Author: Gian-Carlo Pascutto
Date: 04:26:03 06/03/02
Go up one level in this thread
On June 03, 2002 at 04:28:46, Tony Werten wrote: >On June 01, 2002 at 16:14:51, Gian-Carlo Pascutto wrote: > >>[D]8/5r1k/7p/4K3/6Q1/8/8/8 w - - 0 78 >> >>Qe4+ is a tablebase-mate in 31 and is seen at ply 2 in 0 secs >> >>According to the game record Sjeng played 78. Ke6 here >> >>Hmmm. This is no good. > >There were some very unfortunate coincidences here. > >The best improvements I made lately in XiniX were taking out a lot of bugs. >Unfortunately in the first 4 rounds I used my last compile wich was a debug >version. This version is about 2 times slower than the release version, but >because I just got my new computer wich is 6 times faster I didn't notice. > >A side-effect of my debugversion is that it basicly disables my egtb's, >returning always a draw. > >So in this position XiniX said draw, Sjeng said white wins. With 2 minutes on >the clock, Sjeng allocated 20 secs for his move, wich lead me to believe that >Sjengs score was an evaluation score so XiniX' score was correct. > >I was waiting for my opponent to offer the draw, wich I thought ( because of the >time difference ) I did not have to offer. > >He didn't and lost on time. I have some more information as well, and I can make sense of it now. Vincent set up Sjengs machine with the book and exe I had supplied according to my instructions. However, he noticed there was no sjeng.rc file (which contains config info like directory of the EGTB and hash size), and assumed he had to make one. Because Sjeng was running under UCI, it was not needed, which is why I didn't include it. A 'very unfortunate coincidence' was that because of the presence of the sjeng.rc file, Sjeng initialised it's EGTB's on startup. The combination of CPU/HD/OS on the tournament machine did so just a tad too slow, causing the engine to time out when being loaded in the Shredder 6 interface. In desperation, the EGTB's were disabled, so the engine could be loaded in the interface. The book problems were caused by a similar issue. Vincent assumed the book was a book for Sjeng, but it was really a book for the Shredder interface. For some reason we haven't figured out yet, the first moves were played instantly during testing, causing us to believe all was ok. I'm not sure what caused the loss on time. At a guess, the operator was not very familiar with the Shredder interface and didn't know how to change the clock (not possible via the menus) or missed the fact that the window to do so shows used and not remaining time. There were bugs in Sjeng which caused problems during pondering as well. -- GCP
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.