CEGT 40/120 Quad testing started

[Heinz Van Kempen]

Hi Ray ,

tests are on Core 2 Quad E6600 (5 men) - 4 GB RAM.

To have consistency between all our lists we adapt the time controls to AMD x64 4200+, what comes down to 40 moves in 80 plus 20 moves in 40, plus 20 minutes for the rest of the game, a bit more than I would have to give according to benchmark. Of course I could overclock but I think this is a good time control, while I still think to give some matches with even longer time controls when the lists are advanced.

Thus time controls are 10 times those of CEGT 40/20. In the same time we run one game 40/120 we could also give 10 for CEGT 40/20.

[Heinz Van Kempen]

Hi, it sound great!!
It wolud be nice if you test each engine with own book to , since the Rybka-zappa match I'm still wonder which's the real diference between both engines. so if we take both test ,under generic book and own book we can bring some conclusions, Deep Junior wolud be stronger in own book than generic book, don´t you think so?. Thanks to all CEGT team
Regards

Hi Armando ,

the question if own books or shorter generic books had been discussed a few hundred times in all fora.

I repeat that we do not test with own books, because we want to test the engine itself and not the quality of the opening book. It is also awful to see when after around 30 moves from own books each side has only one rook, bishops of opposite colors and a few pawns left. Hard to show than that an engine is superior. So a shorter generic books with for example 8 plies for each side with resulting good positions apted for engines fit our needs.

In general I think that a good book for a "weaker" engine can make it similar in strength to a better one with bad book. The ELO gain can even be enormous. But we want the pure engine strength like CCRL also, while others think that an own book is part of the engine. So better not repeat the discussion again. There is no compromise.

[Spock]

Hi Ray ,

tests are on Core 2 Quad E6600 (5 men) - 4 GB RAM.

To have consistency between all our lists we adapt the time controls to AMD x64 4200+, what comes down to 40 moves in 80 plus 20 moves in 40, plus 20 minutes for the rest of the game, a bit more than I would have to give according to benchmark. Of course I could overclock but I think this is a good time control, while I still think to give some matches with even longer time controls when the lists are advanced.

Thus time controls are 10 times those of CEGT 40/20. In the same time we run one game 40/120 we could also give 10 for CEGT 40/20.

Excellent, thanks ! Good luck with the testing