AI-Based Empirical Asset Pricing: A Solution to a Long-Standing Challenge in Hoperation Fund’s Investment Portfolio

As the manager of Hoperation Fund’s investment portfolio for more than three years, I have consistently faced one major challenge: the portfolio’s performance curve has been volatile. Although the portfolio’s value has increased year after year, its growth has not followed a smooth, steady trajectory. Instead, the performance curve has fluctuated significantly, with frequent rises and declines, as shown in the figure below.

Of course, from a purely investment-oriented perspective, I do not view volatility itself as a major concern. As Warren Buffett and Charlie Munger have often emphasized, volatility is not the same as risk.

However, Hoperation Fund’s investment portfolio frequently experiences cash inflows and outflows. When we successfully complete a fundraising campaign, new capital flows into the portfolio. When we donate a portion of the portfolio’s profits to nonprofits and educational organizations, capital flows out.

This creates a serious practical challenge. If new funds enter the portfolio when the stock market is overvalued, investing that capital at elevated prices could expose the portfolio to greater losses during a subsequent market correction. Conversely, if funds must be withdrawn when the market is near a low point, we may be forced to sell holdings at depressed prices, further reducing the portfolio’s overall return.

Therefore, from a portfolio-management perspective, I have always disliked seeing the portfolio’s return curve rise and fall so sharply. I have long hoped to find a strategy that could produce a smoother, nearly linear growth trajectory. With a more stable return curve, cash inflows and outflows would be less likely to create unnecessary adverse effects on the portfolio’s long-term performance.

Recently, I finally discovered a promising solution: constructing a portfolio with AI-based empirical asset pricing.

As shown in the figure below, from the beginning of 2026 through today, a monthly-balanced portfolio constructed using this framework (indicated by "New Portfolio") would have produced a smooth growth curve. Its trajectory closely resembles a steadily rising straight line. Over this period, the portfolio had a maximum drawdown of -3.5%, an average drawdown of -1.1%, annualized volatility (measured as the annualized standard deviation of daily returns) of 10.8%, a Sharpe ratio of 4.13, and a cumulative return of 29.4%!

By comparison, the growth curves of the broad-market index funds SPY and QQQ fluctuated much more substantially. Their maximum drawdowns were -8.9% and -11.7%, respectively. Their annualized volatilities were 13.8% and 21.2%, while their Sharpe ratios were 1.02 and 1.05. Their cumulative returns over the same period were 9.5% and 13.6%, respectively.

I also extended the backtest to earlier periods, and the results were broadly consistent. Most notably, as shown in the figure below, the new portfolio would have generated a positive return of 7.9% during the 2022 bear market, while SPY and QQQ returned −18.1% and −32.6%, respectively.

In 2022, the new portfolio also had a maximum drawdown of just −4.1%, an average drawdown of −1.6%, and annualized volatility of 6.3%. By comparison, SPY and QQQ had maximum drawdowns of −24.5% and −34.8%, respectively, with annualized volatilities of 24.2% and 32.1%.

So far, I am satisfied with the strategy, particularly because I believe it is well suited to the second half of the year, given that Hoperation Fund’s Market Top Signal has recently issued a warning. In response, I have liquidated approximately half of the short-term holdings in Hoperation Fund’s investment portfolio and used the proceeds to build a new portfolio based on this strategy. I will present the new portfolio at our next regular meeting in August.

Going forward, I will continue investigating this methodology and searching for more effective AI algorithms to implement it, with the goal of further reducing volatility while improving returns.