Isn’t this academic curve-fitting at its worst? The lagged-returns factors are clearly correlated (despite one of them not being in the “Seasonality” category), using the debt-to-market factor would mean you’d increase your position every time a company’s debt increased, and the intrinsic value measure here is a million miles from what investors actually use. These guys aren’t looking at the factors themselves to figure out what might go with what and what would work in the real world, but are doing a crazy number of multiple linear regressions, which is a much worse way to figure anything out than using combined ranking. If I were teaching finance (God forbid), I’d give this paper a C.