Implement headstart search. Try
out your player on a standard game of your choosing and demonstrate the value
of the strategy once the game starts. Find a game for which the strategy is
not so good. In general, when is headstart search helpful and when is it
Construct an end-game book. Again find a game
for which this works. Discuss when this technique is effective and when it
does not work so well.
Implement a player that uses your Monte
Carlo program to evaluate various heuristics during the startclock and then
uses a linear evaluation function during game play with weights determined by
the results of your Monte Carlo results. Discuss your results.
Think of a way to improve your Monte Carlo player and implement it.
Your improvement does not have to be particularly complicated. Run your
program on at least two games. Describe your motivation and your approach. Do
you get consistent results on a single game with additional time?
Extra Credit will be given to novel metagaming techniques that go
significantly beyond what is required for the preceding exercises.
Make sure your improved player is ready to compete. We will test your
player in timed mode using a three minute start clock. It's okay to
lose, points off if your player fails to play (unless it is our fault).