I want to try an n=1 experiment where I do a random Fermi problem in some time limit, once or twice a day. After several days of this, I'd like to graph my estimation error over time, to see the effect of practice on my performance. I could even try varying certain treatments, like caffeine or chattering background noise, to test their effects on my estimation error, after adjusting for the practice effect.
Some Fermi problems are straightforward to verify - the classic "piano tuners in Chicago" was answered with a phone book. Others can only be verified with another expert's estimate, and the question might be too obscure for an Internet search to yield an instance where an expert had already done the math. I think I need a source of Fermi problems that I know a priori to be verifiable.
* To verify that a question has a solution, I usually have to do a Google search, which exposes me to the answer.
* If I search for a new one each day, I fear I would quickly become biased towards selecting easier questions.
* If, before starting the experiment, I cobble results together across many Google searches that seem likely to be verifiable, I fear I would have more days to dwell on the ones I do later in the experiment, exaggerating the practice effect.
Does anyone have any advice?
