Author: Robert Hyatt
Date: 17:48:43 07/11/98
Go up one level in this thread
On July 11, 1998 at 18:49:39, Roberto Waldteufel wrote: > >On July 11, 1998 at 13:44:32, Bruce Moreland wrote: > >> >>On July 11, 1998 at 11:45:25, blass uri wrote: >> >>>I am interested in positions only one program can solve >>> >>>Vincent Diepeveen posted on June 29 the following 2 positions >>>r4rk1/pp1n1p1p/1nqP2p1/2b1P1B1/4NQ2/1B3P2/PP2K2P/2R5 w - - bm >>> >>>the solution is Rxc5 and diep found the move after 6:42 >>>and found the advamtage for white after 13:14 >>>he used 60MB hash tables >>>and I think he used pentium200MMX >>>my programs found only Bf6 with draw. >> >>I thought I posted something about this last week. >> >>Mine finds this in two minutes on a P2/300. The score it gets isn't great, it's >>a little negative. Other versions have found this quickly with scores of +2 or >>more. >> >>>can another program solve this position in 1 hour? >>> >>>the second position: >>>r1b2rk1/1p1nbppp/pq1p4/3B4/P2NP3/2N1p3/1PP3PP/R2Q1R1K w - - bm Rxf7 >>>diep found the move after 22 minutes and 43 seconds >> >>This one takes four minutes to get a slight positive score, five minutes is +1. >> >>I like the Nolot positions, I've been fiddling with them for a few years. If >>you want to see more from me, look for "Nolot" in the "old" archives in r.g.c.c, >>or perhaps even just rec.games.chess. >> >>I can get solutions in some of the other Nolot positions as well. >> >>>can another program find this move in 1 hour? >>> >>>(according to Robert hyatt deep thought solved the position after 2 minutes but >>>I cannot buy deep thought so it is not important for me) >>> >>>in Rebel's homepage there is mate in 30 Rebel can find(all the moves of the >>>loser are forced). >>> >>>can another chess program find the mate? >> >>I can't find it, feel free to post it. >> >>Long mates are often the result of positions where one side is in check and has >>only one legal way out. >> >>One very old extension in computer chess involves extending when you make a move >>that gives check. >> >>If you also extend on the reply, when there is only one way out, you get the >>checking move and the reply without decreasing depth at all. >> >>If there is a long forcing sequence you can find mates in very large numbers of >>moves in just a few nodes. This has been well-known for years. >> >>It is hard to find long mates that don't involve this extension. >> >>bruce > >Hi Bruce, > >I never reduce the depth when the side to move is in check, but I find that if I >do the same for moves that administer check it sometimes blows my search sky >high, hanging the machine for an inordinate amount of time. It's a pity, because >I also found many times that this method discovered long mates as you describe. >How do you overcome the problem of situations where the number of checks becomes >excessive, eg a king being chased around the board by a queen that lacks >supporting pawns or pieces to deliver a mate? > >Best wishes, > >Roberto Bruce and I are similar here I think. I extend on a checking move, rather than extending to get out of check. At the next ply, I can extend if there is only one legal move and I am in check... The thing that should limit this is being careful. 1. No more than 1 ply of extension at a given ply. More will guarantee a non-terminating search. 2. repetition code will stop a long series of checks due to a two-time repetition. 3. some positions will get very shallow searches because the checks will extend like mad. There's nothing you can do for those cases. This is why I use 3/4 ply for the one-legal-reply-to-check extension, because that means that I only extend 3 of every 4 one-reply positions... which gives me a better limit...
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.