well i decided to go 15 rounds against the new homebrew nes 15. with less than encouraging results
i used save states soboth the AI and I got the same puzzles
here is how well i did
*=winning score
game 1   AI-92*  Me-102
game 2   AI-141* Me-161
game 3   AI-119* Me-125
game 4   AI-108* Me-122
game 5   AI-132* Me-162
game 6   AI-119  Me-107*
game 7   AI-141* Me-187
game 8   AI-149* Me-181
game 9   AI-141* Me-275
game 10  AI-145* Me-197
game 11  AI-92*  Me-154
game 12  AI-158* Me-236
game 13  AI-145  Me-133*
game 14  AI-132* Me-136
game 15  AI-120* Me-172

all in all a very solid AI. and i don't consider 2/15 too terrible 

Nice comparison ! I'm assuming the AI is based on an algorithm, but should be beatable by someone with a bit of experience. It would be cool to implement an optimized algorithm, but probably a bit complex for the NES.

Brad - I haven't looked at the game you guys are discussing, yet, but sometimes the optimal algorithm is actually the easiest thing to implement (Nim, for instance).

game 1 AI-103* Me-175
game 2 AI-130* Me-132
game 3 AI-99* Me-131
game 4 AI-129= Me-129=
game 5 AI-134* Me-144
game 6 AI-157* Me-185
game 7 AI-124 Me-104*
game 8 AI-133 Me-69*
game 9 AI-144* Me-180
game 10 AI-131* Me-183
game 11 AI-157 Me-109*
game 12 AI-155= Me-155=
game 13 AI-127* Me-155
game 14 AI-133* Me-135
game 15 AI-104* Me-150