Shogi engine invalid moves (Dan and HGM)

[hgm]

Indeed, WinBoard needs quite some massaging to play mini-Shogi. It is also important to set the -pieceToCharTable to inform it that it should not reserve space for L and N in the holdings (pushing S and G out of it, so they will demote to Pawns).

The WinBoard install should contain a file mini.ini, which contains the required settings to configure WinBoard for mini-Shogi. You can put '@mini' on the engine line of an engine that you want to play mini-Shogi. Just to be sure, the file contains this:

Code: Select all

;
; special settings required for mini-Shogi
;
; overrule board dimension and holdings size
;
/variant=shogi
/boardWidth=5
/boardHeight=5
/holdingsSize=5
;
; tell WB which pieces can participate (with normal abbreviation)
;
/pieceToCharTable="P.BR.S...G.+.++.+Kp.br.s...g.+.++.+k"
;
; give file for defining initial position
;
/loadPositionFile="mini.fen"
;
; opening book (not included, but exists!)
;
/polyglotBook="mini.bin"
/usePolyglotBook=false
/firstXBook
/secondXBook
;
; Since board is so small, use large square size
;
/boardSize=bulky
/showCoords=true
/coordFont="Arial:10 bc1"
/clockFont="Arial:24 bc1"
/overrideLineGap=-1
/flipBlack=false
/allWhite=false
/whitePieceColor=#ffffcc
/blackPieceColor=#202020
/lightSquareColor=#c8c365
/darkSquareColor=#77a26d
;
;
;
/saveSettingsOnExit=false

Essential are the board/holdings-size overrides, the variant and the pieceToCharTable. The rest if just for graphics.

[Ferdy]

Thanks it works now, I have mini.ini already, and I commented out 2 lines from winboard.ini file then add @mini.ini to one of the installed engines.

[hgm]

Nebiyu 1.45 still has a problem: it doesn't know that in mini-Shogi repetioins that are not perpetual checks count as a loss for white (sente). Instead it claims a draw on the 3rd occurence (while in Shogi the rule is 4-fold repetition!, so claiming a win on the third when your opponent repeats also results in a forfeit!) See http://minerva.cs.uec.ac.jp/~uec55shogi ... English%29 .


This gives Shokidoki a very easy win when it plays black:

Code: Select all

[Event "Computer Chess Game"]
[Site "MAKRO-PC"]
[Date "2013.11.21"]
[Round "-"]
[White "NebiyuAlien_1.45"]
[Black "Shokidoki Yokohama 2013"]
[Result "0-1"]
[TimeControl "1200"]
[Variant "shogi"]
[FEN "rbsgk/4p/5/P4/KGSBR[-] w 0 1"]
[SetUp "1"]

{--------------
r b s g k
. . . . p
. . . . .
P . . . .
K G S B R
white to play
--------------}
1. Bc2 {+0.00/24} Bc4 {-0.12/16 55} 2. Sd2 Sb4 {+0.32/16 1:11} 3. Gb2 Gd4
{+0.40/15 52} 4. Kb1 {+0.00/25 7} Bd5 {+1.12/13 25} 5. Re2 {+0.00/24 29}
Bc4 {+2.84/14 27} 6. Re1 {+0.00/25 5} Rb5 {+0.68/15 43} 7. Ba4
{+0.00/24 28} Ra5 {+0.64/15 9} 8. Bc2 {+0.00/26 28}
{False draw claim: 'Draw by repetition'} 0-1

Perhaps you can still fix this? It will never hurt to claim only on the 4th repeat in all games, and if all else fails you can always make a dedicated version of NebiyuAlien for mini-Shogi for the UEC Cup that doesn't count repetitions as a draw.

[Daniel Shawul]

I have fixed it now so that it claims losses on the 4th repetition only when playing minishogi. I have also noticed that Nebiyu seemed to have difficulties winning with black after I added the rule but couldn't figure out why. I guess this misunderstanding is probably the reason but I will have to run some games to see if there is any change. Btw the rule is not perfectly implemented since it does not differentiate between perpetual checks and standard repetitions, but i guess the former is rare enough not to cause a problem.

I will work on its search and improve it for the tournament since it seems to stand a real chance of winning it.

[hgm]

Beware! Perpetual checks are not all that uncommon (for engines that do not know it is forbidden. If they know that it loses, they of course never do it)..

[Ferdy]

Is there any rule on stalemate, is this also a win like checkmate?

[hgm]

Yes, stalemate is a win.

[hgm]

Daniel, what is the current state of Nebiyu and its participation in the UEC Cup? Did you contact prof. Takeshi Ito? You posted here that you fixed Nebiyu for the repetition draw claims, but you did not provide a link. Your website still says version 1.45 updated in May 2013. They have Nebiyu 1.45 in the WinBoard package I sent them, but that is the version that still makes the false claims.

Did you send them a corrected version yourself? Note that tomorrow around this time the tourney will have finished!

[Daniel Shawul]

I have contacted Prof. Takeshi the same day but I didn't get a reply. Well I thought he may have been busy, so I didn't ask again. My google sites page have problems these days so I was not able to update Nebiyu. The one currently on the google pages had a bug which doesn't allow it to make any drop move. I think what you have is a previous version which works fine except for the repreating rule. Here is a link to the new version which fixes everything https://dl.dropboxusercontent.com/u/55295461/Nebiyu.zip
Hope it is not too late.

[hgm]

Indeed, it seems that Takeshi Ito did have problems last week with his mailing system. I also didn't get any replies to my messages either. This seems to have been fixed now, as I did receive confirmation from him that he received my package with WinBoard ans all the engines. (But it contained Nebiyu 1.45.)

I guess I will mail him the NebiyuAlien.exe of the 1.46 package, with instructions to overwrite the one in the install with it. Can this version still run with the same alien.ini file as the old one, or would I have to send that one too. I remember I had to make some changes in the alien.ini file, to rename minishogi back to 5x5+5_shogi, because that is how some of the other engines still call it. (And I think the package I send to Takeshi Ito contained the standard WinBoard 4.7.2, which doesn't handle engine-defined variant names yet.)