Author: blass uri
Date: 10:37:19 04/06/99
Go up one level in this thread
On April 06, 1999 at 12:07:48, Keith Kitson wrote: >On April 06, 1999 at 07:13:20, Laurence Chen wrote: > >>On April 06, 1999 at 00:47:21, Brian Smith wrote: >> >>>Any opinions on what program is best positionally and what has the best >>>combination of tactical and positional skills against humans (I know this >>>question is also in the REBEL poll, but the reults might (or might not) be >>>influenced by the many REBEL owners at going to the site. >>> >>>Thanks >>> >>>(sorry if this has been discussed a millions times already... I guess I came >>>late) >>There's no such thing as the best positional chess engine. If you want to get a >>clear picture without any bias then my advice is to get as many different chess >>engines as you can afford to buy. I mean different chess engines, and not >>different personalities or settings one can modify in a single chess engine. The >>reason is all chess engines are weak in this department, they all suffer from >>"horizon effect", and they don't give the same evaluation for the same position, >>some engines are stronger in some chess positions than others. So if you have >>many different engines, you'll be able to better appreciate the chess position >>and be able to get to the true evaluation of a position. Don't fall into the >>pitfall of falling in love with a chess engine, remember that love is blind, and >>the blind leading the blind, then both of them will fall into the ditch. :) >>Laurence > > >An interesting point of view. In my opinion it is possible to determine the >strongest positional program to be the one that comes out top the most on >positional 'decisions'. A lot of work has already been done in some circles in >this area, and although not particularly impressive compared with good human >chess players the 'knowledge programs' seem to do best (i.e. Hiarcs 7.01, >Rebel10c and MCP8...possibly in that order) > >I reckon chess program writers have a long way to go with strategic decision >making before they approach the capabilities of human players, however, at the >present time there appears to be two main schools of thought in engine approach >1. Speed >2. Knowledge I do not agree about it. It is not so simple. Hiarcs7 is good at tactics and fast searchers can evaluate positional factors. I saw many cases when Fritz evaluated positions with material advantage to one side as an advantage to the other side and not because it knows more than only to count material. In one of the ssdf games Fritz5.32 sacrificed 2 pawns for space to its pieces and won. It did not sacrificed the pawns because of tactical reasons. I think that it is possible that the only difference between slow searchers and fast searchers is that slow searchers see more plies by evaluation. I tested Fritz/Junior at 1 ply depth and they can do stalemate with positive evaluation. They need to search 2 plies to see the stalemate and avoid the move that cause stalemate. If a slow searcher see the stalemate in 1 ply then it is seeing one more ply in this case by evaluation and it has nothing with positional unserstanding. It is possible to evaluate other tactical factors by evaluation(for example if you can do a fork in the next move). Uri
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.