Author: Robert Hyatt
Date: 09:00:48 10/11/02
Go up one level in this thread
On October 11, 2002 at 07:51:42, Vincent Diepeveen wrote: >On October 11, 2002 at 07:43:40, Uri Blass wrote: > >But he's an old idiot. He's still talking about computerchess, >forgetting how bigtime the weakest chain has gone up last few >years. In 1997 games were decided by programs blundering away >material. All programs were very weak in endgame by then too, >any random rook endgame i could win from any program in 1997. Absolutely bull. I have a program that would be _happy_ to play you in a "random rook ending". I actually lifted my KRP vs KR evaluation code (in Crafty) directly from Cray Blitz. Other programs were even better as they used endgame tables in addition (we did not). HiTech comes to mind and I personally saw both HiTech _and_ Deep Thought outplay GMs in endgames on ocacsion. Your hyperbole is actually worse than that coming from the ChessBase crowd prior to the Kramnik match. > >How things have changed there... > >Even theoretical you can proof that without nullmove and >without 'junior' type of forward pruning you can't get >18 ply fullwidth at all. > No you can't, and your statement is totally ridiculous. There are _plenty_ of non-null-move ways to forward prune. Go back to the early 1970's and read about those programs. None of us were doing full-width stuff, and none of us were using null-move as it is used today. >Knuth in fact proved a lot about that already. It's so >easy. Only because this guy has 'professor' before his >name doesn't mean that he can do something that theoretical >is impossible. > And what must we say about someone that declares that anything that _he_ can't do is completely impossible to do??? >Apart from that the statements from the deep blue team >are very clear in 1999 in the IEEE advances where they >show a 12 ply search. The 12.2 average depth in their >artificial intelligence thesis. > >Then last but not least. No one got 12 ply in those days. > Deep thought got 11 plies in 1989 in Alberta. I was there. Cray Blitz was doing 10 plies that same year. I was also _there_. There were plenty of cases where we did get to 12 plies, so that is simply a false statement. I don't have any electronic logs, but I certainly have a file folder full of printed logs from games dating back to 1981 or so. Anybody could stop by and see that 10-11 plies was certainly doable, at speeds of only 400-500K nodes per second. Because we were doing that... >I only remember fritz3 which by very dubious means got like 11 >ply. Basically some preprocessors did get 11 to 12 ply thereby >forward pruning last few plies a lot. > >None of them had things like simple mobility in the leaf eval >even. > >Now deep blue got in the past when it was deep thought with >500k nodes a second after 3 minutes sometimes a search depth >of 8 ply. With 126MLN, from which like 95% was wasted to >parallel search, they got 12.2 ply. That is utter bull snot. Belle, at 160K nodes per second, searched 8 plies _all_ the time and to depth=9 quite frequently. I have a log file Ken gave me from a Belle vs Cray Blitz game that shows this quite clearly. Belle had no null-move whatsoever, no forward pruning, just pure brute force plus typical search extensions like check and recapture. It did have hash tables in the hardware also. And it _did_ reach 8-9 plies in 40/2hr games, _evrey_ time. Deep Thought was _way_ faster than belle and was going 2 plies deeper the first time it played. As chiptest it was doing 10 plies. The multiple-chess-chip deep thought versions were doing 11. Again, I was _there_ looking. > >I see that as a good achievement. >And it makes sense to me! That doesn't say a whole lot, now does it? > >>On October 11, 2002 at 07:12:34, Vincent Diepeveen wrote: >> >>>On October 11, 2002 at 04:08:49, Uri Blass wrote: >>> >>>Isn't his article clear enough yet? >> >>Bob Hyatt still claims that it was 12 plies software and 6 plies hardware >>so I prefer to hear an answer directly from Hsu. >> >>> >>>reporting a 12.2 average search depth fullwidth. >>>I guess you never searched with a decent program fullwidth >>>with extensions. If you did, you would understand that >>>getting 12.2 ply fullwidth with loads of extensions is already nearly >>>impossible. Every extended line is searched to the full depth, >>>no pruning happens! >> >>I agree that 12.2 plies with a lot of extensions and no pruning is impossible >>for normal programs and also is impossible for deep blue in case that >>there were real 6 more plies in the hardware. >> >>The only case when it may be possible is if the 6 more plies in the hardware are >>real selective search and it means more pruning than null move with R=3 and in >>this case the 6 plies in the hardware may be eqvivalent to only 2 plies in >>software because of big probability to miss things. >> >>> >>>The interesting 2 questions are >>> a) did DB use 'no-progress pruning' in SOFTWARE (we know >>> already it used it in hardware). >> >>They explained in the article that they did not want to take risks of missing >>something in the first plies so it is clear that they did no pruning in the >>first 12 plies. >> >>If they did some pruning in the software it is clearly after it. >> >>I do not know what is exactly no progress pruning. >>Is there a difference between it and null move pruning? >> >>Is no progress pruning more aggresive than null move pruning? >> >>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.