I use what is called a Poisson distribution, assuming that all purchases are effectively random.
When the number of tickets purchased is very small compared to the odds, a Poisson distribution will give numbers that are close to the numbers obtained by dividing the number of tickets purchased by the probability. However as the number of tickets purchased increases, there is a greater probability that each new purchaser will purchase a ticket with a set of numbers that has been already purchased by someone else. We all are intuitively aware that the larger the number of tickets that are sold, the more frequently we will see jackpots won by more than one person. These possibilities are measured by Poisson calculations.
The mathematics of Poisson distributions are available on the internet and in many other places and I encourage you to investigate them.
I have built a series of spreadsheets that do all of my calculations, and produce the results that I provide here from them. I have arranged it so that I merely enter the sales for each drawing in a box on these spreadsheets and all of the values pop out. I then cut and paste them here for informational purposes.
Note that I only report values that have a probability of 1 hundredth of one percent. The probability of there being six winners in the upcoming draw is not zero. It is possible that there will be six winners, not very likely, but possible.
Sources of error in my calculation include the fact that purchases are not always random. For instance people play their birthday numbers frequently, which tends to skew the probability of rollovers in someway by creating more doubly covered sets of numbers (those with all of the numbers being less than 31) and creating more uncovered sets (those which do not represent birthdays.) There are undoubtedly other such effects, however the Poisson calculations are a useful first approximation.
The long term probabilities I post are based on a mathematical technique known as curve fitting. I have graphed the sales figures for various jackpot sizes, made an estimate of the type of curve that fits, for MM an exponential function, and for PB a linear function that switches at a certain point to an exponential function, and have calculated certain constants. These constants are automatically updated for each drawing by a series of linked spreadsheets.
I hope you find them useful and I hope you find these comments useful.