The study used machine learning methods to analyze how different factors affect option returns. By creating portfolios based on these factors, they found that portfolios reflecting volatility risk had higher returns compared to other strategies. This approach does not rely on specific assumptions about how factors influence returns, making it a valuable tool for understanding the relationship between characteristics and investment outcomes.