@PinChusion below........... im impressed brother thats some good shit you just brought to the table and makes good sense when looked at in a logical manner..... thing is though is that i can only speak for myself with sust, some guys bloat like water buffalo... maybe its because my AI approach is totally different or the diet tactics maybe! it would certainly make for a good mass experiment to see who does well on sust and who bloats bad........ interesting stuff indeed and its got me thinking about creating a mass experiment within the community but i dont have the time i used too have for stuff like this....... maybe someone else would be interested in starting this thing off........ CryHavoc comes to mind as a guy who thrives on details, maybe a 10 guy experiment would prove to show a varied result range.
This solution impliments a recursive, depth-first search of the possible values unfilled sudoku cells can take. The search tree is pruned using logical deduction rules and takes about a minute to solve some of the more difficult puzzles. This code can be cleaned by making the main code blocks, denoted by "%% [Block Title]," into their own separate functions. This can also be further improved by implementing a Sudoku class and making this solver a member function. There are also several lines of code that can be vectorized to improve efficiency, but at the expense of readability.