Suppose that I live on a holodeck but don't know it, such that anything I look at closely follows reductionist laws, but things farther away only follow high-level approximations, with some sort of intelligence checking the approximations to make sure I never notice an inconsistency. Call this the holodeck hypothesis. Suppose I assign this hypothesis probability 10^-4.

Now suppose I buy one lottery ticket, for the first time in my life, costing $1 with a potential payoff of $10^7 with probability 10^-8. If the holodeck hypothesis is false, then the expected value of this is $10^710^-8 - $1 = $-0.90. However, if the holodeck hypothesis is true, then someone outside the simulation might decide to be nice to me, so the probability that it will win is more like 10^-3. (This only applies to the first ticket, since someone who would rig the lottery in this way would be most likely to do so on their first chance, not a later chance.) In that case, the expected payoff is $10^710^-3 - $1 = $10^4. Combining these two cases, the expected payoff for buying a lottery ticket is +$0.10.

At some point in the future, if there is a singularity, it seems likely that people will be born for whom the holodeck hypothesis is true. If that happens, then the probability estimate will go way up, and so the expected payoff from buying lottery tickets will go up, too. This seems like a strong argument for buying exactly one lottery ticket in your lifetime.