ChangeLog

2000-12-25  Alejandro Dubrovsky  <s328940@student.uq.edu.au>

	* unches: initial version.

2001-02-07

    changed to version 0.1.1
    
    ai.c: fixed evaluation of best value when queening pawns
	introduced heap sorting for all moves at first level ordered by evaluation
	
2001-02-08
    ai.c:  lots of bugfixing of heap
    utils.c:  added printBoard and pieceToChar
    ai.c:  added board printing if setboard was used
	    fixed bug where new wouldn't play
	    made finer grained time cuttoffs possible
    board.c:  fixed bug where castling was possible when rooks weren't there
    created hash.c
    
2001-02-09
    ai.c:   output error for non-implemented features
    
2001-02-12
    played on fics for a while.  much better than expected.
    utils.c:  parseResult stupidity fixed
    ai.c: print result if it is passed from the server
	changed relativeheuri to positionheuri
	replaced relativeheuri by a white / black value (normalised on log scale)
    board.c:  stupidity where if one person moved their rook, the opponent couldn't castle
    ai.c:  added continuous reseeking for fics
    left playing on fics overnight (lightning c 1850, blitz c 1550)
    
2001-02-13
    utils.c:  added parseRatings
	    added printing ratings and time controls to the game header
    board.c:  fixed bug where king next to king moves were allowed in some cases

2001-02-14
    changed to version 0.1.2
    
2001-02-15
    unches.c:   changed the relative flag to be a modifier and added position flag
    ai.c:	added a position and a relative position evaluation functions.
    
2001-02-16
    ai.c:	changed the timing so that increments account for the no-increment-on-the-last-
	    move effect
    unches.c:	added option for hash size
    hash.c:	wrote initialiseHashTable, clearHashTable and getHashKey
    board.c:	modified movePiece to modify the new bitboard (used in hashing)
    ai.c:	modified iterativedeepening to insert hash values and newalphabeta to retrieve
	added hard time limit for searches

2001-02-17
    board.c:	fixed bug where board.fifty wasn't being reset to 0 so any game with over 100
	non-pawn and nontake moves was considered internally as a draw
	Added movePieceAndHash which also adds the current hash value to the board.history
    ai.c:	changed iterativedeepening and newalphabeta use movePieceAndHash
    board.c:	Added isRepetitionDraw.

2001-03-11
    board.c:   Fixed isRepetitionDraw.
    ai.c:	added check for repetition draw
    
2001-03-13
    cmdline.conf.in:	Increased default hash size to about 2 million
			Changed verbosity level to be an integer
    ai.c:		debugging info sprinkled
	changed all position values back to default
	increased range of returned values by the relative evaluators by ten

2001-03-14
    ai.c:	reintroduced random variance
		when tree is pruned return cut values instead of current best to avoid
		bad moves getting in through random variance
		Play forced moves immediately
    Changed version to 0.1.3

2001-03-17
    hash.c:	changed hashing scheme to Zobrist hashing
    board.c:	changed movePieceAndHash to make use of the possible speedup of Zobrist hashing

2001-03-23
    created procvalues.py
    board.c:	Added loading value matrices from file
    
2001-03-25
    ai.c:  changed popHeap and insertIntoHeap into negamax (always maximise)
	    changed iterativeDeepening and newalphabeta to negamax
	    
2001-03-26
    ai.c:	added basic negamax for comparison        
    hash.c:	split hash tables into white and black hashtables
    ai.c:	according changes to newalphabeta and iterativedeepening
    hash.c:	add storage of best move in hash table

2001-03-27
    ai.c:	use hashed best suggested move and beta cut off
		store best move and beta cut off
		add debugging for hashed events

    
2001-03-28
    ai.c:	fix hash initialisation for Zobrist hashing
		use suggested move even if betacut

2001-03-31
    split ai.c into ai.c, search.c and eval.c
    created search.h and eval.h
    added include files to Makefile.am

somewhere in between:
    added storing of alpha cut off, pruning on hashed alpha cut off, 
    using hashed alpha and beta values, storing only on worthless information.

2001-04-15
    search.c:	fixed the extremely annoying hash bug (b->blackturn not attempt.blackturn)
		fixed random to adapt to beta cutoffs returns
		
2001-04-16
    Changed version to 0.1.4
    Added compname command line argument to choose engine's name
    Changed variable passed to io function to be the engineParameters
    
2001-04-17
    board.c:  Fixed a repetition check bug
        
hmmmm....long break

Mon Oct 01 01:09:49 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	search.c:	Shortened code used for trying each piece when coronating.
	Added libale as a subdirectory.  Made it a compilable separate GNU tools style
	package.
	
Mon Oct 01 03:57:04 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	Bumped version to 0.1.5
	
Mon Oct 01 08:44:50 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	A bit of documentation all over the place
	search.c:  Hopefully fixed the hashtable loading and storing. (maybe i broke it though)
					Added more debugging
	Increased default hashsize to about 4 million
						
Mon Oct 01 09:33:19 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	Bumped version to 0.2.0
	Split search.c into basicsearch.c, alphabetasearch.c and mtdf.c
	
Tue Oct 02 05:06:59 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	alphabetasearch.c:  Fixed flags returned, to work properly with random variations.
	
Tue Oct 02 07:49:01 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	mtdf.c:  Finished first pass at a working iterativemtdf, mtdf, and unrolled alpha-beta.
	Made mtdf default engine.  Added flag for old alphabeta.
	
Tue Oct 02 09:41:04 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	mtdf.c:	Got rid of move passing optimisation, since it can't handle stored moves
				easily
				
Tue Oct 02 11:46:40 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	mtdf.c:  Reinstated move-passing-as-int, and handled finding index of new moves.
				Now check previous best move first.
	mtdf.c:	Modified when the hash table is overwritten
	alphabetasearch.c:	Same
						
Wed Oct 03 06:30:05 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	Bumped version to 0.2.1
	
Wed Oct 03 10:07:37 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	mtdf.c:  Added verynewalphabeta.  Does internal iterative deepening.  Templated on
				newalphabeta.  MTDf uses it now.
	mtdf.c:	Use exact hash values better.  Fix bug where flags weren't set properly if
				the hash value was not going to be overwritten.
	alphabetasearch.c:	Same with newalphabeta.
	
Wed Oct 03 13:14:44 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	board.c:	Fixed ancient bug in en-pasant hashing in movePieceAndHash!
	
Wed Oct 03 13:25:22 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	board.c: Another bug in movePieceAndHash, bitboard setting when castling.
	
Wed Oct 03 18:57:04 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	board.c: Added generateMoveList to generate all legal moves at a node
	mtdf.c:  verynewalphabeta now uses generateMoveList
				unrolledalphabeta now skipped. Unnecessary. So jump straight to
				verynewalphabeta.  This caused a massive decrease in nodes searched
				and in some very obvious bugs. 
	
Tue Oct 09 21:05:52 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	mtdf.c:  Fixed obvious bug in hash-value storage.  
				got rid of unrolledalphabeta (not needed)
				
	Bumped version to 0.2.2

Tue Oct 09 21:24:52 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	alphabetasearch.c:  Fixed the same obvious bug in hash-value storage

Thu Oct 11 15:55:04 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>				
	all:  Changed GPL wording to just version 2.
	mtdf.c:	got rid of allmoves storage and all associated stuff in iterativemtdf
	
Thu Oct 11 16:43:22 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	Recreated search.c for common search functions.

Thu Oct 11 20:22:29 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	search.c: First implementation of quiescentsearch.
	mtdf.c:  make verynewalphabeta use quiescentsearch.
	
Thu Oct 11 21:02:44 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	mtdf.c:  Fixed stupid bug when calling quiescentsearch.
	mtdf.c:	Fixed initial best to MINVALUE - 1 in iterativemtdf and verynewalphabeta
	alphabetasearch.c:  Same in newalphabeta and iterativedeepening
	search.c:  Same in quiescentsearch.
	alphabetasearch.c:  Make newalphabeta use quiescentsearch.
	
Fri Oct 12 14:17:49 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	board.c:  Added generateQuiescentMoveList
	split common.h into ai.h, pieces.h, piecesstruct.h, communication.h, hash.h,
	hashstruct.h and board.h.
	search.c: changed name of functions to capitalise wordBeginnings
	alphabetasearch.c:  Same
	mtdf: Same
	basicsearch: same

Fri Oct 12 14:47:58 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	search.c:  Make quiescentsearch use generateQuiescentMoveList
	mtdf.c:  Make iterativemtdf use generateMoveList
				Make verynewalphabeta skip the 0th branch and jump straight to quiescent
					
Fri Oct 12 16:20:32 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	added heap.h and heap.c for the heapsort handling
	search.c:  Added orderMoveList (to order a list of moves, does not much for the moment)
	mtdf.c:  Simplified verynewalphabeta by using orderMoveList
	search.c:  Same as above with quiescentsearch
	Too many changes.  Buggy.

Fri Oct 12 17:10:52 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	search.c:  Fixed silly bug in orderMoveList where it always pointed to first move.
	
Fri Oct 12 21:17:10 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	alphabetasearch.c:  Fixed bug in hash storing (yet again).
	mtdf.c:  Same
	search.c:  Same
	search.c:  Fixed logic of quiescent search.
	That seemed to do the trick.  Consistently getting to depth 6 with standard weights
	(4-5 with more complicated evaluation tables)
	
Fri Oct 12 22:51:26 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	search.c:  Added the possibility of a bit of randomness through move ordering
	mtdf.c:  Added randomness if in parameters on iterativemtdf

Sat Oct 13 19:47:50 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	Bumped version to 0.2.3
	Moved orderMoveList from search.c to eval.c

Sun Oct 14 14:52:34 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	board.c: Tiny optimisation to inCheck		
	mtdf.c:  Fix nasty bug in verynewalphabeta where the checkmate move wasn't being returned.
	search.c:	Strip checkmate and draw checks from quiescentsearch.
	
Mon Oct 15 16:09:02 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	ai.c, mtdf.c:  Added hash, evals, node and quiescent node count and reporting
	
Mon Oct 15 17:37:49 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	eval.c:  Added move ordering based on piece captured and capturer.
	common.h:  Added history tables to engineParameter structure.
	mtdf.c:  Added storing of beta cuttoffs in history.
	search.c: same as above
	eval.c:  Use history tables for moveordering.
	mtdf.c:  Fix checkmates to prefer earlier if ours, and later if theirs (not the other
				way around)
				
	
Fri Oct 19 05:37:53 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	mtdf.c:  Added better time management to mtdf. (max of 2 times supposed time)
	
Fri Oct 19 06:35:29 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	mtdf.c:  Added timecheck in verynewalphabeta (seems needed)
	
Fri Oct 19 23:21:15 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	utils.c:  Added moveToSan to write moves in SAN format.
	ai.c:  Changed move output to use moveToSan.
	mtdf.c:  Same
	utils.c:  Added printPGNHeader.
	ai.c:  Use PGN header.  Mostly PGN compliant output.
	
Sat Oct 20 20:33:18 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	Bumped version to 0.2.4
	ai.c: Stupid bug in release (crash on startup when not recording)
	
Sun Oct 21 00:54:50 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	ai.c:  Changed getting current move from parameters->movenumber to board->ply to
			fix some problems with adjourned games.
	alphabetasearch.c:  Same
	mtdf.c:	Same
	
Mon Oct 22 03:44:37 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	doIO.c:  Added new unches specific interface command to convert from traditional move
				to SAN
	communication.h:  Corresponding changes
	ai.c:	Same
	Converted all played games to PGN!

Tue Oct 30 04:39:57 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	eval.c:  Added randomisePieceValues and added an original copy of piece values.
				Randomising scheme changed so that the piece values are slightly randomised.
				Got rid of random character of orderMoveList
	ai.c:		Call randomisePieceValues at beginning of each game

Tue Oct 30 05:02:49 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	mtdf.c:	Modify checkmate score to hopefully prevent checkmate shakes
	
Wed Oct 31 02:13:06 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	board.c:  Modified hashing to get rid of the bitboard keeping, and switch all 
				random numbers and the key to 64-bit numbers.  Changed from keeping tabs of
				king and rooks moved, to queen and king castle availability.
				Added modification of hash depending on castling availability.
				This has the sideeffect of making the hash size much more flexible (no
				need for it to be a power of two any more)
	boardstruct.h:  same
	hash.c:	same
	hashstruct.h:	same
	alphabetasearch.h:	same
	utils.c:	same
	mtdf.c:	same
	search.c:	same

Thu Nov 01 14:53:09 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	mtdf.c:	No need to order moves on first iteration.  Don't store result in
				hash table.
	eval.c:	Make sure both wvalue and bvalue are positive before dividing in
				realtivepositionheuri
	hash.c:	Changed creation of 64 bit random numbers to use all 64 bits (random() returns
				31 random bits, not 32)

Fri Nov 02 20:39:24 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	mtdf.c:	check if out of time after reduced depth search
	
Fri Nov 02 22:12:25 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	search.c:  Think i fixed the instability bug that has been causing stupid moves
				to be played. (it wasn't setting best to the current in quiescentsearch
				if it wasn't better than alpha) (Note:  False alarm, instability still there
				almost certainly caused by hashing)
	mtdf.c:	turned off reduced search depth for a while.  not sure it's any good
	
Mon Nov 05 02:08:13 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	pieces.c:	Fixed long standing bug where the king could castle into an opponent's 
					piece.
	utils.c:		Fixed notation on en-passant in moveToSan.

Sat Nov 17 01:15:07 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	board.c:		Add pawn captures to generateQuiescentSearch.
	bumped version to 0.2.5
	
Sat Nov 17 02:53:42 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	unches.c:	Added valuefile used to default name
	Made my own position values.  Jumped to 2080 in lightning, 1850 in blitz on FICS
	
Sun Nov 25 05:37:34 EST 2001 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	utils.c:		Fixed printResult to take proper care of undecided games.

Sun Feb 03 06:35:24 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	utils.c, doIO.c:  Fixed 'deprecated use of label at end of compound statement' warning
	from gcc
		
Sun Feb 17 00:11:17 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	Since it's been so long:
	bump version to 0.3.0
	Changed name to 'Small Potato'
	
Sun Feb 17 02:49:48 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	Wrote some benchmarks.
	Around 1.4 million move/unmove cycles/sec in Duron 1 gig
	Around 7-800000 moves generated/sec

Sun Feb 17 13:53:01 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	board.h, board.c:  Modified board so that moves are reversible.
	Added unMove.  move/unmove gone up to 5 million cycles/sec, moves generated
	to 1.2 million/sec
	mtdf.c, search.c, board.c:  Start using unmove instead of copying board
	bench.c, bench.h:  New files, for benchmarks and tests.  Put benchmoveunmove,
	benchmovegen in there.  Added perft to test move generator, similar to crafty.
	benchmovecycle around 980000 /sec
	
Sun Feb 17 19:58:29 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	ai.c:  Implemented Undo and Remove from the xboard protocol
	
Wed Feb 20 06:08:55 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	board.c:  Correct small bug in isRepetitionDraw.

Wed Feb 20 15:36:17 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	mtdf.c:  Don't time out unless depth > 2
	Looking at the logs, totalnodes /sec have gone up to around 150000
	bench.c:  Added sp_bencheval.  relativePositionHeuri gets around 1 million/sec.		
	
Wed Feb 20 17:49:37 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	mtdf.c:  In veryNewAlphaBeta, don't check for stalemate or checkmate until the node
		where it happens, therefore saving the canMove call, but it means that 
		checkmate detection is reduced by one.  Total nodes/sec are at around 230000

Tue Feb 26 01:16:33 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	pieces.c:  changed kingCanMove to use move/unmove.
	Released 0.3.0
	Bumped version to 0.3.1			
	
Sun Mar 03 05:26:17 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	board.c, board.h:  Added generateTentativeMoveList to generate moves that also would
		leave the player in check.
	mtdf.c:  Changed veryNewAlphaBeta to use generateTentativeMoveList and return
		ILLEGAL_POSITION in case of -1 moves being generated.

Mon Mar 04 04:21:39 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	board.c, board.h:  Added generateTentativeQuiescentMoveList same modification of
		normal version as above.
	search.c:  Changed quiescentSearch to use generateTentativeQuiescentMoveList.
	mtdf.c:  Added check for no legal moves for veryNewAlphaBeta.
	After all that, it seems to reach shallower depths. Total nodes per second are down to
	160000 /sec as well.  
	generateTentativeMoveList is twice as fast as generateMoveList though.
	search.c:  Don't store hash if depth is 0 and same key (ie, same situation)
	
Wed Mar 06 02:38:12 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	pieces.c:  Changed pieceCanMove to an array of function pointers.
	
Wed Mar 06 19:35:13 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	pieces.c:	Created array of moves pieceMoves
	alphabetasearch.c, basicsearch.c, board.c:  Use pieceMoves instead of getMoves.

Fri Mar 08 16:52:01 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	alphabetasearch.c:  Fix to use move/unMove instead of copy/move.
	basicsearch.c:  Same.   (But in a very dodgy way, so not sure if it works)
	Stating again:  move/unmove 5.2 million/sec, movegen to 1.5 million moves/sec (off
	initial board, should be slower later on), movecycle to 1.1 million /sec

Mon Apr 01 04:52:01 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	mtdf.c:  Changed percentage of time being taken before another depth is searched
	from 20 to 25%  (yes, daring :)

Wed Apr 03 18:12:29 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	Due to an unexpected windows version appearing in the winboard forum, released 
	version 0.3.2 with improved documentation
	Bumped version to 0.3.3
	
Thu Apr 04 23:22:44 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	communication.h, utils.c: Changed a couple of feature enum constants to make them 
	ammenable to the cygwin environment.
	Makefile.am:  add cmdline.h so that it can be compiled without the presence of clig
	common.h:  include system.h (doh!)
	configure.in:  check for the existence of error.h
	libale/xmalloc.c:  if error.h does not exist, write a crappy hacky version
	system.h: clean up to include only once.  define pthread_yield as sched_yield in case of
	cygwiness
	Should now compile straight out of the box in cygwin (except for the clig thing)
	
Fri Apr 05 03:21:30 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	ai.c:  Modify play to perform a blocking getAction when not pondering and not its turn
	to try to avoid using cpu time on badly behaving OSes (win98 only reported case)
	
Sun Apr 07 01:09:16 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	communication.h:  added SP_EPDLINE and SP_EPDSUITE
	bench.c:  added benchEPDSuite and benchEPDLine to be parse epd lines and suites
	mtdf.c:  iterativeMTDF: Added handling of epd cases.
	Can now do EPD testing for the "avoid move" and "best move" ops.  Checking if the
	move is the one sought is very dodgy though (moveToSan and strcmp), should right
	a SAN parser.
	
Sun Apr 07 20:33:52 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	utils.c:  Swapped the results and result comment printing so that scid picks it up.
	book.c and book.h:  New files to add opening book support.  Trying to keep the API nice
	and clean.  Implemented initialiseOpeningBook for opening a book for reading, 
	createOpeningBook for writing, closeOpeningBook for closing the book.
	
Mon Apr 08 00:26:07 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	utils.c:  Bit the bullet, wrote sanToMove.
	bench.c:  Use sanToMove in the epd testing.
	
Mon Apr 08 03:59:54 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	doIO.c, ai.c, mtdf.c:  Added support for POST and NOPOST from the IETF protocol.
	Post just the first move since that's all we've got.
	bench.c:  benchEPDLine: Use the current timelimit per problem if one exists.  keep
	a tab of total/solved in benchEPDSuite.
	mtdf.c:  Use all the time available if we are using timelimit instead of the percentage
	
Tue Apr 09 03:37:35 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	book.c:  Added addBookMove, getBookMove and getBookHash.  addBookMove adds the move to
	the current board, or, if the move is already there, increase its ranking.  getBookMove
	gets all moves for current board and returns a weighted random move according to its
	ranking.
	cmdline.cli.in:  added -nobook option for avoiding using a book and -book <file> to 
	specify the name of the book.
	communication.h, doIO.c:  Added sp_modifybook and sp_closebook.
	ai.c:  play:  when sp_modifybook is called, a book is created or modified and the 
	engine switches to book mode, which is similar to force mode, except that every
	move is added to the book.  Also added sp_sanbookmove to add book moves in san
	mtdf.c:  try getBookMove before everything.
	utils.c:  fix small bug in san parsing
	
Tue Apr 09 10:10:09 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	left it reading the first ten moves of Arturo Ochoa's p2600 from Dann Corbit's ftp site
	Now, we've a book.
	Also, tamed down my weights.
	
Wed Apr 10 00:54:40 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	cmdline.cli.in, hash.c:  Added -hk option to specify hashtable size in kilobytes.
	Default is now around 32 MB.
	utils.c:  Fixed a bug in sanToMove where it didn't disambiguate properly if one of
	the moves left the mover in check.
	Now recreates the book without errors.
	
Thu Apr 11 04:35:32 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	Made it load aleweights.out by default
	Released 0.3.3
	
Tue Jul 02 19:50:33 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	After long hiatus, reassess.  Change versioning so that it is bumped just before
	release.  v0.3.3 came 24th/33 on 9/32 in WBEC's 5th division and so got relegated to 6th.
	Also 37th/46 on RWBC's Open Class.
	
Wed Jul 03 00:03:40 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	Make program nicer.  Get rid of total* globals (incorporate into board struct), move
	a couple of variables from engineParameters to Board

Wed Jul 03 18:27:16 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	Massive code modification.  Remove engineParameter structure, put everything in Board
	(which should probably be called State now).  Input and output handling threads now get a
	threadParameter.  
	Split doc and tools directory.
	
Fri Jul 05 02:55:58 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	ai.c:  Added handleActionWhileInSearch to handle input during the search.
	mtdf.c:  Updated to check for input during search.  Move Now works correctly now.
	
Fri Jul 05 04:12:08 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	mtdf.c, ai.c:  Now handling analyse mode correctly too (even though very uglyly)
	
Mon Jul 08 01:49:54 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	search.c, mtdf.c:  Made calculateFinishTime into a function.  Called from mtdf.
	
Mon Jul 08 22:22:20 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	search.c, mtdf.c:  Converted iterativeMTDf into a generic timed iterative framework,
	making MTDf a special case.  Plonked it in search.c.

Wed Jul 10 03:09:27 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	Transforming move generation into 0x88 style (basing on Bruce Moreland's page).
	This touches pretty much everything.  Hash, book, movePiece(andHash) and 
	generate*MoveList functions done (since most work is now pushed onto the piece 
	generation functions)
	procvalues.py now pads the missing 64 with zeros
	
Wed Jul 10 22:49:14 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	InCheck and isLegal modified.  Lots of utils transformed.  
	All the piece generation done.  Something not working as perft is too low.
	(small change to the quiescent search nodes.  Now pawnGenerateCaptures will also
	produce coronations.)
			
Thu Jul 11 03:19:52 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	Stupid bug in move generation by the bishop fixed.  Perft now seems to be
	about 80% faster.  Maybe worth it.
	Stamped out another million x88-move related bugs all over the place.  Decided
	for the moment to not modify the book format since i don't feel like remaking it.
	This means that all accesses to the book are slower than they should be since the
	notation format is necessary, but the book isn't called very often anyway.
	Got rid of all the pieceCanMove functions, replaced by a generic slow one.  Shouldn't
	be called from anywhere fast in any case.
	There's probably a couple of bugs left, but there shouldn't	be too many (for the x88
	related stuff that is.  Nothing else fixed).  Also, nothing but MTDf has been moved over.
	Will do so soon.
	
	Official benchmarks (duron 1 Ghz):  perft 5 at 2.3 million/sec, move/unmove at 4.9 
	million/sec (why the slowdown?), benchmovegen at 2.6 million/sec, benchmovecycle at
	1.6 million/sec (ok, so that's about 50%, worth it?)
	Still producing illegal moves though.

Thu Jul 11 23:47:29 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	ai.c, doIO.c:  changed sp_sanbookmove to sp_sanmove and its corresponding code to a copy
	(yuk) of usermove.  This means that we can now use standard san (this is for pgn parsing)
	The usermove bit should be converted into a function but i'm too lazy to round up all the
	local variables at the moment.
	Wrote setboardfrompgn.py to set the engine at a specific point from pgn for debugging
	(at least the illegal move bug is reproduceable)	
	Caught the stupid bug (in movePieceAndHash)

Fri Jul 12 06:43:28 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	utils.c:  Fixed just-introduced-FEN-parsing-bug based on perft numbers posted on CCC.

Sun Jul 14 21:39:12 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	hash.c:  Fixed bug introduced in getHashKey  
	bench.c:  Fixed bug introduced in benchEPDLine (freeing board)	
	mtdf.c:  Fixed accounting of quiescent nodes in veryNewAlphaBeta

Tue Jul 16 03:54:24 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	board.c:  Optimised code generation calls (mainly just check for illegal board while
	in loop)  Shouldn't make much of a difference.

Wed Jul 17 06:16:04 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	alphabeta.c:  Substituted everything for a simple shell to veryNewAlphaBeta (which
	is now moved to search.c) using the timedIterative framework.
	basicsearch.c:  Fixed randomSearch, oneStep.  multistep and complete negamax still 
	broken.
	communication.c:  commented out isThereSomethingToRead as it is not used and the 
	semRead function isn't available in cygwin.
	configure.in:  Bump version to 0.4.0 for release

Wed Jul 17 23:09:17 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	configure.in:  Don't output cmdline.cli, so that clig isn't needed at compile time,
	now have to remember to change the version number in cmdline.cli
	With the fix in accounting, the nodes per second count are relatively stable at about
	250000 /sec
			
Sat Jul 20 00:11:19 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	search.h:  Increase MAX_DEPTH to 110.
	search.c:  Check if hash depth is greater than 100, if so, it's a draw.

Mon Jul 29 04:46:35 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	negascout.c:  New file.  Implemented negascout. 
	unches.c, cmdline.cli:  Made negascout the default algorithm.
	While testing it seemed to perform better than MTD(f) except that it seemed to prefer
	to stalemate instead of checkmating, and it played illegal moves when it was about to
	get mated
	
Tue Jul 30 00:49:56 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	negascout.c:  When updating alpha from hash values, initialise best move to the stored
	move (it's a guess, but it sounds right).
	search.c:  IterativeDeepening:  When in analyse mode and maximum depth is reached,
	return with state->anomaly set to the new WAIT_FOR_EVENT.
	ai.c:  Handle WAIT_FOR_EVENT properly (no, action, just block on the input queue)
	search.c:  IterativeDeepening:  Check when returned value is a checkmate, and stop
	iterating (should make for faster playing when withing checkmate zone).
	
Tue Jul 30 04:26:23 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	search.c:  quiescentSearch:  Modified hashtable replacement to only replace when theres
	nothing there (which basically means don't replace, since it fills pretty quickly).
	This is only temporary
	negascout.c:  Hopefully fixed stalemate bug by checking for the existance of a legal
	move (but there's still something wrong there, the evaluation of checkmate scores)
	
Tue Jul 30 06:02:52 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	search.c, negascout.c:  Modified so that hash retrieval of checkmates gets modified
	depending on ply it gets retrieved.  It's working strangely though.
	utils.c:  fixed bug in printBoard (maladapted to 0x88 board)
	
Tue Jul 30 20:02:25 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	hashstruct.h, negascout.c, search.c:  Increased size of hashing chunk to include
	promoting piece.
	search.c:  Added recoverPV to recover the PV line from the transposition table
	search.c:  Check for out of time while in quiescentSearch.
	bench.c:  Fixed freeing of the board when the epd line does not parse
	utils.c:  Fixed bug in FEN and EPD parsing when en-passant square was set.
	ai.c:  When receiving a setboard command, aside from printing the board, also write
	the FEN.
	negascout.c:  Fixed by keeping separate best and tempalpha (so that even if no score
	breaks alpha, we still got a meaningful hash entry)
	search.c:  recoverPV:  don't print from =0 and to = 0 entries produced during 
	quiescent search when no move is better than sitting.
	
	 
Wed Jul 31 04:18:56 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	negascout.c:  don't even look at a hash entry if it wasn't produced in the main tree
	negascout.c:  store hash if the result is exact (makes wonders for PV recovery)
	ai.c:  nailed that stupid lack of response bug (wasn't initialising lastcheck)

Sat Aug 03 06:59:31 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	negascout.c:  And now nailed the "repeating a position is the same as checkmate"
	problem.  Now, for testing.
	
Sat Aug 03 22:44:05 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	Moved site to sourceforge <http://smallpotato.sourceforge.net>
	configure.in, cmdline.cli: Bump version to 0.5.0 for (suspended) release

Mon Aug 05 22:36:48 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	Profiling now seems to indicate that evaluation is the top cpu user with about
	20%.
	
Mon Aug 12 23:35:13 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	eval.c, unches.c, negascout.c, search.c:  Fix a couple of things that wouldn't
	compile on gcc 2.x
	
Wed Aug 14 03:18:22 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	board.c, board.h:  Keep track of total values for white and black pieces on the
	structure.  So, movePieceAndHash and unMove were modified, unMoveSimple is a 
	copy of the previous unMove and is called in case movePiece was used (utils.c, board.c,
	pieces.c).  Add setColorTotals to initialise.  Call from initialiseBoard and from the
	FEN and EPD parsers.  Internalise piece values into the board structure.
	unches.c:  Changed default to position piece evaluation (no relativeness).
					
Thu Aug 15 05:05:21 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	eval.c, unches.c:  Removed heuri and relativeHeuri since there is no longer any benefit 
	on their continued existence.  Calling with -fixed will still behave in the same manner
	by loading the default values into the value tables.
	board.h, board.c:  Added piece count counters for each piece type.
	eval.c:  Changed eval functions to use running totals and piececounts.  Also, slightly
	improve the draw to insufficient material problem, and improve the non-relative version
	for the corner endgame cases (eg 2R vs R)
	(weird unreproduceable bug found during FICSing.  evaluation kept plummetting for
	no real reason)
	
Thu Aug 15 22:55:50 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	board.c:  Reproduced the unreproduceable bug from yesterday (always use same random)
	Small bug in unMove fixed, and problem is gone.
	unches.c, ai.c, search.c, cmdline.cli:  Added resigning.

Fri Aug 16 05:55:59 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	libale.[hc]:  Moved parseInteger from utils.c.  Added parseDouble.  Copied nextToken
	tokeniser function from Amy 0.8.1 by Thorsten Greiner (thorsten.greiner@web.de).
	unches.c:  Started to add configuration file support.
	
Sat Aug 17 00:50:02 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	unches.c:  Finished configuration file support.
	sp.rc:  New file, demo configuration file
	
Sun Aug 18 01:45:35 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	configure.in, cmdline.cli:  Bump version to 0.5.1 for release
			
Mon Aug 19 05:09:51 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	board.[hc]: Add isInsufficientMaterialDraw
	ai.c:  Check for claimable insufficient material draws and claim them.

Sun Sep 08 22:19:34 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	doIO.c:  Changed input to accept special commands without the sp_ prefix if not in
	xboard mode.  Also changed unchesmovetosan to sp_movetosan. Updated doc/commands.txt
	to reflect the changes.
	utils.c:  If move is obviously illegal, don't perform move, unMove.
	utils.c:  Fixed stupid bug in FEN/EPD parsing.	


Mon Sep 09 07:35:59 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	eval.c:  Split evaluations into phases.  For now there's only normal phase, and end-game
	phase.  Add checkPhase function to check if we should switch phases.
	search.c:  Check if we should switch phases in timedIterativeDeepening before starting
	the search.
	utils.c:  Set phase after parsing epd or fen.
	procvalues.py, aleweights:  Switch to take opening and endgame tables.  Enforce 
	symetrical (white/black) tables by only taking one.
	
Mon Sep 09 20:59:43 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	ai.c, doIO.c:  Added evaluate command.
	
Thu Sep 12 00:50:36 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	eval.c:  orderMoveList search crowning knight before rook
	
Thu Sep 12 21:20:13 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	basicsearch.c:  Fixed multinegamax and multistep to work with "new" utility functions
	(generateMoveList mainly).  Cut down lines of code by 70% or so
	
Sat Sep 14 04:23:36 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	Tested on WAC (10 seconds, but 9 seconds really), 194/300, but many that were solved
	were not recognized as such since sp doesn't understand multiple solutions.
	Tweaked rook and pawn endgame evaluation
	negascout.c::negaScoutDriver: Added simple aspiration search.
	
Sat Sep 14 17:37:52 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	utils.c,board.c,search.c:  Added handling of multiple epd moves to avoid/find.
	Rerunning the WAC test, it now gets 205/300 (most differences should be due to 
	multiple solution recognition and not due to the aspiration search imo)
	Changed movePiece to movePieceSimple.  movePieceAndHash to movePiece

Sun Sep 15 04:40:19 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	negascout.c, board.c:  Implemented null-move pruning (R=2 for now)
	WAC solved jumped to 222/300	
	Allowed recursive null-pruning.
	Reduce window size of the aspiration to +-40
	
Sun Sep 15 19:54:52 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	aleweights:  Penalties to kings in the middle in the opening, up values for
	king in b1,a1 and h1.  Reduce bonus for pawn on 7th.  Bonus for most pieces for
	being in the centre during the endgame. 
	search.c::calculateFinishTime:  Fix issue with 1-second reduction while doing epds.
	And now 232/300 on WAC

Mon Sep 16 17:00:58 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	aleweights, eval.[ch]:  Added mating phase evaluation.  Fixed small bug in evaluation
	that ignored king values when in mating mode.
	doIO.c:  if unknown command, say so.  Give some help if help is asked for.
	Change version to 0.6.0 for release
	
Wed Sep 18 20:34:22 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	doIO.c:  More helpful initial message.
	sp.rc:  Toned down default randomness from 3 to 2.
	sp.rc, ai.c, doIO.c, search.c:  Add epdmindepth value, default to 6, minimum depth at
	which an epd can be considered solved to try to avoid false positives, at the request
	of Dann Corbit.
	cmdline.cli: Also add epdmindepth to command line
	sp.rc,negascout.c,eval.c,unches.c: Add another option to set the draw value.
	
Thu Sep 19 22:04:55 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	search.c, sp.rc, unches.c:  Made time management minimum percentage time configurable
	through the configuration file (TimeAggression), and increased default to 35%
	Added doc/help.txt to distribution.  It's just the output of 'smallpotato -help' for
	the people who can't see that (i think win9x series, not sure).
	
Fri Sep 20 06:52:00 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	board.c::isInsufficientMaterialDraw:  Fix not recognising king vs king as draw
	
Sat Sep 21 00:36:31 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	unches.c::main add code from Dann Corbit not to use buffering on input and output
	on windows system.
	
Sat Sep 21 19:48:23 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	common.h, all over the place: replace unsigned long long with U64 and typedef it
	(trying to make it more portable).
	board.h:  Start of massive increase in state.  Attack and moves structures added, now
	to modify everything in the code...
	board.c::setPieceLocations:  New function, assign a number to each piece.
	board.c::setMoves: New function, set all available moves, attacks, and set attack board.
	pieces.c::generate*Destinations:  New functions to set possible destinations for each
	piece from each position.
	
Mon Sep 23 01:29:23 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	board.c::movePiece:  Modified to update new structures.
	board.c::updateMoves:  New function, updates moves of specified pieces
	pieces.c::generateMoves:  Modified to produce moves and captures separately, and
	to update destination and attack bitboards (won't use generate*Destinations)
	board.c::isLegal, inCheck:  Modified to take advantage of attack bitboards.
	pieces.c::pieceCanMove:  Modified to take advantage of destination bitboard.
	pieces.c::genearate*Captures, generate*Destinations removed.
	pieces.[ch]::Made static and removed from pieces.h all pieces tables, and a couple of
	functions
	ai.c::handleActionWhileInSearch:  handle help request and unknown commands without
	exiting search.
	board.c::unMove:  Changed to update new structures.
	utils.c::parseFen/EPD:  Call setPieceLocations and setMoves.
	On early tests, it seems that perft is 8 times slower!  If this translates to nodes/sec
	on game, i'll have to revert everything.
	
Thu Sep 26 15:59:12 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	board.c::checking:  New function.  Is to-move player checking?
	search.c::quiescentSearch:  Switch back to generateTentativeQuiescentMoveList, check
	if legal before recursing.
	negascout.c::negascout:  Switch back to generateTentativeMoveList, check
	if legal before recursing.  Dont null-move if in check.
	That got it up to 180000 nps.  Maybe an acceptable performance hit, if i can use the
	attack information elsewhere.  WAC goes back down to 218/300 in 10seconds (mindepth 6)
	
Sat Sep 28 00:18:40 EST 2002 Alejandro Dubrovsky <s328940@student.uq.edu.au>
	pieces.c::?PawnGenerate:  don't add moves taking your own en-passant.  This fixes
	the stuffups in perft count when using the "real" move generator (generateTentative
	and cull if checking)
	board.c::movePiece:  Fix lunacy in movePiece where if the piece moving was a pawn, 
	many of the crowning behaviour was performed.
	board.c,board.h,pieces.c:  Massive search & replace of piece for entity when referring
	to a member of the entityboard.

Wed Dec 25 03:18:38 AM EST 2002 ale <s328940@student.uq.edu.au>
	pieces.c: bKingGenerate:  fix stupid bug that was checking for wether white had castling
	rights.
	pieces.c:  ?KingGenerate:  Rook squares should also be monitored by the king.
	
Thu Dec 26 01:45:22 AM EST 2002 ale <s328940@student.uq.edu.au>
	board.c:  updateMoves:  Always recalculate kings, too tricky otherwise.
	That seems to have fixed the move generator.  Down to 200 knps.
	eval.c, search.c, negascout.c:  Limit history values (betaers) to 1 << 20.  Overvalue
	piece exchanges so that it mostly weighs higher than history value for ordering.	
	pieces.c:  pieceCanMove: Make it slightly more stringent by checking that source and
	destination squares are of different colour, and that the source is a real piece.

Thu Apr 24 03:24:43 PM EST 2003 ale <s328940@student.uq.edu.au>
     Decided all this newfangled structures are too complex for what they are worth.  Roll 
     back everything to around the 20th of September, 2002
	Lots of playing with autoconf and automake.  Reading the manual for the first time.
	Updating to automake 1.7 and autoconf 2.53

Sun Apr 27 01:37:04 PM EST 2003 ale <s328940@student.uq.edu.au>
	doIO.c: fix small memory leak in doInput (thanks to valgrind)	

Tue Jul 22 05:26:16 AM EST 2003 ale <s328940@student.uq.edu.au>
	search.c: timedIterativeDeepening:  calculate branching factor, and output on log.
	eval.c:  create capture ordering tables, and make orderMoveList use them instead
	of the table values when the move is a capture.
	Go into end phase slightly earlier.
	
Tue Jul 22 06:12:08 PM EST 2003 ale <s328940@student.uq.edu.au>
	aleweights:  turn down the value of the rook in the opening phase a tad (to try to 
	avoid the bishop + knight vs rook + pawn swaps)
	eval.c:  Rook vs minor piece is a draw

Thu Sep 25 05:53:46 AM EST 2003 ale <s328940@student.uq.edu.au>
	moveorder.h, moveorder.c:  new files.  moved orderMoveList from eval.[ch] to here
	board.h:  Added Killer struct, and killer array to Board struct.
	moveorder.c:  wrote updateKiller, getKiller, and clearKillers.
	negascout.c:  update killers.  clear killers at depth + 1 on new move
	mtdf.c, alphabetasearch.c:  deleted
	search.c, search.h:  remove veryNewAlphaBeta.
	unches.c, cmdline.cli:  get rid of all MTD(f) and Iterative alpha-beta engine options.
	moveorder.c:  Use killer moves for move ordering
	
Thu Sep 25 09:33:02 PM EST 2003 ale <s328940@student.uq.edu.au>
	moveorder.c:  Lower priority of killer moves
	(not seeing an improvement from killer moves, but might as well keep them for now)
	Bump version to 0.6.1 for release
	
Sun Sep 28 11:54:03 PM EST 2003 ale <s328940@student.uq.edu.au>
	WACing at 257/300 at 10seconds now.	

Tue Jul 20 06:26:46 AM EST 2004 Alejandro Dubrovsky <alito@organicrobot.com>
	moveorder.c, negascout.c: Implement divide history table by two when max value reached 
	idea by Tord Romstad.
	
Tue Jul 20 05:00:35 PM EST 2004 Alejandro Dubrovsky <alito@organicrobot.com>	
	aleweights:  give more centering value to queen, rook.  Rook bonus for 6th rank too.
	Don't penalise king so strongly for being on third rank.
	eval.c:  Rough king safety implemented (just looks at the three pawns in front of the
	king and their distances)
	
Tue Aug 10 09:41:32 PM EST 2004 Alejandro Dubrovsky <alito@organicrobot.com>
    eval.c:  Add fianchetto bonus to try to avoid predilection for Kf1/8 when bishop in
    fianchetto
    
Tue 2010/09/13 Alejandro Dubrovsky <alito@organicrobot.com>
    board.h, moveorder.h:  Move a couple of functions to the includes to help get them inlined.

Tue 2010/09/14 Alejandro Dubrovsky <alito@organicrobot.com>
    WACing at 272/300 at 10 seconds on a Core 2 Duo E4300

Tue 2010/09/16 Alejandro Dubrovsky <alito@organicrobot.com>
    pieces.c: Make the generate functions take a const Board pointer.
	eval.c: Get rid of king safety.
	
Tue 2010/09/17 Alejandro Dubrovsky <alito@organicrobot.com>
    hashstruct.h: Make the hash chunk 4 bytes smaller.

Tue 2010/09/17 Alejandro Dubrovsky <alito@organicrobot.com>
    unches.c: Make logging line-buffered
    search.c, negascout.c:  When the last move wasn't a capture, don't do a full quiescent check. 
    Only check for legality, and evaluation.
    board.c (generateTentativeQuiescentMoveList): More hypothetical microoptimisations.
    doIO.c: Linebuffer the debug file.
    tools/procvalues.py: Wrap it nicely in optparse.
    negascout.c: Fix case where we do an illegal move on the 100th ply draw case.
    
Tue 2010/09/21 Alejandro Dubrovsky <alito@organicrobot.com>
	negascout.c, search.c: That experiment with not doing a full quiescent search was a failure. Revert.
	
2014/01/26
    Move code from SourceForge to GitHub.
    Moved code to src subdirectory

2014/01/27
    Make code run in 64-bit OSes.
    Make book code more resistant to compiler struct-packing choices
    WACing at 280/300 at 10 seconds on an i7 870 @2.93GHz

2014/01/30
    Releasing 0.7.0


2024/12/20
    Fixing compilation thanks to report and code by Roelof Berkepeis
    Releasing 0.7.1

2024/12/21
    Added compile script needed for new autotools setup
    Releasing 0.7.2