Author: Richard A. Fowell (fowell@netcom.com)
Date: 11:41:02 11/29/98
Go up one level in this thread
(from r.g.c.c - here is Mark Uniacke's comments on Enrique's post) In article <73rmmo$j4a$1@diana.bcn.ibernet.es>, Enrique Irazoqui <enir-nospam@intercom.es> writes >I have been playing some blitz games with Hiarcs 7 beta against Junior 5 and >Fritz 5. Results at game in 5 minutes on PII-400s: > >Hiarcs 7b - Junior 5 25-9 >Hiarcs 7b -Fritz 5 14-9 > >At first I didn't expect to see such a slow searcher defeating this way the >NPS monsters, but it did. Taking into consideration that game/5 on a PII-400 >is roughly like game/90 on the 486/50 of not so long ago, the very good >scores of Hiarcs 7 may be less of a surprise. Dear Intel is helping the slow >searchers... Slow searching does not mean weak tactics or positional play - speak to Kasparov ;) Seriously, Hiarcs 7.0 seems to be able to compete tactically with the other top programs while examining far fewer nodes. > >I still believe that a relatively weak point of Hiarcs 7b is king safety. It >seems that Mark Uniacke improved it in the final version, but the beta does >things like this every now and then: > > In the 6th game Hiarcs7b-Junior5, after move 21 they reached this position: > 1r1q1rk1/2n1bpp1/2b1p2p/3pP3/1p3P2/pP2B1N1/P1PQB1PP/1K1RR3 w - - 0 22 > Here H7b played 22.c3, which looks quite unsound to me. With the rook in the > open b column and these mean looking bishops, black could try to build a > smashing attack. > > This is not a problem that Hiarcs has all that frequently. I saw it no more > than 3 or 4 times in over 100 games, but it still is a weakness. When it > happens, against other programs Hiarcs 7b can get away with murder, but > against humans I wonder. It does'nt happen that often, but it is there. I spent 3 weeks trying to improve things here. Sure I could improve move selection in these type of positions greatly, however, for some reason the auto test results were worse, so I settled on a small improvement here and in the pawn majority, pawn pawn storm type positions which showed best overall in auto testing and human/computer testing. The overall results seem to point to Hiarcs 7.0 being a few points stronger than Hiarcs 7.0 beta. The improvement is significant enough to measure over a small number of games like 60-100. Hiarcs 7.0 beta looked from beta testing and various tests to be a GOOD margin stronger than Hiarcs 6.0. It is always difficult to put a figure on it, but I will stick my head out.... at blitz it could be as much as 100+ Elo better, at normal tournament time controls it is more like 60 Elo in the testing done so far. See http://www.acc-ltd.demon.co.uk for further details. > >Enrique > Best wishes, Mark Uniacke
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.