I actually see that as using a bayesian approach to statistics. Maybe not as formal, but still utilizing some form of statistics there. I can't say for certain how you formulated your priors (E(u_t+1|earnings) = -2%) > E(u_t+1|earnings) = +2%), but again, though it may not be formal, I would think there was a lot of historical data already that went into those formulations (led you to those beliefs). FWIW, I am a bigger proponent of bayesian approaches over frequentist, when it comes to trading and markets.
I've found mine through data and "testing". Also testing gives you invaluable metrics which you cannot find out by philosophizing while smoking a cigar.
An edge picked out of a chart like a head and shoulder pattern is much better than an edge brute-searched over a dataset , an edge should have a concept that can be exploited, testing is not going to give you a concept, a little different way of thinking you may say
And that has massive subjective bias. You'll pick out the trades that worked for the pattern and quietly discard (or assign lower weight) to the ones that did not work out. I've done this myself many times. I stare at charts for hours and have one those epiphanies but when I actually crunch the numbers to remove my bias, the profitably is far lower or even negative.
Testing might not provide a concept, but it can determine if a concept is likely invalid or possibly valid.
Its not about picking and discarding, the pick should be enormously profitable than the discards. The basic thing is working on data reveals low quality and sub standard edges, working on charts and on taming randomness produces wonderful results, that work. You need to change the way you think i guess, but when you get it , you will know it.
Testing to find an edge is like trying to break a password by brute force, Finding an edge by chart exploration,derivation is like breaking the encryption of the password