Artificial Intelligence and Portfolio Management: A Comparative Backtest of an AI-Assisted Portfolio and an Equal-Weight Portfolio
0 viewsAbstract
This study examines whether an artificial-intelligence-assisted asset allocation can improve portfolio performance relative to a simple equal-weight portfolio. Using a fixed set of five U.S.-listed exchange-traded funds representing large-cap equities, growth-oriented equities, gold, long-term Treasury bonds, and real estate, an AI-generated allocation was compared with a 20%-per-asset equal-weight benchmark over the 2021-2025 period. The AI-assisted portfolio allocated 30% to SPY, 30% to QQQ, 15% to GLD, 10% to TLT, and 15% to VNQ. Annual total-return data with distributions reinvested were used to calculate annual returns, cumulative return, compound annual growth rate, volatility, worst annual return, maximum drawdown, and a descriptive Sharpe ratio. The AI-assisted portfolio reached a cumulative return of 76.16%, compared with 56.57% for the equal-weight portfolio, and produced a higher descriptive Sharpe ratio (0.67 versus 0.59). However, it also showed higher annual volatility (20.49% versus 18.49%) and a slightly larger worst annual loss. The results indicate that, in this historical sample, the particular AI-generated allocation had a different risk-return profile and a higher cumulative outcome than the equal-weight benchmark. The study does not establish that AI consistently predicts markets or outperforms traditional methods, and the findings are limited by the small sample, single AI prompt, annual frequency, and exclusion of transaction costs.
Keywords
Artificial Intelligence, Asset Allocation, Backtesting, Diversification, Investment Performance, Portfolio Management, Sharpe Ratio