Creationist: "Evolution simply doesn't work. Take a fairly typical protein, 200 amino acids long. There are 20 different amino acids, so the probability of this single protein arising at random is 1 in 20200 - that is, 1 in 1.6*10260. Since the number of protons in the universe is less than 1080, if every proton was a universe containing the same number of protons, and if every one of those protons were also a universe also containing the same number of protons, there is a chance that in one of these universes, this protein might appear at random if all the protons were involved in sequencing amino acids (rather than basically being fuel in stars). The short way for describing this level of improbability is 'impossible'. And that's just one protein."
Evolutionist: "Ah, but you are making unnecessarily restrictive assumptions. Yes, the appearance of that specific protein at random might be very improbable, but firstly you wouldn't need to specify that precise sequence of amino acids to get a protein with the same functionality. Also, there is no reason to think that there aren't lots of different possible proteins, derived from lots of different routes, that would have the required functionality. Finally, once the protein is partly specified, natural selection can work on it to improve its functionality."
Me: "OK, but how specified does a protein have to be to have limited functionality? Let's say that in a 200 amino acid protein, only half the amino acid sites are required to specify the protein. Also, that as long as one of four possible amino acids is specified at each of these sites, the protein will have its required functionality. The probability of this protein arising at random is then reduced to approximately 1 in 5100 - that is 1 in 7.9*1069. So we have gone from the impossible to the astronomically unlikely - we are still talking about an improbability of a similar order of magnitude to one in the number of protons in the universe. And we are still only talking about one protein. Don't like these figures? Suggest some different ones, based on empirical data. And remember that before a protein has some functionality, natural selection won't work on it to improve it. If you can't suggest plausible figures, given the amount we know about proteins and so on today, then you don't have a theory."