Author: Dan Homan
Date: 06:24:08 02/02/00
Go up one level in this thread
On February 02, 2000 at 07:23:47, Steffen Jakob wrote: >To all authors of programs which play the ICC tournament. I would like to know >if you make changes between the rounds. I guess that most amateurs are doing so, >but the commercial ones don't. I talked to Chrilly Donninger and his opinion is >that a main difference between the commercial programs and the amateurs is the >stability, which I think is true. E.g. I had a stupid permanent brain bug which >made Hossa blunder against Nimzo in the first round in a position which was >probably won: > >[D]6R1/7p/4ppk1/p2p4/3P1B1p/2n1P3/q5PK/1R6 b > >The result of the bug was that Hossa played almost immediately with a short >search, and the move was Kf7. If Hossa would have thought one second about this >position it would have played Kf5. The problem was that I saw this bug two times >in the game against Nimzo. The first time it occured he played a good move but >the second time he ruined his position (and I don't think that I wont get such a >chance very soon again). > >This bug was very ugly, because playing 2 or 3 moves with a very short search in >a standard game is almost deadly. So I implemented a very quick and dirty >workaround even before round 2 (there really was not much time) and a much >better workaround for round 3 + 4. > >What about others? Do you avoid any changes or did you also find some serious >bugs which you try to fix between the rounds. > >Best wishes, >Steffen. In round 2 I lost a rook in a bad line from my book, I put in some code that quickly blunderchecks book moves. There is a "gambit threshold" which I currently have set to 0.5 pawns. I made these adjustments following round 2 - it was scary though, because I didn't know if my quick blunder check search could introduce weird bugs. For example, it just occured to me yesterday that the blunder search might screw up my pondering or hash tables - I don't think this is a problem, but the situation is complicated enough that I haven't convinced myself yet. I've noticed other things about my program's play, and it is tempting to try to 'fix' them, but I think only the book problem was important enough to risk introducing a bug. - Dan
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.