changelog

一款运行在linux上的象棋游戏。用GTK/GNOME环境下用GLADE开发。

     pgn.c     - have PGNSaveGame record player's names based on "name" command
     common.h  - adjust knight and bishop values
     iterate.c - was not taking up to 2 x nominal time per move when in
		 bad situations. fixed.
     book.c    - for random choice amongst top N moves of preferred book
		 now sorts all moves by stored book score and picks
		 randomly amongst N based. Previous algorithm was flawed.
     eval.c - add penalties for attacks against ones king in
            - add OPENGAME heuristic (1/4 pawn reward for positions
	      in which root position either side has more than 6 pawns
	      and in the leaf position both sides have <= 6
	      (idea due to Larry Kaufman and Don Dailey of Socrates)
	    - lots of work on development scoring.
     - Win-at-Chess test: 233/300 at 5 seconds per move. 14,940,780 nodes.
     - Rating went to FICS 2050 with all the positional changes on 1/31/99.
     - make king attacks per piece to reward multiple attacks to
       same square by multiple pieces
     - anti-stonewall
     - credit protectors in front of king and penalize attackers.
     - penalize repeat pawn moves before development complete.
     - rating maintained at FICS 2049.

Version 5.00d (late February)
     - cleaned up version prior to all the mods to get rating back.
     - one night result:
	Games: 		29
	Ave.opp:	1616
	Percent:	89
	Result:		25W + 2D + 2L = 89 %
	Rating:		1967
     - fix up PGN save-to-file routine. Was putting black in result.
     - ScoreDev returns only if fully developed.
     - Diminish connected passed pawn penalty.
	Games: 		104
	Ave.opp:	1797
	Percent:	77
	Result:		80W + 2D + 22L = 77 %
	Rating:		2008

Version 5.00e (beginning of March)
     - try increasing pawncover values -- no good.
     - try diminishing underattack values -- seems good.
     - add pawn majority/minority calculation
     - add in better connected passed pawn evaluator
     - double penalty for open files adjacent to or on king's file if 
       enemy queen present.
	Games: 		91
	Ave.opp:	1763
	Percent:	87
	Result:		79W + 2D + 10L = 87 %
	Rating:		2085

Version 5.00f
     - BUGFIX: Kong provided a fix to prevent segmentation faults in
       certain rare positions:
		6Q1/8/8/4K3/8/8/p7/1k1 w - - bm 1; id 1;
     - based on analysis of 7 lost games last night it was
       noted that 6 of these were in positions where the program
       played other than double king-pawn or double queen-pawn
       instead opting for non-classical openings, so...
       fix to always play double-pawn as long as the evaluation resolves
       to that order of moves once book has been scored.
       this is done by having the program choose from between the 
       top 2 moves for white's first move (excluding Nf3, but including
       e4 and d4) and only the top 1 move for black's first move
       (excluding non-kp/qp replies to KP/QP by white). For odd
       openings like Reti or Nc3, fairly good moves are picked with
       the above algorithm, that may be preferable to just having
       only kp/kp or qp/qp for this algorithm).
     - BUGFIX: fix problem that resulted in friendly king not being
       penalized enough for being far from enemy passed pawns.
     - FEATURE: Add DistMap[m][n] array. This array has all
       bits turned on that are n squares away from the m square where
       m is A1..H8. Aids in rapid diagnosis of distances.
     - FEATURE: Add penalty for every enemy pawn within 3 squares
       of the friendly king in the opening and middlegame: KINGNEARENEMYPAWN.
     - FEATURE: Add penalty for no friendly minors or queen within
       3 squares of king in opening or middlegame: KINGNOSUPPORT.
	Games: 		27
	Ave.opp:	1631
	Percent:	96
	Result:		26W + 0D + 1L = 96 %
	Rating:		2132

Version 5.00f
     - Kong contributed rotated bitboards move generator.
       It tests 21% faster than the old move generator,
       yet overall program speed is about 8% slower based
       on testing.
     - Added ROOKTRAPPED code for rook trapped by king.

Version 5.00g
     - Add Crafty-style bad-trade code for bad piece exchange penalty.

Version 5.00h
     - book.c: when in XBOARD mode, the book needs to be quiet without status.
     - eval.c: some formatting
     - Win-at-Chess test: 219/300 at 5 seconds per move. 11,864,883 nodes.
			  1441.7 seconds (out of 1500 permitted).
     - eval.c: penalize knight on rim even more.
               up undevelopment penalty.

Version 5.00i
     - use GNU 4.0 eval instead. toss out evaluation.
     - Win-at-Chess test: 229/300 at 5 seconds per move. 10,949,825 nodes
	1507 seconds
     - Add Queen+Piece threat extension from Kong.
     - WAC: 235/300 at 5 seconds per move. 9,656,161 nodes. 1525 seconds.
     - Add the whole GNU 4.0 style calling sequence in Evaluate (lazy,
       single lone king, etc.)
     - WAC: 213/300 at 5 seconds per move, 17,911,670 nodes. 1437 seconds.
     - FICS result:
	Games:          34
	Ave.opp:        1836
	Percent:        85
	Result:         28W + 4D + 2L = 85 %
	Rating:         2132

Version 5.00j
     - integrated 1.21k from Kong.
     - added 
	traditional anti-computer locked pawn structure penalty
	traditional anti-computer stonewall pawn structure penalty
	crafty's eight pawns penalty
	old cobalt's ahopen traditional anti-computer edge-of-board attack
	center pawns
	shortened time until queen penalty (4 moves instead of 8)
     EICS 5/31/99
	Games:          83
	Ave.opp:        1882
	Percent:        77
	Result:         63W + 4D + 16L = 77 %
	Rating:         2093
     FICS 6/15/99
 	Games:          125
	Ave.opp:        1766
	Percent:        81
	Result:         100W + 10D + 15L = 81 %
	Rating:         2017

Version 5.00 *** RELEASE ***		(September 30, 1999 - cracraft)

o Found that after first ^C interrupt to running program a second one would
  exit it. Added a new signal (..) call in EndSearch. Okay now.

o Structure the tree (doc, book, src, etc.)

o Rename the tree from gnuchess5.0 to chess-5.00

o Revise version.h.

o Revised copywrite.

o Revised COPYING (update FSF address).

o Default to post mode (statistics) if not xboard.

Version 5.01

  Sun Dec 12 11:42:10 EST 1999 - src/version.c
    Patrick Gundlach <[e-mail]>
    fflush stdout in Xboard startup (i.e. "Chess" herald)

  Sun Mar  4 11:02:52 EST 2001 - book/book.*
    Stuart Cracraft <[e-mail]>
    Expanded opening book

Version 5.02

  Various bug fixes.

Version 5.03

  Book handling rewritten, and other major changes collected and applied
  by Stuart Cracraft before use of CVS. Book need to be remade from PGN.

  List of Grandmasters we "trust" in generating the opening book was
  extended.

  Contributed Bug fixes - various.

  Code placed under CVS after some minor changes to 5.02.
  Opening book not placed under CVS - too large.

  Other changes managed under CVS

  Installation instructions corrected "book create" to "book add".
  Makefile modified to work with OpenBSD make.
  Use of C++ reserved words as variable names removed for ease of reuse. 
  The C Macro i386 replaced with Endianness test for portability.
  Lots of unreachable code removed to make it more readable.
  "#ifdef debug" and "#ifdef never" code removed.
  Removed problem with "space" on the end of filenames.
  Handle XBoard SIGINT gracefully (Caused Solaris Xboard problem).
  Reduced default hashtable size.
  Deleted Crafty style bad trade code as it was incorrect.

  The calculations for hung and pinned pieces were using stale
  information in the calculations. All relevant code was removed
  from the position evaluation routine. 
  
  Fixed loop bound error that allowed King to remain in check!!!

  Made input buffers consistent to stop crash on long commands.
  Correct hashtable handling of En Passant.
  Fixed EPD output when loading a game from file with Black to move.
  Fixed file handling problem extending opening book.

  Various modifications to SortMove to encourage examination of
  pushing passed pawns, centralisation, before other moves. Some
  modification have little effect, as the calling of sorting routines
  is quite complex.

  Added various EPD files for developer testing. These files cannot be
  included in the released tar file until copyright issues are resolved.

  Removed Outside passed pawn bonus from Evaluate as it was
  incorrectly in the pawn hash section.

  Removed the chess_faq.html file from the distribution, as this
  belongs on the GNU web site.