Hi y'all! Wanted to share some test results of a Pick 3 system I've been working on. Feel free to digest and pick it apart, any and all feedback is appreciated! The system relies primarily on "repeats from the previous draw", with a few twists and turns thrown in to (presumably) improve performance.
Most of us know that about 50% of the time the next drawing will contain exactly 1 digit repeat from the previous draw. This is true of the next draw compared to ANY previous, not only the most recent draw. So I asked myself this: what if I could set-up some filters that capitalize on the above concept, to narrow down the field of numbers for the next draw to something manageable? Here's what I came up with:
I've set-up a spreadsheet with three filters that I call CWP, CWX, and OUT, and these filters are intended to capture the "footprint" of the next draw.
- CWP (short for "Common With Previous") is the number of digits the next draw has in common with the most recent draw. It ranges from 0 to 2.
- CWPX is the same concept, but the number you're comparing to is different: it is the number of digits the next draw has in common with the most recent Pick 4 number (for example: if you're playing the Pick 3 Evening you'd be comparing to Pick 4 Midday, or if you're playing Pick 3 Midday you'd be comparing to the previous day's Pick 4 Evening). It ranges from 0 to 3.
- OUT is again the same concept, but you're comparing to a list of numbers that have not shown up, in any position, in 3 drawings. Why 3? I use 3 drawings to "tune" the median size of the OUT list to 5 numbers, but there's no reason a smaller (or bigger) window can be used, I haven't tested that aspect yet. This filter ranges from 1 to 3.
How do I use these filters? Simple. If, for example, I set CWP = 1, CWX = 0, and OUT = 2, I'm saying that out of the 1,000 possible numbers for the next drawing, only the following numbers are "viable": numbers that have exactly 1 digit in common with the previous Pick 3 draw (CWP), exactly 0 digits in the common with the most recent Pick 4 draw (CWX), and exactly 2 numbers in common with the "haven't shown up in 3 draws" list (OUT).
To determine the initial unfiltered pool of possible numbers for the net drawing I would use my "best guess" of the Root Sums that might come out in the next drawing, and then start putzing around with the filters, setting each at a broad range and gradually tightening them to achieve, as in the example above, a specific number. I would, for example, start CWP at 0 to 1, CWX at 0 to 2, and OUT at 1 to 3, and then toggle each filter individually to look for "indirect evidence" of which filter setting might be the correct one. In this example, my first pass would entail setting OUT to 1 to 2, seeing how many combinations were viable, then setting it 2 to 3, seeing how many combinations are viable, and make the leap that the range that yielded the highest number of viable combinations is likely the "correct setting". I would then move on to CWPX and do the same. Then back to OUT to do the same type analysis and reduce the range to one number. And back to CWPX to also reduce to one number. Now that CWPX and OUT are set to one number each, by the same method I would determine the "optimal" setting for CWP.
It's not as arduous as it sounds, but there are two problems: 1) I found that the order of tightening affects the final answer, and 2) the "accuracy" of this method is adversely affected by the more Root Sums are used. When I try this with 7-ish Root Sums it seems to add "noise" that hides the correct answer, whereas only 2 to 3 Root Sums seems to work much better.
So I gave up on the "gradual filter tightening" and have taken a different tact; right now I am running a macro that determines the viable combinations for EVERY possible filter combination, and plan on looking for things like what filter combinations most frequently reveal the wining number, and does that state of the number preceding the wining number (as measured by Sums, Short Sums, Root Sums, Odd/Even structure, Lo/Hi Structure, Singe or Double or Triple) give any clue as to what the next winning filter combination might be.
That's all well and good, but there's still the issue of "garbage in / garbage out". Unless I can figure out how to improve my
determination of possible Root Sums for the next drawings, all is for naught . . .