Crafty v25.2 Release

[petero2]

Bob Hyatt and team - Tracy Reigle, Peter Skinner and myself, are please to announce the release of Crafty-v25.2

Binaries for windows and linux are available here.

The same notes as for 25.1 apply.

[Graham Banks]

Bob Hyatt and team - Tracy Reigle, Peter Skinner and myself, are please to announce the release of Crafty-v25.2

Binaries for windows and linux are available here.

The same notes as for 25.1 apply.

Thanks Peter.

[Canoike]

Thank you Mike B, it worked. One note for the Ubuntu/Debian users, /bin/csh must be replaced by /bin/sh because csh does not exist.

[Canoike]

Why does not Crafty try to win instead of offering a draw by 3 fold ?

[Event "Computer Chess Game"]
[Site "HAL2014"]
[Date "2016.10.30"]
[Round "-"]
[White "donjon"]
[Black "Crafty-25.2"]
[Result "1/2-1/2"]
[TimeControl "300+3"]
[Annotator "6... +0,06"]

1. e4 c5 2. Nf3 d6 3. Bb5+ Nc6 4. d4 cxd4 5. Nxd4 Bd7 6. Be2 e5
{+0,06/26 13} 7. Nb3 Be7 {+0,00/26 13} 8. c4 f5 {-0,18/26 11} 9. f3 Nf6
{-0,32/24 22} 10. Nc3 fxe4 {-0,22/25 16} 11. fxe4 Be6 {-0,07/26 10} 12. O-O
O-O {-0,05/27 6} 13. Be3 Qd7 {+0,00/26 11} 14. Qd3 Nb4 {+0,00/25 11} 15.
Qd1 Nc6 {+0,00/27 18} 16. Qd3 Nb4 {+0,00/31 9} 17. Qd1 Nc6 {+0,00/27 18}
{Draw claim: 3-fold repetition} 1/2-1/2

My .crafyrc file :

mt 4
hash=2048M
hashp=128M
book random 1
book width 5
bookw freq 0.5
bookw eval 0.5
bookw learn 0.5
egtb
egtbd
tbpath=/home/donjon/jeux/echecs/syzygy/
ponder on
learn=100
log off
swindle on
smpgroup=12
smpmin=7
smpgsd=10
smpgsl=7
exit

[rcmaddox]

Bob Hyatt and team - Tracy Reigle, Peter Skinner and myself, are please to announce the release of Crafty-v25.2

Binaries for windows and linux are available here.

The same notes as for 25.1 apply.

Thank you!

[rcmaddox]

Bob Hyatt and team - Tracy Reigle, Peter Skinner and myself, are please to announce the release of Crafty-v25.2

Binaries for windows and linux are available here.

The same notes as for 25.1 apply.

Thank you!

[Peter Berger]

Really - how about documenting things like MAJOR changes at least ?

After I finally managed to get one of these binaries running on my computer I noticed that there obviously have been some changes that should be explained IMHO.

Has Nalimov endgame table support been discontinued ?

It looks as if support of Roland's tables has been added - any documentation on this?

Peter

[Matthias Hartwich]

Really - how about documenting things like MAJOR changes at least ?

After I finally managed to get one of these binaries running on my computer I noticed that there obviously have been some changes that should be explained IMHO.

Has Nalimov endgame table support been discontinued ?

It looks as if support of Roland's tables has been added - any documentation on this?

Peter

As usual, all can be found in main.c:

Code: Select all

 *    25.1   Cleanup of NextMove() plus a minor ordering bug fix that would    *
 *           skip counter moves at ply = 2. Added NUMA code to force the hash  *
 *           tables to be spread across the numa nodes as equally as possible  *
 *           rather than all of the data sitting on just onenode.  This makes  *
 *           one specific user policy important.  BEFORE you set the hash size *
 *           for any of the four hash tables, you should ALWAYS set the max    *
 *           threads limit first, so that the NUMA trick works correctly.  Of  *
 *           course, if you do not use -DAFFINITY this is all irrelevant.  The *
 *           -DNUMA option has been removed.  I no longer use any libnuma      *
 *           routines.  A new "smpnuma" command is now used to enable/disable  *
 *           NUMA mode (which really only affects how the hash tables are      *
 *           cleared, all the other NUMA code works just fine no matter        *
 *           whether this is enabled or disabled.  Fixed a bug with the xboard *
 *           memory command that could overflow and cause preposterous malloc  *
 *           requests.  Change to LMP that now enables it in the last 16 plies *
 *           of search depth, although only the last 8-10 plies really have    *
 *           a chance for this to kick in unless there are more than 100 legal *
 *           moves to try.  Minor change to hash path in HashStore() that made *
 *           it hard to store entries on the first search after the table was  *
 *           cleared.  Removed Nalimov DTM EGTB code and converted to SYZYGY   *
 *           WDL/DTC tables instead (courtesy of Ronald de Man).  This         *
 *           correctly handles the 50 move rule  where the Nalimov tables      *
 *           would walk into forced draws (by 50 move rule) while convincing   *
 *           the search it was winning.  Swindle mode now also activates when  *
 *           in a drawn ending with a material plus for the side on move, as   *
 *           well as when the best root move is a "cursed win" (forced win,    *
 *           but drawn because of the 50 move rule).  This gives the non-EGTB  *
 *           opponent the opportunity to turn that 50 move draw into a loss.   *
 *           There are some changes in the scoring output as a result of this. *
 *           The usual +/-MatNN scores show up for real mates, but when in     *
 *           EGTB endings, the scores are of the form Win or Lose with the     *
 *           appropriate sign correction (-Win means black is winning, +Lose   *
 *           means white is losing.)  Basil Falcinelli contributed to the new  *
 *           syzygy code used in this version.  Minor change to skill code to  *
 *           avoid altering search parameters since the speed reduction and    *
 *           randomness in the eval is more than enough to reduce the Elo.     *
 *           Minor change to HashProbe() where I now only update the AGE of an *
 *           entry that matches the current hash signature if the entry        *
 *           actually causes a search termination, rather than updating it     *
 *           each time there is a signature match.  If the search is not       *
 *           terminated on the spot, we have to store an entry when the search *
 *           ends which will also overwrite the current exact match and update *
 *           the age as well.  Suggested by J. Wesley Cleveland on CCC.        *
 *                                                                             *
 *    25.2   Minor bug in the fail-high / fail-low code.  Crafty is supposed   *
 *           to deal with the case where the first move produces a score, then *
 *           a later move fails high but then produces a worse score.  We are  *
 *           supposed to revert to the better move.  An "optimization" made    *
 *           this fail, but it has been fixed here.  "new" command removed as  *
 *           it is pretty difficult to restore everything once a game has been *
 *           started.  To start a new game, quit crafty and restart.  Crafty   *
 *           now notifies xboard/winboard to do this automatically so using    *
 *           those interfaces requires no changes to anything.                 *

[tmokonen]

Does Peter Skinner no longer makes Windows compiles of Crafty? He used to make the official ones at one time.

[bob]

Thank you Mike B, it worked. One note for the Ubuntu/Debian users, /bin/csh must be replaced by /bin/sh because csh does not exist.

You can always install it easily. And sometimes "tcsh" or "zsh" is installed which is basically the same shell.