Done 2 years after version 0.1a, this is a higly anticipated release. Using highly refined techniques, this release promised to do even better than its successor. Castling, History Tracker and Repeat Moves detectors are all possible in this, making the AI one of the most unpredictable that ever was programmed.
1) Features up to 3 kinds of AI: Alpha-Beta, Min-Max and NegaScout.
2) Move Logging and Node Calculation different: Now nodes are calculated only for base nodes only. Meaning the last 'end nodes" will be added. Any intermediary nodes are not counted.
Download and try version 0.2 today!
14 Mar 2007: Version 0.1b is OUT!
Fixed a number of issues, mainly:
1) Memory leaks: there were huge leaks in 0.1a that causes users to crash in the middle of the game (about Move #15 or so) that was really really bad. Used Visual Memory Leak Detector to catch the nasty little bugs and fixed it there and then.
2) Moveset logging and AI statistics logging. Now we can tell how long it takes and how many nodes computed by the AI. Not too meaningful for now but still its a nice feature for AI engineers to know. Or players to know.... "Hey i beat the computer who calculated 132432434 nodes per move. And I was AHEAD of him! Haha.
3) Finally you can pit computer to computer. Unfortunately you can with the version i release. Maybe I'll make a menu for users to choose instead of Pre-processor directives. Haha. Well to make it out short, depth of thought DO make a difference. Often a higher depth thinking computer will beat a lower computer easily. However, when in equal par, the computers arrive in a deadlock situation, doing silly things like moving a single piece back and forward causing the whole game to NOT progress in terms game play. I'll try to resolve the deadlock situation soon either by : introducing better evaluation algorithms (taking positioning not just material weights into consideration) and maybe even some randomness which I'm quite reluctant to do so. Frankly I don't think randomness will help AI do smart choices. But sometimes LUCK does play a part, doesn't it?
Scheduled to be done
A. New AI - Alpha Beta, and Nega Scout and maybe an in-house developed prototype AI
B. Menu system
C. Playback of past games, for players to review a match.
Get the 0.1b here
Pegasus Chess Alpha Version 0.1a
Computes up to 5 levels of thought, with decreasing computations as it goes deeper. Next Milestone will be to optimize the generation of leaf nodes and allow faster computations.
Also set to release:
1) Text Logger to save and record matches for easy debug (since its deterministic machine for now)
2) Sounds to help notify when it has finished computations
3) Allow 2 AI to pit against each other. i.e Minimax VS Alpha-Beta
Get Alpha Version 0.1a here
