As I spent many years in spreadsheets trying countless systems to find a profitable best guess, I realized some were too complex to test in the spreadsheet because I hated the VBA macro language, so I moved to the Python language. While setting up systems was a straightforward task, back testing was always a bit trickier, but possible. When I started being able to run complete back tests across entire draw histories is when I saw just how systems fail on profitability in the long term. I could back test how effective mirrors or Vtracs were right in a spreadsheet, but jackpot games added complexity. Once I started back testing what I considered the better systems against the histories I saw the reality that multi year gaps and inconsistent results were common to ALL systems.

To be fair, I know that some use reduction systems and I was always searching for the one shot best guess single line to play. I never found it, at least not repeatably or profitably. I can say that upskilling to full back tests took the fun out of the hobby. Applying rigorous statistical testing ended up proving to myself that system design for random processes is a fool's errand.... one that kept this fool busy for over 20 years! None of that time was wasted however as my coding/problem solving skills definitely improved. I was usually under the impression that I was just not asking the right questions, then came the realization that there were no questions that would provide the answers I was looking for.

This hobby was a super domain to work with noisy meaningless data. Now that my new hobby is Bioinformatics, I have seen how working with noisy data with hidden structures is possible. My most focused attempt at lottery was using Markov chains on follower distributions. This does not work on the lottery, but I was able to apply Markov chains directly to solve a few problems in finding hidden patterns in DNA that had shifted by mutation. IF the lottery data was predictable, Markov chains were indeed the right tool. But, because lottery draws are truly memoryless, it did not work.

Back testing squashed every fantasy I had about "figuring it out" or "beating the system".