Author: Uri Blass
Date: 21:45:03 03/24/03
Go up one level in this thread
On March 24, 2003 at 15:04:53, Robert Hyatt wrote: >On March 24, 2003 at 14:32:19, Uri Blass wrote: > >>On March 24, 2003 at 13:56:15, Robert Hyatt wrote: >> >><snipped> >>>>>If I had to play Belle today I would _not_ take it lightly. I'd expect to win, >>>>>but I'd know >>>>>that Belle was very capable of beating anything in one game. >>>> >>>>I expect Crafty to win against it even if you use only p90. >> >>I responded to Belle here and not to deep thought here so your comments are not >>relevant. > >OK. I overlooked that. > >Crafty on a P90 is _not_ going to do well against Belle. Belle searched about >170K >nodes per second. Crafty on a P90 is going to be about 1/10th that speed. I >don't think >null-move will make up that much ground, since testing with and without >null-move in >previous years suggested that null-move was worth maybe 50-70 rating points. > >Crafty on a pentium-pro 200 searches about 64K nodes per second on the bench >command, which I just tried to be sure that number was right. > >That is 2.5X faster than a pentium 133, which is about 1.6X faster than a P90. >So the >P90 NPS for crafty will bei n the 16K nps range, which is roughly 1/10th the >speed of >Belle. Belle's speed is better than three doublings, so crafty is going to have >a tough >time of it with that machine. > >Crafty on current hardware would be a totally different animal, but on a P90, it >might >be ugly. > >> >>I responded only later about Deep thought and >>My opinion is that you cannot take lightly Deep Thought but you can expect to >>win against it because you are better in tactics(I say it for the hardware of >>today and not for p90). > >I don't think I'm better in tactics than they were/are. I might have some >advantage >due to null-move, but then they have singular extensions. Evaluation is a >toss-up as >they were not "dumb" by any stretch, since they beat GM players handily. From the book "how computer play chess" page 192 Table10.1 Correlation between search depth and chess rating. Depth of search 8 year 1980 program Belle Rating 2200 Depth of search 9 year 1986 program Hitech Rating 2400 Depth of search 10 year 1989 program Deep Thought 2600 Belle does not seem to be something that beats GM players handily based on this data. There is also a problem to compare results against humans and I prefer to compare moves. I believe that null move is not the only reason for Crafty to be better in tactics than Belle and there may be other reasons like better order of moves. Here are some tactical error of Belle from game that it won in 1978 against Chess4.7 [D]r2q1rk1/1bp1b1n1/1pn1p2p/p3P1p1/P2p2PP/2PB1N2/1P1BQPN1/R3K2R w KQ - 0 20 bm Qe4 Belle blundered by hxg5 but the opponent did not find the right defence. [D]r2q1rk1/1bp1b1n1/1p2p2p/p3P1P1/Pn1p2P1/2PB1N2/1P1BQPN1/R3K2R w KQ - 0 21 am gxh6 bm Rxh6 Maybe finding Rxh6 is not easy but avoiding the mistake of Belle(gxh6) is easy more interesting test position for the programs of today is the following position: [D]r2q1rk1/1bp1b1n1/1pn1p2p/p3P1P1/P2p2P1/2PB1N2/1P1BQPN1/R3K2R b KQ - 0 1 am Nxb4 bm dxc3 Chess4.7 played Nb4 and in the book it is claimed that it i a good move but movei like other programs can fail low on that move after some time and prefers dxc3 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.