Quantum-Assisted Optimal Rebalancing with Uncorrelated Asset Selection for Algorithmic Trading: Walk-Forward QUBO Scheduling via QAOA

Battle Count: 77

Directly addresses core quantitative trading problems: portfolio construction (asset selection, weight optimisation) and rebalancing scheduling. The walk-forward QUBO/QAOA formulation for rebalancing timing is novel and practically relevant. Results show 44.5% transaction cost reduction with improved Sharpe ratio. The framework is end-to-end from universe construction to backtesting. However, the quantum advantage claim is preliminary given the small problem size (W=8) and classical simulation, and the single-year test limits practical deployment confidence.

The framework integrates multiple components: Ledoit-Wolf shrinkage (scikit-learn), Ward's hierarchical clustering, entropy-regularised GA on GPU (PyGAD + PyTorch), QUBO matrix construction with custom penalty terms, QAOA circuit design (Qiskit), COBYLA optimisation with multi-restart, walk-forward scheduling logic, and a full backtest engine with transaction cost accounting. Each component is individually implementable, but the end-to-end pipeline requires careful integration. The QAOA circuit is small (8 qubits, depth 2) but the overall system complexity is moderate-to-high. No code repository is provided.

The paper provides detailed hyperparameters (Table 3), software stack (Table 1), QUBO formulation equations, and walk-forward algorithm pseudocode. Data is sourced from yfinance (publicly available S&P 500). Qiskit and PyGAD are open-source. However, no GitHub repository is provided, the GA implementation details (PyGAD configuration) are partially specified, and the single-year out-of-sample test (2025) limits statistical robustness. The QAOA simulation uses Qiskit Aer statevector (CPU), which is reproducible but the exact random seeds for multi-restart are not specified.