Author: Mark Andreoli
Date: 08:10:06 10/01/99
Go up one level in this thread
On September 28, 1999 at 18:50:31, Ratko V Tomic wrote: >On September 28, 1999 at 15:29:39, Jeremy McComas wrote: > >>I just ordered Hiarcs 7.32 and I was wondering what people think of this >>program. I kinda got scared when people started saying that hiarcs was making >>bizzare moves. > >It is probably the best of the bunch, anyway (and I have played most of them, >only Rebel 10b is close in understanding of positions). Last weekend I got up >and decided to play a few fast games against Fritz 5.32, and after being >outplayed from start to end in three games, I went to breakfest thinking how >Fritz isn't that bad positionally. I wondered how could it have improved so >drastically through its book learning. Well, when I came back to play some more >I noticed it was a Fritz program, but playing Hiarcs 7.32 engine (I was testing >some positions the night before and forgot to return Fritz engine to Fritz >program/shortcut). Well that explained how suddenly Fritz got some positional >sense. The difference in handling quiet positions, especially the blocked ones >is quite noticable. That mixup showed how striking the difference is, that even >being completely primed for Fritz couldn't prejudice my observation enough to >override the objective difference. > >As to flakey aspects, I did run into some opening lines where H7.32 mishandles >the opening as soon as it gets out of the book (I play its full book, not the >narrow torunament one). But other programs have such lines, too (I don't know >which one has the worst book). If you set-up the learning they won't lose twice >in the same line. In few games H 7.32 did pick a greedy or speculative line of >play and then lost, where it had a better position and could have won with a >less ambitious play. While the other programs have similar judgment problem, >with H 7.32 I couldn't reproduce its critical (bad) moves later. It may have to >do with its aggressive re-use of hash tables accros moves, so that no two runs >are identical since each move is affected by the exact timing (and thus the >content of the hash tables) of earlier moves. It could be a bug, too. > >But whatever it is, it is so infrequent, you can't count on winning a game on >it. While this random streak is a borderline flaw, it is minor compared to the >outright positional blunders by, say, Fritz. With Hiarcs you can sense planning, >strategy, as well as its attempt to twart your plans, to nip them in the bud, >well before you're threatening anything; these abilities are completely or >mostly absent with other programs (with partial exceptions for Rebel 10b). One >thing I noticed with H7.32 is something I didn't see with other programs present >or past: in some number of games (perhaps a quarter), somewhere well into the >game, with material still equal, I realize in amazement that my position is >squeezed of any possibility of counter-play, with almost all pieces still on the >board there is nothing remotely hopeful I could try, and I wonder how did it >manage to tie me up so completely. It reminded me of a sense of complete >helplesness I had in some friendly games back in college days with a neighbour >who was an IM at the time and a youth champion of Yugoslavia (Branko Damjanovic, >he became later a GM and is now one of the strongest players in Yugoslavia; I >later got rating around 2100 in few months of competitive play, when I was >graduate student of theoretical physics at Brown over ten years ago, but had to >drop competitive chess [probably before my rating reached its plateau] due to >the time demands of my thesis work). So if you enjoy in being thoroughly >trashed, tactically and positionally at the same time, you didn't go wrong in >picking Hiarcs 7.32.
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.