Thursday, October 25, 2007
Some of you may recall this program was originally named "Variant Shredder." Stefan Meyer-Kahlen, the author of the chess program named Shredder, contacted Ed Trice and the programmer of "Variant Shredder" and requested another name be selected for this Gothic Chess program. The new name is "Variant Pulverizer", a much more graphic depiction of destroying an opponent!
SMIRF vs. Pulverizer
As stated before, I asked the programmers to change their code before sending me their programs. That way, nothing can be "learned" about them prior to the Gothic Chess Computer World Championship. The purpose of the tests was to see if they can run on the hardware that they will each be playing on during the event.
Pulverizer was altered radically, and the play in this game shows some very anti-positional play (reverse developing pieces with moves such as 6...Ai8, unorthodox moves that you rarely see such as placing the Queen where a Bishop could take it with 16...Qb8 to precipitate an exchange of queens). In the end, Pulverizer lived up to his name, and won the game.
This was an unusual game between an updated version of SMIRF and a Gothic Chess newcomer named Tornado. In speaking with the programmer of Tornado, he told me he named his program after the name of the horse owned by the legendary Spanish hero, Zorro.
SMIRF vs. Tornado
The purpose of these "warmup" games is to guarantee that the programs are able to run on the hardware being supplied for the Gothic Chess Computer World Championships. I encouraged the programmers to modify their evaluation functions to make them "weaker" than they are in reality, so that no information can be "data mined" before the start of the tournament. The Tornado programmer complied, and his changes made the program play some odd-looking moves. It was enough to overcome Germany's SMIRF program in this game. The link above will allow you to replay it one move at a time.
Thursday, October 18, 2007
Here is a link to the game (with SMIRF 1.66g score and comments)
Martillo vs. SMIRF
The Martillo programmer has had problems getting his software to run on the hardware that will be used for the upcoming 2007 Gothic Chess Computer World Championship. Finally, this evening, he was able to get it to run. He requested a casual game to be played against SMIRF, which I did do.
The time controls were G/30 + 15 seconds per move.
I will post the complete game here tomorrow for everyone to review.
My comments on the game:
Both programs showed similar evaluations and Principal Variations for much of the opening. Castling did not appear in the P.V. for quite some time as each program moved their pieces around the board. Both were reaching roughly the same search depths, with neither really outsearching the other.
There was some odd form of "Archbishop chasing Queen" where I thought some of the moves were unusual. Martillo finished a pawn ahead as a result of the run-around. As each program moved progressively faster, it seems that Martillo was able to search 1- and 2-ply deeper, until its score grew. SMIRF's score fluctuated +/- 0.4 pawns through all of this, thinking it was +0.4 one moment, down to +0.2, then to -0.1, back up to +0.1, and back and forth, finally down -0.4
Meanwhile Martillo showed about +2.3 or better, growing to +4.3 at one point. Yet, the score never increased for Martillo, and never got worse for SMIRF. Move after move, +4 for Martillo, about -0.8 for SMIRF at that point.
Finally, towards the end, Martillo picked up the material and checkmated on move 63.