Author: Tony Werten
Date: 23:03:09 09/29/05
Go up one level in this thread
On September 29, 2005 at 15:38:30, Robert Hyatt wrote: >On September 29, 2005 at 12:54:27, Uri Blass wrote: > >>On September 29, 2005 at 12:29:45, Robert Hyatt wrote: >> >>>On September 29, 2005 at 09:26:52, Uri Blass wrote: >>> >>>>On September 29, 2005 at 01:29:07, K. Burcham wrote: >>>> >>>>> >>>>>Thanks for your reply Robert. Always like reading your comments about Deep Blue. >>>>>Of course we both know that the h5 move in game five has been discussed many >>>>>times. Here is Fruit 2.1. >>>>>John if you read this, what does chessmaster do with default settings in this >>>>>position? >>>>> >>>>>Fruit 2.1 >>>>>432 hash >>>>>Athlon 64 >>>>>764 pin >>>>>2200 mhz >>>>> >>>>>h5 in five seconds. >>>>> >>>>> [D] r2qk2r/pp3ppp/2p1pn2/4n3/1b6/3P2PP/PPPN1PB1/R1BQK2R b KQkq - 4 1 >>>>> >>>>>00:00:00.0 0.71 5 14645 O-O f4 Ng6 O-O Nd5 >>>>>00:00:00.1 0.60 6 59452 O-O a3 Bc5 Nb3 Qb6 O-O >>>>>00:00:00.2 0.27 7 200589 O-O O-O Qb6 Nf3 Bd6 Re1 Ng6 >>>>>00:00:00.4 0.24 8 407698 O-O O-O Qb6 Re1 Bd6 Ne4 Nxe4 Bxe4 f5 >>>>>00:00:01.1 0.18 9 1011530 O-O O-O Nd5 Re1 Qc7 d4 Ng6 c4 Nf6 >>>>>00:00:02.5 0.04 10 2475965 O-O O-O Nd5 Re1 Qc7 Qe2 Ng6 c3 Bc5 Nc4 >>>>>00:00:04.2 0.11 10 4088578 h5 a3 Bd6 Ne4 Nxe4 Bxe4 f5 Bg2 Bc5 Bd2 Kf7 >>>>>00:00:07.8 0.11 11 7518463 h5 Qe2 Bd6 f4 Ng6 Nc4 h4 f5 Bxg3+ Kd1 Nf8 Bd2 >>>>>00:00:19.3 0.00 12 18830729 h5 O-O h4 d4 Qxd4 c3 Qb6 cxb4 hxg3 Qb3 gxf2+ Rxf2 >>>>>Qd4 b5 >>>>>00:00:41.3 -0.05 13 40615296 h5 O-O h4 d4 Qxd4 c3 Qb6 cxb4 hxg3 Qe2 Ned7 Nc4 >>>>>gxf2+ Rxf2 Qxb4 >>>>>00:02:17.0 0.00 14 133102696 h5 Qe2 Qc7 O-O h4 Ne4 Be7 Be3 Nd5 d4 Nxe3 fxe3 Nd7 >>>>>Qf2 O-O gxh4 >>>>>00:05:17.0 0.08 15 307648549 h5 Qe2 Qc7 c3 Be7 Ne4 h4 Nxf6+ Bxf6 g4 Ng6 Be4 Nf4 >>>>>Bxf4 Qxf4 d4 O-O O-O >>>>>00:10:54.5 0.07 16 633368278 h5 Qe2 Qc7 c3 Be7 Ne4 h4 Nxf6+ Bxf6 g4 O-O-O d4 Ng6 >>>>>Be4 Nf4 Bxf4 Qxf4 Qe3 Bg5 >>>>>00:16:44.6 -0.02 17 977058195 h5 Qe2 Qc7 c3 Be7 h4 O-O-O d4 Neg4 Nc4 Nd5 O-O Bf6 >>>>>f3 Nh6 Bxh6 Rxh6 f4 >>>>>00:42:35.1 -0.02 18 -1805878841 h5 Qe2 Qc7 c3 Be7 h4 O-O-O d4 Neg4 Nc4 Nd5 O-O >>>>>Bf6 a4 Kb8 Re1 Be7 Be4 >>>>>01:17:57.7 -0.06 19 267945229 h5 Qe2 Qc7 c3 Be7 h4 O-O-O d4 Neg4 Nc4 Nd5 O-O Bf6 >>>>>Re1 g6 Bf3 Kb8 Bd2 c5 Bxg4 hxg4 dxc5 Qxc5 Qxg4 >>>>> >>>>>Rematch >>>>>Place: New York, NY USA >>>>>Round: 5 >>>>>Date: 10/5/1997 >>>>>White: Kasparov, Garry >>>>>Black: Deep Blue >>>>> >>>>>Kasparov's comments about 11...h5 >>>>>Kasparov said no computer can play 11...h5 without human interference. >>>> >>>>Nonsense >>>> >>>>11...h5 is very logical for computer to play. >>>>It increase the mobility of the rook at h8 so there is no reason for computers >>>>not to play it. >>> >>>Sorry, but I _totally_ disagree there. Most any decent computer chess program >>>has an evaluation component that covers the king-side pawn structure. Playing >>>h5 creates a pawn target for black to aim at. It weakens g5 and g6. So a >>>computer has to see something _else_ to more than offset that king safety issue. >> >> >>I know for sure that mobility of the rook based on number of moves is one of the >>component of Fruit's evaluation. >> >>There are other components but the possibility that mobility of the rook is more >>important than them is not illogical > >If making a _major_ decision like playing h5 is based completely on adding two >vertical squares for the rook to reach, then that seems wrong. Mobility is >fine, but there is "mobility" and there is _MOBILITY_. The latter is really a >case of useful mobility in multiple directions... > >The idea may be just fine. But again, it is easily possible to play the right >move for the wrong reason. If mobility is the only thing that makes "h5" jump >to the top of the search, then that is simply broken reasoning... Otherwise you >would be seeing h5 pop up all the time and as often as not, it is a bad move. >Every pawn move is inherently bad in fact, and making one over there has to have >a point to offset the loss of control at g6/g5 at least... > >> >> >>> >>>I suspect _many_ strong computers will _not_ play this move, for the reason >>>given above, since any long-term gain by the move is way beyond any search >>>horizon, and is going to be purely based on positional scoring numbers.. >>> >>>Mobility of the rook is a lousy reason for h5. Otherwise black would always be >>>playing a5 and h5 to give both rooks more mobility... >> >>mobility of knight and bishop is more important in the opening so Fruit prefers >>to develop them in the opening but it certainly gives a bonus for having more >>moves with the rook. >> >>Here is part of the code of fruit2.1. >> >>It gives bonus of 0.02 pawns for every possible rook move in the opening(it is >>practically average between 0.02 and 0.04 based on the stage of the game) when >>it gives bonus of 0.05 pawns for every bishop moves and 0.04 pawns for every >>knight move >> >>Mobility of the rook is not the only reason that push fruit to choose h5 but it >>is probably one of them. >> >>static const int KnightMobOpening = 4; >>static const int KnightMobEndgame = 4; >>static const int BishopMobOpening = 5; >>static const int BishopMobEndgame = 5; >>static const int RookMobOpening = 2; >>static const int RookMobEndgame = 4; >>static const int QueenMobOpening = 1; >>static const int QueenMobEndgame = 2; >>static const int KingMobOpening = 0; >>static const int KingMobEndgame = 0; >> >>Uri > > >there is nothing wrong with any of those, just so the "RookMobOpening" is not >able to force moves like h5 for no other reason. It should be compensated by a king safety minus, but its main line shows castling queenside, so the minus falls away. Making h5 a perfectly sensible move. I doubt there are engines playing h5 without castling queenside. (don't know if that happened in the game) Tony
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.