Author: Robert Hyatt
Date: 06:13:26 03/15/99
Go up one level in this thread
On March 14, 1999 at 22:25:57, blass uri wrote: > >On March 14, 1999 at 21:30:02, Robert Hyatt wrote: > >>On March 14, 1999 at 16:29:07, blass uri wrote: >> >>> >>>On March 14, 1999 at 15:52:27, Mark Rawlings wrote: >>> >>>>This is really an interesting position! Is the final consensus that it was a >>>>sound sacrifice? I let my programs search fairly deep with no real conclusion. >>>>After 1. .. Bh3 2. gxh3 Kf5 3. Kf2 Ke4 black still has a challenge forcing >>>>the win. I know some others must have been running this with faster computers, >>>>tablebases, etc. Any conclusions? >>>> >>>>Mark >>> >>>I think that it is a simple win for black. >>>I suggest you let your programs to play one against the other in slow time >>>control after 1...Bh3 2.gxh3 Kf5 3.Kf2 Ke4. >>> >>>I do not think that white has a chance. >>>4.Bxf6 is losing against 4...d4 and if white does not take f6 then black simply >>>puts the pawns at d3 and f3. >>> >>>Tablebases cannot help in this position because the lines that black is winning >>>are not based on tablebases(there are more than 5 pieces in the board) >>> >>>Uri >>> >> >> >> >>the last statement is wrong. IE the advantage for probing tablebases in the >>search is that the program gets to pick the path to 'enter' the tablebase. I >>can show you _lots_ of positions with 12 pieces on the board where Crafty has >>announced a mate in 60+ because it found a deep way to force entry into a won >>tablebase position. And it has happened with 16 pieces on the board also, once >>against a GM two years ago in a game/30 tournament on chess.net (Roman was the >>victim, mate in 28 was the announcement). >> >>I get thousands of tablebase hits from the position where black plays Bh3, >>so tablebases do influence this significantly. > >I think that in this case tablebases will not help to see mate because black has >more than 5 pieces in the board in the winning lines even after many moves when >the computer understand that black is winning. > >I think that you get thousands of tablebases hits to see draws and not to see >mates in relevant lines. > >Tablebases may help in this position only not to calculate draw lines and help >crafty to go deeper. > >I think that there is a simple test to see if tablebases can help in this >position. >If you can go deeper with tablebases in this position then tablebases help >otherwise they probably do not help. > >Uri That was easy to test. I just ran this for 60 seconds. Got 2972 tablebase hits (from the initial tablebase position where black should play Bh3.) Of these, 1612 hits were 'draws'. 917 were mate-in-N for black winning, and the rest were mate-in-N for white winning. When you search to depth=16-20, it is _trivial_ to turn that position into 5 or fewer pieces...
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.