I've been diligently been running all combinations for 5/39 for different filters.
The first set is what I call static filters because they are not time dependent or reflect the current status of any lotto numbers. Things like consecutive numbers, roots, last digits, odd/even, what I call 4 by distribution, 5 by distribution, columns, postional ranges...etc. The only value I see in doing this is not to make a dumb low percentage bet. There are obvious flags of higher probablies in the resultant data, but to narrow the combinations based on this and the associated error rates that come with them seem premature.
The second set of what I call dynamic filters reflect the current status of the individual lotto numbers. Things like how many times they've been picked over a history range, most occuring next pick numbers, pairings...etc. Although not as cut and dried as the static filters, these filters seem to give more impressive results.
The dilema is in the strategy of filters. The mix of how many high pass (98%+) filters combined with low pass filters(lower %) to achieve an acceptable amount of combinations to bet on. Although the LP addesses alot of the filters, nobody seems to really address overall filter strategies in any detail except on individual filters, one by one.
With a new program in the works, the debate continues. I guess we'll decide after the high pass filter program construct is completed and we'll see the average amount of combinations there are left to deal with.