---

Copyright ©

Pseudo 0.7 WinBoard by Jan Klima,CZ
irenicus@volny.cz

characteristics:
-0x88 board representation
-principal variation search
-null move pruning R=3, generates checks in qsearch after null move
-internal iterative deepening
-futility pruning,lazy evaluation...
-attack tables
-SEE qsearch pruning and SEE-directed ordering of captures

Some remarks:
-it's possible that this experimental version is actually much weaker than 0.6h
-if you are running matches under Arena, it's safer to reset engine before every game, I still have some bugs in winboard protocol implementation
-it's questionable if pondering gives any improvement
-lazy eval is also very questionable...especially in positions where king safety is an issue it completely ruins knowledge, so you might want to switch it off
-if there's anyone who could and would like to create a strong tournament book for Pseudo somehow(in any format), please let me know
-btw, anyone understands why SlowChess WV stands for "Weaker version"? ;)

0.6->0.7
Well, I didn't have any free time during the past few months, so there are only few changes made in the last two or three weeks. In parallel I began to work on a complete rewrite because I didn't like nps at all...but now it seems that it isn't that critical, so I will postpone this. For 0.8 I want to write better evaluation and finally change hash tables from replace always to something better, and then I will see...

Evaluation:
-fixed KRPKP endgame eval
-static eval cache(doesnt seem to help much when lazy eval on)
-some basic bad trades antibonus finally added
-re-added some knight mobility code
-larger bonus for center pawns
-temporarily removed mate pattern evaluation because it was buggy
-fixed king safety code,  there were at least two problems:1)with opposite castling "pawn storm" code gave sometimes terribly high scores even in dead draw positions 2)with opposite castling the side with king on the queenside was encouraged to have the worst possible pawn shelter 
-endgame piece-square tables fixed,i hope there wont be any drawn KRK endings now
Search:
-first check now extended with only 1/2 ply, recapture extension is disabled on default
-more aggresive prunings and reductions added, the most interesting thing here is history pruning-it gives usually 1-3 plies in the middlegame, but at the cost of ignoring both strong tactical and positional lines...it seems to be a gain, but I didnt have time to do much testing(and search without reasonable evaluation is good for nothing anyway ;()
-move legality is now tested before making move(speedup), the same for the test if the move gives check(slowdown)

Other:
-some _basic_ time control added
-note that while .ini settings weren't for much use in 0.6, here they might be interesting to experiment with

older versions info:

0.6g->0.6h:
-fixed(i hope so) cases when engine lost on time and crashed
-recapture extension added(1/2 ply)
-changed futility pruning, now it's more safe(and maybe it gives no speedup at all):) current search setting improved performance in some tactital test suites
-in this version nodes are counted in a different manner(nodes=number of MakeMove calls... position after making a null move is not counted as a new node)
-some minor eval tweaks

0.6f->0.6g:
-bug that might cause problems(crashes) in KX-K positions fixed
-possible i/o bugs fixed


0.6->0.6f:
-mate threat extension fixed
-fifty moves rule fix
-book fix
-lazy eval modified(less nps probably)
-now writing to only one log file, which is deleted when program starts:)
-some minor eval fixes
-safe handling of mate scores(temporary)

many thanks to:
-Robert Hyatt, Crafty source is nice to see(SEE) "what-is-this-all-about"
-Bruce Moreland, his pages are really outstanding for beginners
-Filip Tvrzsky, for all his replies to my dumb questions
-Ed Schroder, for many great advanced ideas published on his web
-Fabien Letouzey, his open source program Fruit proved to be excellent partner for many test games, it also inspired me to write simple drawish endgame and candidate passer evaluation
-all other chess programmers, whose engines I used in tournaments to test every new build
-Colin Frayn and Dann Corbit, I stole their Beowulf's opening book without asking :(

-and of course to all other people from CCC forum

© COPYRIGHT 2001 - 2012 All Rights Reserved wbec-ridderkerk.nl