Due date: Friday, April 6, by 11:59pm

This is the second of three projects to build an online chess game. This project focuses on building a computer opponent, named Shallow Blue (not to be confused with Deep Blue). Shallow Blue is an artificially intelligent chess player that knows the rules of chess and should play a reasonable game. The last and final project, project 3, will develop the online component (Networking) for the Chess Board and Shallow Blue.

Note: As for the first project, this project is to be done individually! You are not to work in groups of any kind. You are encouraged to talk about solutions with your class mates and can even help each other debug code. However, cutting and pasting and mailing code to each other is not allowed.

You are encouraged to build upon your first project, the Chess Board. If your first project really did not go well, contact the instructor for how to proceed.

Description

You are to make a computer-controlled chess opponent, called Shallow Blue, for your chess game. Shallow Blue and the player should play a game using the Chess Board developed for project 1.

The game should support:

• Legal chess moves (as in Project 1) taken in turn, by Blue.
• "Smart" moves by Blue. In other words, Blue should try to win and not just move pieces randomly around the board.
• Levels of difficulty. Blue should have settings, making it an easier or harder opponent. You are free to implement the levels of difficulty as you'd like (see the Notes below for suggestions). The GUI front end should be modified to allow the user to change the level of difficulty.
• Advanced:
  • Tunable "think" time. Blue does not always need to spend the same amount of time per move, but there should be checks in place to make it so it doesn't inadvertently think too long.
  • Sophisticated Evaluation function. Basic count of material (ie- pieces) works really well, but there are a lot of additional enhancements that might be added.
  • Opening vs. Mid-Game vs. End-Game. Evaluation of opening moves or end-game state may proceed differently than mid-game state.
  • Personality. Different versions of Blue (selected randomly, is fine) may be more aggressive or defensive or ...

You will use a MinMax search algorithm produce a move. The algorithm will have AlphaBeta pruning and a board evaluation. See the lecture notes from class for more details.

To implement a "dumber" Blue, consider:

• Not searching as deep.
• Not having as sophisticated a board evaluation.
• Adding randomness to the choice (ie- the best choice is not always taken).

If there are choices that have equal value (for example, in an opening), you should "roll a die" to see which choice Blue makes.

For different personalities, consider different board evaluation strategies (ie- assigning more weights for an exchange in pieces in some cases, and less in others).

For opening and end-game, consider having separate phases and, again, the board evaluation may be different in these cases.

As for the first project, you are free to implement your game in your choice of language (C, C++, Java, ...). You should use a language you are familiar with as you don't want to spend too much of your development effort fighting with the language.

Links

There are lots of links on MinMax trees, some specifically for chess. Here are a few:

  /cs/bin/turnin submit <course> <assignment> <file1> <file2> ...

  /cs/bin/turnin submit imgd4000 project2 proj2.zip

  /cs/bin/turnin verify imgd4000 project2