Well, I finally worked out the last bugs in my dumb AI separated into a module, so I am now starting testing. I am now a little more than an hour into what I estimate will be a six-hour test. I am picking 1000 random words from the dictionary and having the dumb AI play each one, with the word and number of tries being recorded in a log file. I'll do the same thing with the smart AI next. Then I'll compare the lists. Yes, I probably should do a complete test, finding averages for each word in my dictionary for both AIs, but that would take a long time, considering there are almost 6000 words in my dictionary. Of course, I could use the pared-down dictionary, but we're still talking 3000 words, several times each. And the dumb AI doesn't take very long. The smart AI is probably going to take more than a day to run the same program, maybe several days. Maybe I should do its first trials with very low threshholds, so it doesn't turn on until the dictionary size is down to 200. That won't take much longer than the dumb AI. So, yeah, we'll see.
So, I plan on taking the numbers and finding the average and standard deviation of the number of tries for each AI. Maybe use this data to figure out graphics programming and put out some nice-looking graphs.