Tried something different with the filtering. Normally, I'd filter it according to the respective trend. The MOST filter gets deflated against the most likely trend. The LEAST filter gets deflated against the least likely trend.
However, I wanted to develop a better way to find cold/due numbers. By inverting the filter, I should be able to see that. For example, running the MOST numbers against the least likely filter can show a weak number because even though it manifested itself in the MOST group, it's matching a dead trend (which is different than a number simply not matching the most likely trend). Running the LEAST numbers against the most likely filter can show numbers which are due because they haven't been drawn in a while, yet they match the most likely trend.
I'll be interested to see if it works.
I played with some key filters last night, but I'm not convinced it's real math. I think I'm going to calculate the key filters AFTER the drawing results to try and reverse engineer the method. If it starts matching draws, then I'll apply it towards a future drawing and see how it fairs. It's very time consuming.