DavidMLewis wrote:The assumption behind the logic of changing one thing at a time is that the variables are independent, which as you state, they aren't in the case of espresso. So to really understand what one variable does in isolation, you have to change that variable and then re-optimize the related ones at the new setting, I think.
The assumption is that if you change one variable, it will change the pull in a predictable way. It will either improve a result, or not. The corrolary is that if you change it the other way, it will either make it worse, or not. It answer's the question, "What happens if I do this?"
If you start out having bad luck, and you change two or three things, you won't ever be able to figure out which change had the most positive effect in the result, and which one will do what.
Randomly changing variables without any clue as to what does what is shooting blind. Ya can't steer the careening car away from the cliff's edge if you don't know which lever is the velocitator and which one is the deceleratrix, or what that big-roundy-thingie in front of you is supposed to do.
One at a time to discover what does what, and then start pairing them so see how one affects the other, and then, only after you have a fairly keen grasp of what they are, and
how they interact, can you change multiple variables with any hope of improving the result any better than blind chance.
The assumption that one should change all the variables that are off presumes that the operator already knows which ones are off, by how much, and in what direction. Not to mention what level of change will effect what other variables and by how much.
That's purty danged presumptuous...