While I see a few things about this technique that are ignored and not included in the "findings", and I believe significantly better search algorithms already exist, I'm most curious why you chose THREE, specifically, as the number of "characters" to check?  Is there something about THREE that makes it better than TWO or FOUR (or any other number)?  Have you tried your simulations with other prefix sizes to find the optimal prefix size?

Also, it seems very inefficient to take a hash (which is a large binary number), and first convert it to a string representation of its value as represented in hexadecimal, and then use string manipulation to compare "characters"?  Wouldn't it be much faster and simpler to perform a bitwise XOR on some predetermined prefix of bits of the hash itself (or even on the entire hash itself)?