Bayesian search techniques for optimal shaving solutions
When I first learned Go, I couldn’t stop seeing Go positions in the objects around me; now, I can’t stop seeing examples of Bayes’ theorem in action—and let me again recommend McGrayne’s The Theory That Would Not Die.
I just realized that I have based my shave recommendations on a Bayesian foundation, in this sense: I have long recommended—in print—that a beginner’s second razor should be a Slant Bar. (For the first razor, I initially recommended the Merkur Hefty Classic (aka “HD”), but switched to any of the Edwin Jagger DE8x series once that razor was available with the new head design, and that continues to be my recommendation.)
A guy on the Wicked_Edge posted that he started with the Weishi, then tried the Gillette Red-tipped Super Speed, and then a Gillette Slim adjustable. In the course of his description of his razors and their performance, he mentioned that he had a thick, tough beard. That, to me, is Slant Bar bait.
So of course I immediately weighed in to recommend he try a Slant, and wrote:
In my book I stress that the beginner’s second razor should be a Slant Bar (and I recommend one of the Edwin Jagger DE8x series as a first razor).
For much the same reason—gaining experience by trying something different—I would recommend that a beginner’s second brush be of a different sort of bristle than his first: If he began with boar, get a badger; if a badger was first, get a horsehair second; and so on.
I’m reading (with great interest) a book about Bayes’ theorem, so of course I’m now seeing it everywhere, and my recommendation of making a big rather than an incremental change is quite Bayesian and is more likely to lead more quickly to optimal outcomes.
And I realized as I described this to The Wife (who hears, I think she would agree, rather a lot of shaving theories from Yours Truly) that this is a search strategy that quickly gives information on the overall characteristics of the search space by testing widely separated points—for one, a search space of razors; for the other, brushes.
Indeed, now that I have the notion, I can do it better. For razors, for example:
a. Edwin Jagger DE8x
b. Slant Bar
d. Straight Razor
Again: the idea is to avoid trying incremental change (i.e., nearby points in the search space) and instead trying a different neighborhood. In fact, I could be persuaded that “c.” should be the Straight Razor.