Evaluating Short-Term Equity Portfolio Performance: A Comparative Study of AI-Driven and Human-Managed Strategies

Engineering notes

Engineering notes will be added by the aipentium editorial team.

Chinese explanation / 中文解读

中文解读待补充：本站会优先为大语言模型、生成式AI、ChatGPT相关技术、计算机视觉、深度学习等高价值论文补充中文说明。

Original abstract

Despite recent developments in artificial intelligence (AI) for portfolio management, it is still unknown how free AI agents compare to human fund managers in constructing short-term, high-risk portfolios. This study uses an experimental design model with the aim of evaluating whether free AI tools can outperform human managers. Four professional human fund managers and three free AI agents (ChatGPT, Gemini, and Perplexity) created portfolios consisting of ten global stocks, with a starting portfolio value of SGD $10,000. Portfolios were assessed over a one-month period from November 17th to December 16th, 2025. To analyze the data, portfolio returns and risk were calculated using the percentage change in value and portfolio standard deviation through a covariance matrix, respectively. The results showed that human-managed portfolios (-3.33% return) slightly outperformed AI-managed portfolios (-3.46% return), with a lower average risk of 0.964% compared to AI portfolios’ 1.603%; however, AI portfolios demonstrated better risk-adjusted performance as measured by Sharpe ratios (-2.010 versus -4.834). No clear relationship between risk and return was found for either group. AI-managed portfolios were volatile and concentrated in high-growth sectors, whereas human-managed portfolios tended to deliver more consistent returns across diverse sectors. These preliminary findings suggest the importance of human judgment while highlighting AI’s potential to help with investment decisions, though further research with larger samples is needed to confirm these results.

Links and sources

Comments