The "knowledge base" will consist of a group of saved queries, written so that they can be applied to any game simply by changing a few parameters.

Want to dive deeper for the conditions surrounding a drawn number, turning that into a query and running it against each number in each position...

First up will be a second order Markov exploration... looking at the numbers preceding the digit of interest as well as the numbers following it.
Rather than command line queries I want to explore the use of a program called LextEdit, which is used for creating and editing SQL queries.

The spreadsheets are too much at once, anyone looking at the raw data would easily miss a slight bias condition that happens every 500 draws or so.

Python scripts process data based on what you tell it to look for, but what if you do not know exactly what to look for?

SQL will hopefully bridge that gap, helping to identify actual trends (even if just coincidental) that can then be coded into python. Also looking to see if different columns follow certain trends at different times, wide or narrow bands of frequency that could indicate different data regimes where maybe, if identified, can be used to finally solve the column synchronization issues.