249\) trading days) show that the GA + QAOA strategy achieves a Sharpe ratio of 0.588 and total return of 10.1%, compared with 0.575 for the strongest classical baseline (GA with 10-day periodic rebalancing). The QAOA-scheduled approach executes 8 rebalances versus 24 for the classical benchmark, corresponding to a 44.5% reduction in transaction costs. Multi-restart QAOA with 4,096 measurement shots demonstrates concentrated probability mass on high-quality rebalancing schedules, indicating stable convergence of the variational optimisation procedure. These results provide preliminary evidence that QUBO-based rebalancing scheduling can reduce turnover while maintaining competitive risk-adjusted performance relative to classical rules applied to the same portfolio. We acknowledge that a single-year out-of-sample test is an exploratory result, and multi-year rolling evaluation is identified as a primary direction for future work.