Otherwise, my vote is 100% all the time. The way I see it, there’s a lot of text out there. There’s probably billions (actually who knows how many) of permutations of sentence variables indexed. Under the assumption (please tell me if I’m wrong) that Google uses Bayesian filters to detect bad permutations / grammar, I would think that as long as it makes sense grammatically (is a good permutation) and unique, it will work. It’s like a gift that the sentences are better at 100%.

I’m going to go try it now. Thanks for pointing it out!