Gross profits divided by total assets turned out to be a surprisingly durable predictor, and it helped explain why some cheap companies underperformed: they were cheap because they earned very little. Combining valuation with profitability produced better results than either alone.
Profitability was later formalised as one of the added terms in the five-factor extension of the fama-french-three-factor model, alongside an investment or asset-growth term. It overlaps heavily with the broader quality-factor used by index providers.
As with all accounting-based signals, the measurement point matters. Using figures before they were publicly filed produces look-ahead bias and a backtest that could never have been traded. See backtesting.
Related: quality-factor, fama-french-three-factor, value-factor, factor-investing, backtesting, eps