Author: Vincent Diepeveen
Date: 04:42:34 05/27/05
Go up one level in this thread
On May 26, 2005 at 19:54:15, Dann Corbit wrote: >On May 26, 2005 at 19:42:38, Arnor Yogev wrote: > >>Hello. >>Is It possible that the core engine of hydra is just the same of Nimzo latest >>version? > >No. > >>After all, it is written by the same guy, Dr. Chrilly Donninger, who hasnt >>updated Nimzo8 since 2001. > >Crafty was written by Dr. Hyatt as was Cray Blitz. >There will be similarities between them, but they are not the same program. > >Hydra has hardware accelerators. So something like Rxa4 is a CPU instruction -- >hardwired into the FPGA. > >It's a highly similar idea to Deep Blue (or Belle for that matter). > >The fundamental algorithms of Hydra will probably be highly similar to the >fundamental algorithms of Nimzo (or any other chess program, for that matter). There is huge differences between hydra and nimzo, apart from the software versus hardware difference. first of all hydra focuses upon searching as deeply as possible and it really is tactical weak the last few ply in hardware; as a matter of fact it's forward pruning in a very simple way the last few plies in hardware and chrilly tries to search as little as possible plies in hardware (usually just 2). The reason for that is simple, hardware is so so inefficient that you try to get everything as much as possible within software. If you do a simple measurement of any program the last 2 ply with huge forward pruning you'll still see that the vaste majority of nodes happens in those last 2 ply, so "just 2 ply in hardware" is not so simple at all. Yet where Hydra focuses upon really searching as deep as possible, Nimzo definitely was focussing upon getting the branching factor worse and worse when getting deeper; it was extending and extending. Obviously Chrilly somehow cannot do this with Hydra. Secondly Nimzo was basically 1 big piece square table where he had made a 'language' Che to initialize piece square tables. Hydra is far more agressive and far more a leaf evaluation based program than Nimzo is. Yet it will be so so complex to even get some simple eval to work in hydra that we must not overreact here. Compared to the 'average' commercial, hydra has little knowledge. its play shows it. What will not be different is the tuning Nimzo vs Hydra. I feel Chrilly is one of the best tuners in computerchess. After a game you always wonder: "how did i manage to lose from this program?" Chrilly's is mister KISS principle in own person. So what hydra and nimzo share is that they are real strong for 1 year and the next year they are completely outdated. Yet that's not so important in this matter, as Hydra will be most likely playing just humans, no more computers. If you improve Nimzo's evaluation a bit, let Chrilly retune it and put it at a fast A64 processor, it just will show a similar performance in such a match against Adams like Hydra will. Yet no one will understand that except some big computer experts. The important surprise Chrilly has for the human opponents is the big agression of his programs, way more agressive than others are, towards the king side. Additional it runs with all pawns. Though that isn't positional that good, it gives a totally different playstyle from the commercial programs where GM's are used to. Another thing Chrilly seems to be good at, according to Frans Morsch, is to create a lot of noise around him and get all the attention. Yet i feel Chrilly deserves that this time. Vincent
This page took 0.01 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.