Enter Your Reply

Humans are strategic players, computers are tactical players. Computers can
follow a few pieces in long sequences, humans are better at evaluation
whole board situations. Thus a game where strategy counts and that the
evaluation function quickly engulfs the whole board, would be the hardest
for computers. Another thing would be is that if the differences in the
relative worth of the positions are subtle, and that the resolution is
sufficiently in the end game that the program is forced to evaluate more
positions within any one ply. 

However there is another approach, that is to have a game of sufficiently
high complexity class and sufficiently scalable, then one just need to
increase the size of the board to keep it out of the computer's reach,
especially if the time complexity is beyond exponential, or space
complexity is beyond polynomial. In this kind of game one can think of
leveling as computer and humans battle at increasingly higher levels.