A worry about Pascal's Wager that I've had for a long time is that there are different sizes of infinities.
Let f(R) be the cardinality of the set of pleasurable experiences promised by religion R. It seems that for any religion A, once could make up some religion B such that f(A) < f(B). Now, my credence in B will probably be less than that of A, but credences are all real numbers as far as I'm concerned. So, as long as my credences are nonzero, if f(A) < f(B), then I should always favor religion B to A. But, if for any religion, there is another religion I should follow rather than that one, then I have no way to choose which religion to follow.
This worry has never been huge in my mind, since for any any infinite set of pleasurable experiences, I believe that Christianity is my best chance at attaining a set of pleasurable experiences comparable to that set. But this solution depends on my specific credences being lined up in a certain way. For other distributions of credences, one could easily be led to decision-theoretic gridlock. Here's a solution that hopefully bypasses this worry:
We have the ability to consider two states of affairs and decide whether one is preferable to another. But we haven't considered absolutely every state of affairs. (For example, until I sit down and really think about it, I am not sure whether I would rather spend an eternity at a beach or in a forest.) Maybe there exists some state of affairs that would obviously be preferable to all other states of affairs if we only considered it (but alas we haven't). Let's call such a state of affairs the Ultimate Good (UG). If we are being consistent, we should assign a nonzero credence to the proposition that the UG exists (even if we have no idea what it would look like). Now, this means that a rational agent will only pursue courses of action that have the highest chance of actualizing the UG. Any other good state of affairs -- even if infinitely desirable -- aren't worth the time of day. A rational agent goes straight for trying to actualize the UG.
