Predicting copyright fair use

I've always thought that copyright's "fair use" factors are little more than a guide for lawyers to organize their arguments (and judges to organize their decisions). In my opinion, those factors are not a useful tool for successfully predicting, in advance, whether any particular use will be deemed fair or infringing. Loyola of Chicago law professor Matthew Sag has a different opinion. He thinks some factors do lead to predictability, and he explains why in an article titled Predicting Fair Use. The abstract:
Fair use is often criticized as unpredictable and doctrinally incoherent - a conclusion which necessarily implies that the copyright system is fundamentally broken. This article confronts that critique by systematically assessing the predictability of fair use outcomes in litigation. Concentrating on the characteristics of copyright litigation that would be apparent to litigants pre-trial, this study tests a number of doctrinal assumptions, claims and intuitions that have not, until now, been subject to empirical scrutiny. This study finds that commercial uses are not disfavored in fair use cases, nor is the unauthorized use of unpublished or creative works. In contrast, complete copying by the defendant does appear to be a significant predictor that the defense of fair use will fail. This article also presents empirical evidence that transformative use, or at least specific kinds of transformations, makes a finding of fair use more likely. This article concludes that, based on the available evidence, the fair use doctrine is more rational and consistent than is commonly assumed. The fair use is more than simply the right to hire a lawyer and take one’s chances in court.