What you ask for is non-trivial but I suggest working in reverse. First decide what sort of curve you want then work your varibales from that. Skewed curve probabilites can be calculated using Binomial Distrubutions, but when the population is large enough (as in your case) then normalized approximations can be used.

Try this link for more info:

http://online.moraine.cc.il.us/Summer/2002/MTH139-303/lesson17.htm

Hope it helps.

Try this link for more info:

http://online.moraine.cc.i

Hope it helps.