Comparing Traditional and AI Small-Cap Stock Selection

Summary

The document compares a traditional Chinese equity strategy that buys the smallest market-cap stocks with a StockRanker model that ranks stocks using market capitalization as its sole feature. Both are evaluated over 2015–2016, illustrating how a machine-learning ranking approach can be applied to a familiar factor. It explains that small-cap exposure has historically been associated with excess returns in the domestic market, while also being common among funds and potentially contributing to crowded, similar portfolios.

The reported backtest gives the AI strategy a 289.46% total return, a 35.22% maximum drawdown, and a Sharpe ratio of 5.77; the traditional approach is reported to grow 1,000 yuan to 3,120 yuan. The AI version had somewhat higher drawdown, with similar return volatility. These are historical results from a short, dated period, and the document provides no details on costs, benchmark, validation design, or robustness. Its claim of AI superiority therefore should not be treated as evidence of persistent live performance.

Key ideas

• The traditional strategy forms a portfolio from stocks with the smallest market capitalizations.
• The AI strategy uses a StockRanker model with market capitalization as its only feature.
• The reported AI backtest has higher returns and a higher Sharpe ratio, alongside somewhat greater maximum drawdown.
• Both strategies share exposure to the size factor, which can make portfolios crowded and similar.
• The comparison covers only a historical two-year period and does not establish future performance.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.