Divide the winning lottery draw list in half.
Use the first half (the oldest draws) to generate which random seed does the best.
Use the second half (the newest draws) to see if it continues to produce at the same percentage of hits.

My research into this so far has eliminated the possibility that there is just one seed that has chained 1570 draws. (I ran at 1 million as whole integers AND 1 million by .5 increments... anything less than .5 on the seed generates the same number, so .5 is the sensitivity threshhold)

Now the test must move more to the "reverse engineering" phase.

There are several things I can try, most center around guessing when they click "Run" on their program and estimating the timestamp. I actually thought about dividing the list up by years, but this thing is either going to work against the whole list or not at all. The random numbers are based on a pre-determined list, not sure if it's provided by Microsoft or Intel... If they are using a source of entropy other than the CPU's built-in list, then this won't work anyway.

my next move is to swapout the INT command which just uses the integer portion for the ROUND command which will round up when needed, it will change about 1/2 the list.

Good luck with your experiments.   I remember when you talked about this idea way back.    I think it's tougher than the proverbial needle in a haystack.

Thanks.
It does seem impossible, but it is a great exercise in programming for me. Even if I get nowhere with results... I have learned a great deal about Arrays, reading from and writing to text and CSV files, global vs. local variables, UI construction... Educationally, I did win!

Next attack will actually test my Date-as-seed theory AND capture the possibilities of them doing test draws... more on that later.