Author: David Dory
Date: 02:14:26 03/29/02
Go up one level in this thread
On March 28, 2002 at 01:21:56, Jouni Uski wrote: >On March 27, 2002 at 09:27:51, David Dory wrote: > >>On March 27, 2002 at 05:28:28, Jouni Uski wrote: >> >>>[D]8/B5pp/4k3/P7/6n1/8/P6P/7K b - - 0 41 >>> >>>Crafty as black played Nxh2!!(??). Yes the game ended draw so the move is >>>probably correct, but I doubt, that Crafty's evaluation really understand this. >>>E.g. if white plays a6 then black has no way to a8 square... >> >>I don't see that. Black king is INSIDE the "square" of the a5 pawn after you add >>one more move because the White bishop has to get out of the way of the pawn. >> >>It may LOOK crazy, but just count the squares and moves up! >>What does Crafty 17.14 like as a move for Black in this position?? Does it do >>any better? >> >>If Nxh2 was such a bad move, Yace would have immediately jumped on a6 and won!! >>But it didn't, so although it looks bad perhaps, to humans, it's a good move. >>With White pawns on both edges of the board, the Bishop would certainly be >>favored to win against a Knight. >> >>I haven't seen any matches where 17.14 was better than 18.14, have you? >> >>Since you have them both, why don't you run a match with standard time controls >>and see how each one does? Be fair about hash table size and book! >> >>>Crafty 18.14 is clever endgame player, but still I am disappointed, when >version 17.14 solves 4 positions more in my 40 position endgame test and also >17.14 is on average about 30% faster in all postions! >> >>Test suites are crap for comparing engine to engine strength OTB. IMHO. >>Very crude indicators, only. >> >>I think your enthusiasm is great, but knowledge is just not something you can >>readily see and measure easily in a chess program. Not like NPS. >> >>My belief is forget the NPS, forget the test suites. >> >>Play a match or three, VERY FAIRLY! ** Then you'll know **. >> >>I'll be surprised if 18.14 is worse than 17.14. >> >>Dave > >I played also 40 games match and 17.14 won 18.14 by 21,5-18,5... > >Jouni So out of 40 games, in one match, the score was almost equal. Sorry, Jouni, that doesn't sound like 17.14 is stronger. That sounds like they're equal and your data can barely separate the strength of the two. More games or a greater difference would be required for a real validity. What was the time control of the match, and other conditions? Hash table sizes, etc. Of course, blitz (faster time controls in general), will nearly always favor a faster engine, like 17.14. From Dr. Hyatt's posts, it's very clear that he is trying to add more knowledge to Crafty right now. Optimizations to further increase 18.14's speed will undoubtedly be looked at in upcoming versions. As you know, Crafty is a test bed and Dr. Hyatt is not someone to stop testing out idea's on Crafty. It's good that you did the match. I just disagree with your opinion of the match, relative to Crafty's strength of 17.14 versus 18.14. It doesn't mean I'm right. Just my opinion. Dave
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.