Benchmarking Constraint-Preserving QAOA for Renewable-Energy Portfolio Selection (Academic research)

Summary

• From May 1-October 12, 2026, V2 returned -3.41%, while renewable/SPY blends returned +2.50% and +5.46%, versus +8.41% for SPY.p.1
• Across 18 settings per mixer, trained probabilities imply 5.849 expected X recoveries and 13.320 expected XY recoveries with 64 shots; one corrected rerun observes 5 and 16.p.1
• A 72-run smaller-penalty sweep does not establish a reliable remedy for the penalty-X baseline.p.1
• Mean pairwise return correlation ranges from 0.31 to 0.52, revealing limited diversification despite seven or eight nominal assets.p.2
• The financial study executes 91 nontrivial QAOA decisions, 23 from V2 and 68 from the expanded universe.p.12

Key insights

1. No quantum speedup or independently validated investment alpha is established.p.1
2. Blended gains reflect market exposure rather than demonstrated quantum investment value.p.1
3. Effective correlation rank ranges from 2.62 to 4.61 across evaluation windows.p.2
4. Version 3 increases market exposure, with mean cash falling from 71.2% to 42.6% in the observed 2026 window.p.6
5. During the extension window raw return rises from 28.58% to 36.67% while Sharpe falls from 2.20 to 1.81.p.6
6. For May 1-Oct 12, V3 returned -12.01% against 8.41% for SPY.p.10
7. The 25/75 blend returned 5.46% for May 1-Oct 12, but all blends lag SPY.p.11