Author: Vine Smith
Date: 13:18:44 05/15/02
Go up one level in this thread
On May 15, 2002 at 15:47:50, Harald Faber wrote: >On May 15, 2002 at 15:09:27, Vine Smith wrote: > >>On May 15, 2002 at 11:28:37, P. Massie wrote: >> >>>My version of Hiarcs8 (running in the Hiarcs8 GUI) has a 9% probability of >>>playing Nc6 on move 2 as below. Following that it has a 100% probability of >>>playing 3 ... Be7, although e5 is listed as a 0% probability. The opening book >>>is exactly as arrived on the CD, with no changes. >>> >>>Paul >>> >>>On May 15, 2002 at 08:13:53, Chris Taylor wrote: >>> >>>>On May 15, 2002 at 05:14:21, Vine Smith wrote: >>>> >>>>>Hi -- >>>>> >>>>>This is in reference to some issues about the book, particularly a recent post >>>>>by Harald Faber. I had seen a game posted (at game in 30 minutes) where Hiarcs >>>>>as Black played 1.e4 e6 2.Qe2 Nc6 3.Nf3 Be7? Faber says that when he looks at >>>>>the Hiarcs book (I think he's just using the Chessbase GUI function), it says >>>>>Hiarcs will always play 2...c5. For this purpose, the time control and system >>>>>speed are unimportant, just try playing an actual game from the position after >>>>>White's second move, and see if it immediately responds with 2...c5. If so, I >>>>>suppose the game I saw posted was played with a different book, or is spurious. >>>>>If not, then the program is not understanding its own book correctly. >>>>>Thanks in advance, if you have the time for this. >>>>> >>>>>Regards, >>>>>Vine Smith >>>> >>>>Playing against Hiarcs 8 with >>>>1. e4 e6 >>>>2. Qe2 c5........It played this move sort of expecting either f4 or Nf3 as >>>>Whites third. Nimzo 7.32 gui......Identical in H8.gui >>>> >>>>Looking for 2....c5 and browsing the book we see >>>>N=4182 and 50% chance of being played >>>>2....Nc6 has N=261 with 49% >>>>Also in is 2....Ne7 N=257 50% >>>> >>>>The plot thickens.. >>>> >>>>Chris >> >>This is weird -- between your response, and Taylor's and Faber's, it looks like >>nothing has been cleared up. I'm almost tempted to buy the program myself just >>to experiment more and see what exactly is going on here. >>Other lines that Faber disputes Hiarcs would ever play are 1.e4 e5 2.Nf3 Nc6 >>3.Bb5 a6 4.Bc4?!, which was posted by Walker under message number 228941 on this >>forum, and 1.e4 c5 2.Nf3 e6 3.d4 cxd4 4.Nxd4 Qc7?!, posted by Taylor under >>message number 229441, game 3 of a series he conducted against Genius 3 on >>separate machines. Under the Walker post, I also noticed Hiarcs playing 1.e4 c5 >>2.b3, which isn't exactly terrible or anything (White probably can't get an >>advantage after Black plays ...d6 and ...e5, though), but might be another >>unusual line which Hiarcs shouldn't be playing. >>Is it possible that different versions of Hiarcs were distributed, possibly one >>with some kind of defect in program-book communication at first, followed by a >>"silent" bug-fix soon after? >> >>Regards, >>Vine > >At the moment it seems to me that Taylor and Massie do not know what all the >values and columns in the opening book window mean. >What I cannot explain is why such games with the described openings could >happen. Maybe someone has adjusted some bars in the book options? I have read some more posts, and I think the situation is actually what Thorsten Czub theorized in another branch of this thread. Walker, in a message above, says he just switched from using Hiarcs under the Junior interface (because he didn't like the Chessbase 8 GUI), and is suddenly getting better testing results. The Taylor game was played against another computer using Auto232, which I don't have or use, but I imagine it must be configured for each GUI being connected. In that case, he might have already had it set up to communicate with Nimzo (he mentions having this), and so it would be tempting to have Hiarcs play under the Nimzo GUI. If so, for inexplicable opening play, that possibly leaves Massie (unless he is really misreading his data), and the Chigorin French game which was posted very early after Hiarcs' release, so the message has fallen off the board and into the archives. It seems very likely to me now that Hiarcs MUST use its own interface to read its book properly, and for engine vs. engine matches, the other engine must visit Hiarcs at its "home", not the other way around. Chessbase should probably make note of this, for instance in Lopez's T-Notes column or something like that. Also, anyone posting games should always include details about which GUI any match was conducted under. Regards, Vine
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.