Replace parsimony with Marcomony!
Reverse Ockham's razor and replace it with Mahkco's bag of 50 disposable ones.
This is critical with the historical sciences especially, and not really relevant for physics at all. Basically with quantitative models where calculations yield repeatable results one should certainly pick the simplest model which makes the numbers work precisely. Not only that, but any new theory which complicates a model of this nature needs to have a heap of evidence and the burden of proof is against the new, more complicated theory. In these cases we should be Jack McCoys - It's not what you know, it's what you can prove. This is Ockham's Razor, or parsimony.
With historical sciences evidence is scattered in time and space compared to the "experiment" (ie. nature has done the experiment for you, but it cannot be repeated precisely to check or "prove" anything), so Ockham's Razor is a cut too deep. Not to mention that the "simplest" explanations of all often involve God, or at least an explanation that relies on faith, because by their nature the hardest questions of history have the least incontrovertible evidence. With these questions we *constantly* need to be detectives. We follow our hunches - we need a gazillion of them. EVERY scrap of evidence, no matter how circumstantial, no matter whether it is something that shouldn't be there, that is, or something that should be there that isn't, peculiarities even if they appear to be random peculiarities. These should be only weeded *roughly*. We rule out any that are impossible based on naturism, mathematical probability etc. *Not ones just because they are impossible to prove or have no direct evidence*. We need to be Sherlock Holmes, not Jack McCoy. We don't keep looking for something in a particular spot because that is where the light is shining - we need to feel our way in the dark and glean complex enough theories that make the tiniest scraps of circumstantial evidence meaningful probability-wise. We cannot stick with one theory because it is the simplest and has the most provable elements. We need a dozen, more complicated, theories that fit the circumstantial evidence in the middle of their probability range.