In The Sledgehammer Approach, I compared theism to believing that a test subject hit an egg with a sledgehammer while blindfolded - in the absence of actual evidence in support of that belief.
There are of course problems with the analogy.
Some theists would say that the nature of god is such that the test subject could not have missed an egg (this is equivalent to the claim that god is necessary, the theological position that it is impossible for god not to exist). Such theists would argue that I was vague in that I said "give (the test subject) a sledgehammer" and "take some eggs" rather than specifying that the experimental protocol would require either an infinitely large sledgehammer or an infinite number of eggs, thus making it impossible for the test subject to miss.
Some atheists would say that if any eggs were put out, then that would be ignoring the fact that the god in question is logically impossible and cannot exist. I'd probably agree if we were talking about a specific god, but strictly speaking when we are talking about theism, the god isn’t specified.
Then there is the problem of "prior probability". If only one very small egg was put out and the sledgehammer's head was tiny (a pinhammer or needlehammer, perhaps), then the prior probability of hitting an egg would be very small. Alternatively, if a large (but not infinite) number of eggs were put out and the sledgehammer was a monster, then it'd be highly likely that an egg would be hit. This wasn't really what I was getting at with the article, I wasn't trying to be accurate about the likelihood of a god existing, I was only trying to explain the shading involved with being an atheist - so something like a 50-50 chance of hitting an egg was fine. This also has the benefit of being a strong-ish version of the theist position.
The size of the sledgehammer's head, however, raises another issue. Evidence. The test subject would be justified in assuming a low probability to hitting an egg if the sledgehammer was light (and the head small) and a higher probability if it was heavy (and the head large). Then there are the sounds and smells involved. The test protocol only called for a blindfold, but an egg being smashed by a sledgehammer might make a distinguishable noise and the albumen and yolk could possibly release a scent that the test subject could pick up. Also, the sledgehammer might have been a bit slippery after the strike, if an egg was smashed. There are enough variations between potential test subjects with respect to their sensitivity that some might smash eggs without noticing while others would believe that they had smashed one when they hadn't. Their beliefs, therefore, would be subjective - overdetectors would detect smashed eggs when there aren't any, underdetectors would miss a lot of smashed eggs. (This is equivalent to "agent detection" - some atheists, like myself, reckon that theists have hyperactive agent detection.)
The point in the analogy is that there isn't any conclusive evidence, which is precisely the position that atheists take. A theist might want to argue that there is conclusive evidence for their god, but the atheist cannot really accept that argument without becoming a theist.