Author: Stuart Cracraft
Date: 07:48:11 08/15/04
Go up one level in this thread
On August 14, 2004 at 12:19:08, Robin Smith wrote: >On August 11, 2004 at 17:30:29, Stuart Cracraft wrote: > >>>On August 09, 2004 at 14:26:18, Stuart Cracraft wrote: >>> >>>>Sorry, the posted position was in error. It is: >>>> >>>>>[D]r1bqkb1r/2pp1ppp/p1n2n2/1p2p3/4P3/1B3N2/PPPP1PPP/RNBQ1RK1 b q - 4 1 >>>> >>> >>Still, even with castling incapable on the king's side, I'm looking for >>a non-book, non-search method to avoid 1. ... Na5. > >Hi Stuart, > >I think the main reason Na5 is bad involves search ... it is not that Black >gives up a center pawn, since without a deep search programs will think Black >can get the White e4 center pawn in return with 1... Na5 2.Nxe5 Nxb3 3.axb3 >Nxe4. The reason 3...Nxe4 does not work is because Black's king is uncastled, on >the now opening e-file. So if you want to fix this without search, add a bigger >penalty for poor development when the king is uncastled. Black is using an extra >tempo to exchange knight for bishop, and that is the most relevant reason why >(without search) it should be avoided. > >Robin Hi, I played around with the evaluation yesterday and after some refinements, Na5 is avoided. In the search below, note the high score is in millipawns. This is because white is castled with protection and black has no real protection. I use millipawns after observing Bob's struggle through the years with centipawn vs. millipawn for the same reason: I can give substantive bonuses and penalties without worry that material will be compromised. The main change I made was a large bonus for a king with pawn protection after the immediate opening. This fixed the whole Na5 issue. My previous king safety term had been disabled. r1bqkb1r/2pp1ppp/p1n2n2/1p2p3/4P3/1B3N2/PPPP1PPP/RNBQ1RK1 b - - d6; id "NA5BAD" (id "NA5BAD" ) BR ** BB BQ BK BB -- BR ** -- BP BP ** BP BP BP BP ** BN ** -- BN -- ** ** BP ** -- BP -- ** -- -- ** -- ** WP ** -- ** ** WB ** -- ** WN ** -- WP WP WP WP -- WP WP WP WR WN WB WQ ** WR WK -- stage 0, black to move, computer plays black hash=4dcc20f7be0544f5 NULLMV TT(524288) Alpha=-425 Beta=75 Maxdepth=9999999 MaxTime=99 1/ 2> f6e4 0.00 75 3 f6e4 1/ 2 f6e4 0.01 876 64 f6e4 2/12 f6e4 0.01 858 477 f6e4 f1e1 3/14 f6e4 0.02 858 2546 f6e4 b1c3 f7f5 4/16< f6e4 0.08 608 11634 f6e4 b1c3 e4c3 d2c3 4/16 d7d5 0.13 -154 18148 d7d5 e4d5 f6d5 b3d5 5/21 f8c5 0.35 -151 62955 f8c5 d2d3 e8g8 b3d5 f6d5 6/23 f8c5 0.88 -167 162217 f8c5 d2d3 e8g8 b1c3 f8e8 f1e1 7/24 f8c5 2.24 -157 420719 f8c5 d2d3 d7d5 b1c3 d5d4 b3d5 c8b7 8/37 f8c5 8.79 -167 1710075 f8c5 b3d5 f6d5 e4d5 c6b4 f1e1 d7d6 d2d4 9/38 f8c5 45.82 -160 4535115 f8c5 b3d5 c8b7 c2c3 e8g8 10/41 f8c5 95.33 -169 14133000 f8c5 b3d5 c8b7 c2c3 f6d5 e4d5 c6e7 d2d4 f8c5 95.33 -169 14133000 f8c5 b3d5 c8b7 c2c3 f6d5 e4d5 c6e7 d2d4 nps=148252 ha=20.00% q=56.0% bc=70% br=NaN% mp=331<>144 pawnx=0 recapx=0 qcheckx=197782 checkx=0 futilx=0 qfutilx=0
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.