Author: Uri Blass
Date: 05:25:02 05/07/03
Go up one level in this thread
On May 07, 2003 at 07:08:16, Kurt Utzinger wrote: >On May 07, 2003 at 06:52:42, Uri Blass wrote: > >>On May 07, 2003 at 06:37:22, Kurt Utzinger wrote: >> >>>On May 06, 2003 at 19:54:18, Rex wrote: >>> >>>>On May 06, 2003 at 16:09:01, Jorge Pichard wrote: >>>> >>>>>>Indeed the engine is with CB for a while now. CB had a technical delay due to >>>>>>Easter - Junior8 will be released in about 10 days >>>>>> >>>>>>Shay >>>>> >>>>>What are the Beta testers saying in comparison to Shredder7 ? >>>>> >>>>>Jorge >>>> >>>>It will be no different than any other program release. It comes out, people >>>>get it, and gets old after 3 days then wonders when the next release will come >>>>out. >>> >>> Hi Rex >>> Just great comment about the computer chess scene. It comes out, people get >>> it, play some stupid blitz games and after 3 days up to a week are crying >>> for the next version. And as soon as we have 5000 Mhz PC's people are >>> playing games at level 0s+1s ... Crazy chesscomputer world. I will never >>> change my mind: quality is more important than quantity. >>> Kurt >> >>Maybe if you start to develop your chess program you will change your mind. >>Bullet games are important for detecting bugs that are not relevant only for >>bullet. >> >>If after changing your evaluation you see that the new version fail to win in >>bullet a simple win that the >>old version had no problem to win then you may know that you probably have a bug >>in your evaluation. >> >>If you play only long time control games you will have the following problems: >> >>1)You will need more time to detect bugs. >>2)you may not discover part of the evaluation bugs because search is going to >>hide them and you will not see the program missing simple win and the mistakes >>of the program may be more complex so you may miss them. >> >>Uri > > > Hi Uri > I would like to make a difference between programmers/beta testers and > the normal chess community if an engine has been released for public. On > the other hand I can tell you from my own experience as beta tester of > program X: it often happens that it is easier to find out [real] bugs with > games played at longer time controls: the behaviour of some engines is not > the same at different time controls. > Kurt I think that it is dependent on the bug I can describe 2 cases when watching bullet games(40 second per game) helped me to detect exsitence of bugs(I had to investigate more to find the exact bug but the main problem with bugs is not to fix them but to know about them). 1)The program failed to win KQ vs K endgame. It has only few seconds to finish the game but I know that usually it win this endgames even with 2 seconds to finish the game. I found that the reason was a bug that distort the evaluation in part of the cases including KQ vs K positions. 2)The program that played against previous version showed significantly different numbers in it's evaluation. 40 seconds per game did not prevent me to see that the old version has evaluation that was bigger by a pawn than the new version for many moves and I knew that I did only a small evaluation change. Again the reason was a bug. I agree that there may be bugs that happen only at long time control and I do not say that games at long time control are unimportant. Uri
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.