Author: Miguel A. Ballicora
Date: 05:44:19 08/21/01
Go up one level in this thread
On August 21, 2001 at 06:27:11, Peter Berger wrote: >On August 20, 2001 at 19:34:10, Federico Corigliano wrote: > >>Hello, I am called Federico, I am from of Argentina and wanted to know Gaviota >>in this game did not waste a victory. I played a game 5 minutes without book of >>openings for WildCat and 8 hashtables of Mb, and in last the thirty moves, >>Gaviota had an advantage of 6 to 8 pawns according to the evaluations of the >>programs and of Crafty. But Gaviota instead of attacking with the king, backed >>down. It had enough time like gaining the game, but it wasted it. Perhaps it is >>an error mine, I am very inexperienced in the chess. In any case they are >>amiable to respond. Thanks. Hola Federico! Voy a escribir en Ingles para que todos entiendan, si te es mas comodo, pasamos al castellano or escribo en los dos idiomas. No hay problema. Gracias por probar a Gaviota!! You are right, Gaviota should have played Ke4 whenever it could. That is a very "human" move. However, gaviota preferred not to put the king in front of the central pawns that were going to be advanced anyway. GAviota had to wait to avoid the checks and advance a pawn. This is not a terrible problem for a computer, they sometimes chose ways to win that scare humans. The problem here is that it lost on time. I can try to improve this situation in the future, maybe speeding things up when there is a big advantage (so far Gaviota does not take into account the position to regulate the time, maybe it should). However, it is very limited what it can be done when games are played without and increment. Sooner or later, one program must lose on time if the game is too long. It should not happen at move 86 with a huge advantage though. Do you still have the log.txt file? it is overwritten so probably you don't. I you still have it, please send it to me. >Hi Federico, > >after a brief look at the game I think it was a convincing game by Gaviota until >its flag fell - it transformed its advantage into a clearly won endgame. Hi Peter, Yes, time management when there is no increment is very basic in Gaviota. As Dann suggested in the other post, many weird things can happen in such short games. Anyway, I am quite happy to see this game. It handle the opening quite well. The manouver Qb6-Qc7 made white to play Qe2 leaving the Bishop in f1 with development problems. Granted, Wildcat played without book in this game, but Gaviota does not have a book either. Gaviota has a basic learning feature that allow it to avoid previous disasters. After a while, it seems that it prefer the french opening and handles that reasonably well (considering that gaviota is not a strong engine). Maybe when I release a book I should take this into account and include the french. People say that french is generally not suitable for a computer... Well, I like the french and maybe Gaviota is copying his dad :-) Thanks everybody for your comments, Miguel > >It seems the time management by Gaviota isn't perfect yet. > >Maybe you could contact Miguel A. Ballicora for more information. > >pete
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.