Author: Vincent Diepeveen
Date: 04:40:00 03/05/04
Go up one level in this thread
On March 05, 2004 at 04:41:22, Odd Gunnar Malin wrote: The most important 3 issues are: a) you need to pay them to get garantuees *when* a new program gets on the list b) not all games get published. in fact CURRENTLY, so in 2004, only 1 tester i can find who publishes all games he plays. What does that say about the other testers? That means that they produce a rating which you cannot verify to be true. c) All the engines competing in SSDF use interfaces and engine tricks that are just too dirty to be true. For example when loading UCI engines playing fritz8, then it will after the first game only get 1MB hashtable for games 2 till the end of the match. Such tricks are too dirty simply. To avoid all these tricks you are working for years. So you are not busy with computerchess then. SSDF should not allow such tricks, which is of course hard when you get paid for that... Right now we cannot verify what happens in the games. Only 1 tester is transparant. >On March 05, 2004 at 04:26:25, Dan Andersson wrote: > >> You are not taking that post seriously are you? >> >>MvH Dan Andersson > >At least the book issue is something. > >Before the Leiden tournament there was some posting here that suggested that a >book author did test hard against opponents books. Nothing wrong with this but >later we see that SSDF use the same book in their testing. > >Another issue is the book learning, this make the list unuseful to check which >engine to bye for human games or analysis. Eg. the importantness of the list is >only for those who like to play comp-comp games. I don't have thought too much >of alternative test forms here but tournaments instead of maches could maybe be >something. > >I will not comment on the other issue in Diepeveen's post since I don't know >anything on SSDF's testers and their ability. > >Odd Gunnar
This page took 0.01 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.