Author: Uri Blass
Date: 08:13:04 11/15/00
Go up one level in this thread
On November 15, 2000 at 10:07:05, Andreas Stabel wrote: >On November 15, 2000 at 09:20:57, Laurence Chen wrote: > >>On November 15, 2000 at 08:31:29, Bob Durrett wrote: >> >>>I have used Fritz a lot for this purpose. >>> >>>Is there something better than Fritz? >>> >>>What is it? >>> >>>What makes it better? >>It depends on the type of position you wish to analyze. If it's tactics, then >>any top chess engine will do the job, if it's positional, that is open for >>debate and personal preferences. It's like using a tool, one needs too know how >>the works and its application. For example, one would never use a hammer to >>screw a screw, although one can do that, nothing in the books which says one >>cannot do it. It reminds me of an episode of the Simpsons where Maggie ask >>Homer to do some house chores, his son Bart volunteers to help. The first job >>was to open a drawer which was stuck in the living room. Bart comes with a >>brilliant solution, he tells Homer to use fireworks. As Homer is getting ready >>to lit up the fireworks, Maggie shows up and asks Homer what he is doing. Homer >>doesn't know what to say, Bart asks his mother if she wants the job done or not, >>she says yes, and Bart lights up the firework. The firework explode and gets >>the stuck drawer open. From there on, Homer and Bart uses fireworks to fix all >>the house problems, except that one job it backfired, they decided to fix a TV >>with fireworks for Lisa, the fireworks destroyed the TV and made a huge mess of >>Lisa's room. Like I said, it depends what you want to use the chess engine for, >>and what type of position are interested in analyzing. IMHO, open positions, >>Fritz does a very good job, so does Chessmaster, and now that Gambit Tiger 1.0 >>is available, I like Gambit Tiger more than Fritz to analyze open positions. >>This engine (Gambit Tiger) surely knows how to attack and create attacking >>chances, sometimes it backfires because it is willing to accept bad pawn >>structures in order to open lines, and when its attack fails then its weak pawn >>structure takes its toll, but it is surely very aggressive. It knows how to >>maneuver its pieces in closed positions with the idea of opening up the >>position. My two cents. >>Laurence > >Another important feature for analysis is how the program works on long >thinking time. When I analyse games I might use the program to run up to >several hours on the same position and then it is very important how the >program handles hash tables etc. F.ex. I have found that Virtual Chess II is >impossible to use for this because it has terrible move ordering on long >thinking times - possibly because it hash tables overflow. > >As far as I can remember the new Rebel Century (or was it Tiger ?) has an >option for over night analysis which might be useful for such things. > >Is there any other programs with a similar feature ? You can use every program to analyze one position for hours. Hiarcs has selectivity parameter and it is claimed that bigger selectivity help the program at long time control. I know that Rebel has over night analysis but I dp not know if this analysis is productive. I read that one of the ideas behind the over night analysis was to reduce the extensions because tactics is not important at that time control and it is better to see more plies forward. I am not sure if it is the case because I know that extensions are important also in long time control(see the nolot test). I think that it is better to use hard test positions in order to decide which version is better at long time control. You can take also positions from the endgame. Here is one example This position is from one of the Gandalf-Fritz games. [D]8/5k2/6p1/3K1p1p/1R3P1P/6P1/3b4/8 w - - 0 1 White can win the game by the plan of getting the king forard. In order to do it it needs to prevent the bishop from getting to the long diagnol. In the game gandalf failed to do it and drew the game by Rb1 that let the move Bc3. Rb7+ Kf6 Rc7 is probably a win for white and programs do not understand it because they are afraid from Be1 that wins the g3 pawn but let the white king to get into the files e and f in order to take the g6 pawn and later to take the h and the f pawns in order to get KRP vs KB winning endgame. I did not prove that white wins against every defence but it seems to me that black has no chances. The process takes many moves but I think that black has nothing to do against it. The problem here is not to find Rb7+(programs can find it not for the right reason) but to win the game and the question is if the program are capable of doing it in long time control against the right defence(you can try 1 hour per 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.