Author: Robert Hyatt
Date: 07:58:31 12/01/99
Go up one level in this thread
On December 01, 1999 at 03:10:56, Ed Schröder wrote: >>Subject: Re: Positional/Real Sacrifice >> >> >> [ Post Followup (without quoting) ] [ Post Followup (with >>quoting) ] [ Computer Chess Resource Center ] [ CCC Home Page ] >> >> >>Posted by Robert Hyatt on November 30, 1999 at 17:45:23: >> >>In Reply to: Re: Positional/Real Sacrifice posted by Ed Schröder on >>November 30, 1999 at 04:40:20: >> >>On November 30, 1999 at 04:40:20, Ed Schröder wrote: >> >>>On November 30, 1999 at 04:12:22, Jeremiah Penery wrote: >>> >>>>On November 30, 1999 at 03:35:54, Ed Schröder wrote: >>>> >>>>>As said we must have a different view on king safety even on the basics as >>>>>this case for me is so crystal clear: QRRB all pointed at a naked king, the >>>>>king having hardly any escapes, >>>> >>>>True, but all it takes to wreck the whole thing is ONE escape for the black >>>>king, as long as black doesn't lose material doing it. In this case, I >>don't >>>>think there is an escape, but it will happen in other positions. >>> >>>And SEARCH will filter that. >>> >>>Ed >> >>not at 1-2-3 plies it won't. If you don't see the escape, you make the >>sac. If I don't see the win, I don't make it. That is a subtle difference. >>You want the search to refute the positional sac. I want the search to >>justify >>the sac... > >I understand your point of view, material safety has the highest priority >for you. Does that mean Crafty will never sac a pawn for positional >compensation?\ It does it regularly. But here is the issue: giving up a pawn for some positional compensation may or may not lose if the compensation evaporates, as in many games, 1 pawn is not enough to force a win. Sacrificing the exchange is a similar thing... since R vs N or B is a draw. But a piece for 2-3 pawns almost _always_ loses, at least in the games I have watched on ICC... I'd say that the sac works maybe 1 of 10 times. This one is different, in that it does appear to tactically win, which takes it out of the discussion on 'sacrifices' since it stops being a sac the instant you see the tactical win of material. > >It's not the way I like to program Rebel. I want the thing to play chess >as humans tend to play it and positional sacs are part of the game. > >I am not saying Rebel is great in this respect playing the all the great >sacs in human chess history but I definitely give it a try. My experiences >with positional sacs are not bad at all on the contrary I would say. It's >all doable if you can define a positional sac to be right in 9 of 10 times. >From there search does a great job in filtering the exceptions. And the >deeper you go the better the filtering works. here's a good test... how did you do on Howards Bxh6 test positions? I ran them at very shallow searches, and didn't do the sac one time, which means I was right 4 of 9 times... It would be interesting to see how you do at very shallow _max=4 ply say) search depths... > >Positional knowledge and search go hand in hand, the one can not >without the other. > >Ed Of course... I have just made this mistake _so_ many times on the chess servers that I decided to add special code to make sure I don't sac a bishop for 2 or 3 pawns unless it isn't a real sac at all... Because as I fiddled with king safety over the years on ICC, I had several versions that were wildly aggressive like that... (and some that weren't so wildly aggressive). And I found myself in trouble more often than not... I want to do whatever it takes to be right _most_ of the time. Because against a fast searcher like Fritz, et al, you had _better_ be right in such sacs or you simply lose... because a bishop for two pawns doesn't leave _any_ room for error...
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.