Author: Uri Blass
Date: 11:30:27 07/20/02
Go up one level in this thread
On July 20, 2002 at 13:14:48, leonid wrote: >On July 20, 2002 at 10:22:38, Uri Blass wrote: > >>On July 20, 2002 at 08:31:35, leonid wrote: >> >>>On July 20, 2002 at 04:28:57, Uri Blass wrote: >>> >>>>On July 20, 2002 at 00:03:49, leonid wrote: >>>> >>>>>Hi, >>>>> >>>>>This mate is, probably, easy: >>>>> >>>>>[D]Bn4bb/1PqPQnP1/2Prrq2/2KPBk2/2NRRq2/1PNqqQq1/1q4R1/q6q w - - >>>>> >>>>>Please indicate your result. >>>>> >>>>>Thanks, >>>>>Leonid. >>>> >>>>Movei says mate in 11 at depth 7 after almost 10 minutes >>>>(574.57 seconds). >>>> >>>>Movei does not look for the shortest mate so there may be shorter mate. >>> >>> >>>Solution is excellent since it is mate in 11. >>> >>>Can you briefly explain what kind of search for mate this program do. >> >> >>Movei is a chess program that I developed. > >Good to know that you don't loose your time like me. > >>There is a version for download but >>it is not the latest version. > >I will lately try to look into what it is through Alta Vista engin. I hope that >it is with graphics interface. I am not very used to engins. Only exception >until now was Heiner's mate solver "Chest". There is simply nothing like this >program for mate search anywhere. Movei can be used under winboard. It has no graphics interface It is a winboard engine. You can see a list of all the free winboard engines in http://home.hccnet.nl/leo.dijksman/EnginesIndex.html You can find a link to download Movei there if you click on M I guess that the public version is going to have bigger problems because I do not do checks in the qsearch in the public version. There are positions when the public version crush but it does not crush in the position that you posted. The public version is weaker in finding the mate in the position that you posted because it does not do checks in the qsearch so it cannot find the mate at depth 7 and needs at least depth 8. I decided not to wait for it to get depth 8 and I do not know what depth is needed for it to find the mate and how much time it needs to find the mate(I guess that it may need hours to find the mate). I did a mistake in my array for bishops and my array for knights and it assumes that the maximal possible number of bishops or knights is 9 so it can crush if you give it a position with 10 bishops or 10 knights. There are some other bugs that are relevant for games but they are rare and it may crush maybe in 1 out of 1000 games. The practical problems that I found it that it can crush in a stalemate position and it can crush in some rare positions when it assumes wrongly that en passent capture is legal when it is illegal. > >>The program simply play chess and also knows to read Fen. >>The program is not supposed to be a mate solver. > >Then your program do perfect work! I doubt very much that mine will find mate in >11 moves, through what I call "material exchange" logic, at your time. > >> I have the >>>impression that I never heard about this program. If this program look for mate >>>with some kind of extentions, or search for mate is done by some independent >>>segment in the program, like me program do. >> >>It simply searches for the best move like >>other normal chess programs. > >So it should be by looking into what move give overwhelming material exchange. Yes At small depthes it saw advantage for black and only at depth 7 it suddenly saw that white can force a checkmate. >Probably few other way are possible. Mine mate solver don't look into material >echange but only into fact or existence, or not of mate at given depth. MAte >will be considered as found not when certain material is accumulated but when it >reach position when one side is under check and have no moves to be done. Movei evaluates mate as 10000-distance to mate in plies when 100 means 1 pawn. It means that mate means almost 100 pawns in it's evaluation. Usually it does not cause problems but if I give it position when white has 9 queens and 2 rooks and 2 bishops and 2 knights against king it may get confused because the total material of white is evaluated as more than 100 pawns. It is not a practical problems because even if the opponent is very weak ,movei is going to find a way to force checkmate a lot before having 9 queens. > >>It is not at the level of the top programs and >>I guess that it may get a rating of 2200 against humans. > >More that good! Would like to be there. > >>Part of the commercial programs like Junior7 >>can do clearly better than movei in your position >>but when I tried deep fritz >>it crushed. > >It is normal for Fritz and many other programs. I was surprised when I found >this. Certain program that are not so smart at playing chess but not alergic in >solving mates. One of them is Chessmaster. > >>Junior7 can see the mate in 11 in only 35 seconds on p800 >>(16 mbytes hash). > >>Blass,U - Junior 7 > >Is the Junior 7 some other version of your program, or Rebel Junior? Junior7 is a commercial chess program of amir Ban. It also does not crush in solving mates but after finidng a mate it does not continue to search deeper for shorter mate(the same is for movei). Chessmaster is different and it continues to search also later for shorter mate. 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.