Adam
The higher the lexi value used the more numbers that will be returned. Setting my my 5-39 game to 500/575 gives
a range of 13 to 35 which is a sharp contrast to setting 000/ 100 which returns 01/02. The last 75,757 draws in the
matrix covers much of 13 plus all draws with first numbers 14 to 35. When playing this range use the same range
for both high and low. Remember that the range is padded by the program even though the user only inputs the
first 3 digits. Setting low to 535 and high to 535 uses draws 535,000 to 535,999. The higher the value the tighter
the range needs to be. If I can select the first digit then the 2nd and 3th are ranged to include my best choice.
Lets say that I think that the first digit will be 4. I set low 400 to 499 and run and get numbers from 8 to 13. Since
this program was originally part of DM's bayesian predictor it would have then checked the digits 890123 to cut the
list down. Next it would remove any numbers from the list based on the digits it had selected, kind of a crosscheck.
Since Bayes was tuned to my 5-39 game and digits 1-2-3 were given priorty it would first check to see if digits 8-9-0
had been blocked. If all were still in play then the probabilities for each digit would be considered. It would
continue to do this checking the outcome against every filter trying to reduce the numbers down to 3 or less with 3 being
acceptable. This took thousands of calculations before making a choice and to do this by hand would be impossable.
What I would do in this situation is this. I would break this down into 5 ranges, 400 to 420 = 08-09, 421 to 440=09-10
441 to 460=10-11, 461 to 480=11-12 and 481 to 499=12-13. Since digit 0 is the lowest hitting digit in my matrix I
might take out group 2 and 3 beacuse every other number could be gotten from the remaining groups and then I
would play what I feel would hit or just select one at random each having a 33% chance of hitting as long as I was
correct about digit 0 not hitting. I have yet to come up with any proven method of selecting the range without the
bayesian program which will never be released and is not working at the time, so I guess for now we are limited to
using the simple tools provided in the program or making a random selection and hope for the best.
RL