[IsaacW]

First you'd need to estimate the probability that a particular pot is killed, let's denote that P(kill).

Assuming that the bets double when the pot is killed and that a particular hand being killed does not affect the probability that the next hand will kill, then the average big bet size (in terms of the normal big bet size BB) is:

BB*(1 - P(kill)) + 2*BB*P(kill)

The second assumption is a bit tenuous, because in some kill games the kill is based on the size of the pot, and obviously the pots are more likely to be over the kill amount if the bets are double-sized. If the kill game was the type that killed after the same person won two pots in a row, then I imagine the second assumption would be OK.