Author: Robert Hyatt
Date: 16:03:26 01/11/03
Go up one level in this thread
On January 11, 2003 at 15:24:24, Kolss wrote: >On January 11, 2003 at 14:01:28, Robert Hyatt wrote: > >>On January 11, 2003 at 12:46:35, Kolss wrote: >> >>>On January 11, 2003 at 11:14:10, Robert Hyatt wrote: >>> >>>>On January 11, 2003 at 07:43:24, Drexel,Michael wrote: >>>> >>>>>[D] 4nnk1/8/6b1/4BP2/2N1B1K1/8/8/8 b - - 0 1 >>>>> >>>>>I just came over this endgame position with Crafty 19.01 as analysis engine in >>>>>CB-GUI. Crafty suggested 1...Bxf5+ with a score of 0.01. >>>>>1...Bf7 seems to be a simple draw. >>>>>you probably told him 3 pieces against 2 pieces without pawns is draw, but I >>>>>doubt this is a good idea in the endgame BBNK vs. NNK. You might be better >>>>>informed but I would think this endgame is almost always a win and would even >>>>>assume that it is almost always a win within 50 moves. >>>>>If I were a tablebase-expert like you, I would fix it immediatly, whether it is >>>>>important or not. >>>>> >>>>> >>>>>In games against Fritz 8 Crafty will resign in not more than 8 moves: >>>>> >>>>>[Event "ATHLONXP"] >>>>>[Site "ATHLONXP"] >>>>>[Date "2003.01.5"] >>>>>[Round "1"] >>>>>[White "Fritz 8"] >>>>>[Black "Crafty 19.01"] >>>>>[Result "1-0"] >>>>>[SetUp "1"] >>>>>[FEN "4nnk1/8/6b1/4BP2/2N1B1K1/8/8/8 b - - 0 1"] >>>>>[PlyCount "14"] >>>>> >>>>>1... Bxf5+ 2. Bxf5 Ng7 3. Be4 Nge6 4. Kf5 Kf7 5. Bd5 Ke7 6. Bd6+ Kd7 7. Ke5 >>>>>Ng6+ 8. Kf6 1-0 >>>> >>>>First, Crafty well knows that KBB vs KN is a win. But it isn't seeing that >>>>here until way deep into the search. In 99.999999999999% of these endgames >>>>with 3 minors vs 2 minors, the game is a draw. Handling that one special case >>>>here and there is not going to make it play better, and would almost certainly >>>>hurt it overall as it really should say KBBN vs KNN is a draw in almost all >>>>cases. >>> >>>I do not agree. IMHO, the stronger side should win in most cases. Especially if >>>you score all positions as 0.01, you will allow the opponent to make progress, >>>drive your king into a corner and seperate the knights. >>> >>>Regards, Munjong. >> >> >>Do not agree with what? That king + 3 minors vs king + 2 minors is almost >>always a draw? or that KBBN vs KNN is mostly drawn? I'd be willing to wager >>that _most_ of the above positions are dead draws. > >Of course, I accept and respect your opinion in this regard. Nonetheless, allow >me do differ: > >I suppose that most positions KBBN vs. KNN are won for the stronger side. In >fact, I would not be surprised if in a couple of years, a properly working chess >program which uses endgame databases (TB) announced forced checkmate for white >after 1. ... Bxf5??. I wouldn't disagree there, but I do not believe that "most" are winnable. A single piece is not enough. IE KQ vs KQN is not going to be winnable. KRB vs KR is not winnable. A single piece is generally not enough, except in rare cases. The two knights can cause real problems in trying to trade the white knight for the black knight. And I still suspect that most are going to be drawn. > >I can probably agree that king + 3 minors vs. king + 2 minors is mostly a draw >IF the stronger side is not up more than one minor in material, whereas the >bishop is scored higher than the knight (B > N), such that 2B - N > B = highest >value for minor. In fact, I believe that this is a pretty good rule for most >pawnless endings. E.g. RB vs. R is very drawish, QB vs. Q or QN vs. Q are >extremely drawish, RBB vs. RN offers good winning chances for the stronger side. > >This is of course only my own, subjective view of the matter. > >>Note that I don't score them as "drawn". I score them as "side with most >>material can't win" which is different... > >The end result is very similar. If you score all such positions with at most >0.01 (actually, Crafty 19.01 always says 0.00 on my machine - please pardon me >if I do not have the most recent version to check) for the stronger side, your >program will most likely lose to any program which tries to make progress, even >if yours outsearches the other program by 4 (number picked at random) plies. > > >Regards, Munjong. that probably needs looking at for some cases...
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.