Factor investing is and always will be an exercise in data mining. It is still performance chasing, albeit over much longer time periods. I think tracking error limits are exactly what factor investing needs. Massive sector bets, loose definitions of factors, academic dissonance of what is a "risk" factor or a not a "risk" factor, behavioral issues, portfolio construction assumptions, market structure changes, and things like turnover assumptions all point to factors being fuzzier than your run of the mill comfy academic will admit. I would wager very small and informed tilts against the market thousands of times over is going to have a much better result than the current "optimized" approach of most major factor investors. You don't really need to look back that far for proof that "optimization" is the stuff of fairy tales.