Author: Peter Berger
Date: 02:45:16 10/09/05
Go up one level in this thread
On October 09, 2005 at 01:10:41, chandler yergin wrote: <snip> >Bruce, our own CTF Programmer explained it very well in a Post >some time ago. >"To min-max chess using alpha-beta would require a horrific tree search." > > >" In 40 moves (80 plies), which is clearly a horrific >under-estimate, the game tree size is on the order of 10^62." > >"Even at a billion billion nodes per second, you could search for a billion >billion seconds, and you'd only be a billion billionth of the way to being a >billionth of the way done." >This is what Computers do.. Brute Force basically. >They ‘evaluate' every possible move in a position and rank every one, >best first, on the feedback they get from the above noted algorithms. A wrong conclusion. This argument was taken from a discussion about solving chess I assume. Chessprograms don't just use a plain alpha-beta search - e.g. , elsewhere you mention pruning yourself . : pruning basically means discarding huge branches of the game tree, preferably irrelevant ones. >Since Opening Books are used.. The true playing strength of the Engine >has taken a back seat so to speak. Opening Books are based on games >‘played'! Does this mean it is the latest theory or the best moves? >Of course not! You can prove this for yourself. Load an Opening Book, >doesn't matter which one, Open it, now add a ‘kibitzer' Engine. >Go through the Opening Book while looking at the PV from the kibitzer, >you will often find the ‘best' move is not even in the top 6 choices. >The Opening books that come with the Programs are outdated. >A high depth search will be based on very few games.. Old ones at that. >Example: What made Fritz 6 stronger than Fritz 5.32? Or Fritz 7 >stronger than Fritz 6, or Fritz 8 stronger than Fritz 7 etc.. Etc..? >The Opening Book of course. What did the Opening Book consist of? >Games published in Chessbase Magazine. With each new release thousands >of new games had been added to the Database & the Opening Book. >Just open the Database and click on Sources. Interesting? >Now, all the Top Programs when out of book use basically the same Algorithms. There are many mistakes in this paragraph IMHO. But a few of them you can verify for yourself. Many people play computer tournaments without any books, starting from selected positions. They show that newer Fritz versions are clearly stronger than Fritz 5.32 - and the opening book can't be the reason for that. Actually I even think that you won't get a strong or stronger opening book by the process you describe, but that's another topic. >Bruce has a wonderful Website which explains in great detail all of the >Algorithms and short cuts Programmers have made to Prune the chess tree & >insignificant PV's. >Sorry to say I lost the Link, perhaps he would be kind enough to Post it again? >Having said "A" I must now say "B" >There has not been any significant change in the Programming Algorithms for >30 years? Therefore what can we attribute the fantastic result of some "new" >Programs defeating Programs like Shredder? >He who searches deeper & faster when out of book will find the better moves! >Simple as that! A few better moves in the middle game results in forced wins in >the end game. Faster..deeper.. That is the Key to Fruits Success! Lack of EGTB's >is not a handicap, in fact it speeds up the Search. >Also, it is not encumbered by massive Hash Tables! >Prove it you say? >No Problem! >Give your Engine the highest Hash Tables optimum for your System. >Check your kN/s! >Reduce your Hash tables to 1 MG and check your kN/s! >Now.. If you want to truly check Fruit vs Shredder, get a level playing field! You made an observation here: higher NPS with a much lower size of hashtable, something that makes sense to me. But I don't understand what you are trying to conclude from that here. Peter
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.