Thanks, Alex. Your formula was exactly what I needed. Like I said too much accuracy in curve counting is not critical as the only consequence is you move on to a larger B1 a liitle too early or a little too late  no big deal.
Another ECM puzzle for you: I want to hand out ECM assignments such that I roughly maximize the chance of finding a factor for a given effort. What formula would you use given the quantities total_ecm_effort (the sum of completed B1 values) and N (the Mersenne exponent). Ignore the step function nature of FFT lengths and assume the time it takes to run an ECM curve is proportional to N (log N).
A simpler way to phrase this is: If an exponent has had a total_ecm_effort X, what is the probability the next ECM curve will find a factor (to make life easier, ignore the step nature of assigned B1 values).
Given such a formula I can then use the expected CPU run time to maximize factors found.
