Author: Robert Hyatt
Date: 07:10:21 05/17/00
Go up one level in this thread
On May 17, 2000 at 05:56:36, Oliver Roese wrote: >On May 16, 2000 at 09:54:24, Robert Hyatt wrote: > >>On May 15, 2000 at 23:24:13, blass uri wrote: >> >>>On May 15, 2000 at 22:43:00, Robert Hyatt wrote: >>> >>>>On May 14, 2000 at 18:09:20, Pete R. wrote: >>>> >>>>>On May 14, 2000 at 08:54:35, Jürgen Hartmann wrote: >>>>> >>>>>>On May 14, 2000 at 08:44:57, Jürgen Hartmann wrote: >>>>>> >>>>>>>and ...b5 is a possibility for counter play. Frans Morsch will certainly sweat >>>>>>>blood and tears at the moment. Small consolation that all other chess programs >>>>>>>behave in the same way. >>>>>> >>>>>>Correction: Crafty 17.10/CB plays 13... exf4!, the only chance to avoid the >>>>>>death sentence. >>>>> >>>>>Well, according to comments from Dr. Hyatt, pawn structure evaluation seems to >>>>>be his major tuning focus for Crafty these days. Clearly this is a worthwhile >>>>>goal, since computers are already top notch at tactics. But the consequences of >>>>>pawn moves are often far too deep for tactical evaluation, so things like this >>>>>have to be caught in the positional evaluation. A program *should* be able to >>>>>"see", just as a human can, that white has a great pawn structure and will >>>>>continue to annex space if f5 is allowed. Not easy to program I suppose. >>>> >>>> >>>>Not easy... not impossible... >>>> >>>>Very important against these guys... In this case you simply can't allow >>>>the e4-f5 pawn chain to happen. The center becomes locked, white has all >>>>the space on the kingside, and black is constrained to live in a small shoe- >>>>box and wait for jack-the-ripper to show up... >>> >>>It will be interesting if there is some statistics about the results in the same >>>pawn structure after f5. >>> >>>In how many games the same pawn structure happened and how many games out of it >>>white won. >>> >>>Uri >> >> >>Take a wild guess about how many times this happened to Crafty before I tuned >>the eval to not allow a pawn chain like d3/e4/f5. >> >>:) >> >>I don't have _all_ the answers to such positions, by any stretch. And crafty >>will react violently to such attempts, sometimes so violently that it ends up >>with a weak position (but at least it doesn't get crushed after allowing f5). >> >>It is a very fine line to walk. I still have a lot of work to prevent this >>sort of stuff, even though I have worked on king safety repeatedly. It has a >>very good idea about open lines around the king, and not allowing them. But it >>does, on occasion, defend against a threat that is not serious enough to cause >>concern. And in chess, everything is 'give and take'. If you give more than >>you take, you can get into just as much trouble as when you allow f5 to be >>played... > >Interesting... >Recently i played a lot of games against Fritz 4.01 (100+, i got roughly 20%, >each having 15 m/g). A main avenue to success (amongst some others) was to build >up such a pawn structure. Computers are very bad in such positions and Fritz is >even worse, since he has absolutely no clue there. >(But beware! >Against most humans it is enough to have a strong attack and they will loose >their fate rather soon. Not so against computers. If you sacrifice, every move >must be correct otherwise you loose. I lost many games in this manner. ) >But against crafty i couldnt manage to get into that... >I think this is very important for good performance in a competitive >environment. >Practically a closed position is more worthy than an extra pawn in an opened >position against Fritz under some circumstances. (I mean it.) > >Oliver Roese I agree. So do many IM/GM players on ICC. :)
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.