Battle of the Goths 10x8 Championship 2009

[beachknight]

Thanks for the background info. I didnt know most of them.
I have toyed with ChessV 0.93 few times. Not longer than 60
moves I could resist. I must try Joker and other top engines.

Best,

[smrf]

2008 10x8-Championship had seen BigLion80 being twice as strong as ArcBishop80. This time 2009 the situation seems to be inverted. Is there any explanation for this surprise? Any comment to this - maybe from the author Matthias Gemuh ?

[hgm]

Indeed, this is what struck me too. I had expected BigLon80 to be stronger, and even had considered to drop ArcBishop to make an even number if I could not fix TSCP Gohic to play anything else than Gothic Chess. (Last year I forced opening moves to create other initial positions, but I really hated that, because not every position I wanted to play was reachable from the Gothic opening array, so TSCP had to play different positions than other engines.) Fortunately I found a way to hack the TSCP binary to replace the initial position.

I obtained this version of BigLion from Matthias' web page, and it was of later date than the version I had left from 2008 on my disk, so naturaly I figured it would have been improved, and downloaded the new one. ArcBishop is the same binary as that participated last year. (Joker80, Fairy-Max, and TSCP Gothic are the same versions too; although I have in the mean time released bugfixed versions of Fairy-Max this only concerns portability fixes and winBoard protcol improvements which are not needed for this tourney and do not affect playing strength. And for Fairy-Max 4.8v I had a PGO optimized compiled version, which run 1.75 times faster than the compilations I can make myself, so I let that play again.)

[Matthias Gemuh]

2008 10x8-Championship had seen BigLion80 being twice as strong as ArcBishop80. This time 2009 the situation seems to be inverted. Is there any explanation for this surprise? Any comment to this - maybe from the author Matthias Gemuh ?

There can be many reasons for BigLion80's weakness.
All my 10x8 engines have a built-in profiler that is never fully switched off.
The profiler can be switched on in ini file.
- Maybe previous BigLion80 had no built-in profiler yet.
- Maybe I updated wrong source code.
- Maybe debug mode is ON in ini file.

Anyway, my 10x8 engines have very slow bitboard-based data structures.
My home-made 80-bit bitboard class is pathetically slow.

Matthias.

[hgm]

I used the settings file BigLion80.ini as it came with the download. This contains:

Code: Select all

[Parameter]
CommandLine=
Engine=BigLion80 2.23x
Author=Matthias Gemuh, Cameroon
Debug=False
[Support]
AnalyzeMode=False
PonderMode=True
HasMoveNow=False
CastleAndEp=True
[BigLion80]
NalimovPath=d:\Spiele\ChessBase\tbs
UseTableBases=True
NalimovCache=4
Hash=32
StrangeExtensions=1
CurrentLogNr=9
LoggingON=True
Personality=INTERNAL
[DebuggingAndFun]
bDebuggingAndFun=False
bMaterialOnly=False
bUseSmallEval=False
bTestSuite=False
bAbortEval=True
bUseLazyEval=True
bInDebugMode=False
bStoreFENs=False
bFullPV=True
bQTrace=False
bTrace=False
bDoingPerft=False
bPvsSearch=True
bPvsZeroWindows=False
bNullMoveZeroWindows=False
bFailSoft=True
bNullMoveFailSoft=True
bLessExtensions=True
bTestHashKeys=True
bUseRecognizers=False
bTryHashMove=False
bUseAtkBoards=False
bUseExtensions=True
bUseFutilityPruning=False
bUseQSearch=True
bUseHashTables=True
bUseNullMoves=True
bVerifyNullMove=True
bHashNullMove=True
bUseIID=False
nFixedDepth=5
bFixedDepth=False
bComplexQSearch=True
bChecksInQSearch=False
bQSearchInCheck=True
bEvalInCheck=True
bLegalMoves=True
bLegalQMoves=True
bKibitzEval=False
bNullMovePerft=False
bSeeQPrune=False
bDeltaQPrune=False
bKibitzOrderRootMoves=False
bKibitzOrderMoves=False
bKibitzOrderQMoves=False
bKibitzRootMoveLists=False
bKibitzEvalScaling=False
bComplexOrdering=True
bTestMoveLegality=False
bRotated_BitBoards=True

If you see anything obviously detrimental there, please let me know.

[Matthias Gemuh]

I used the settings file BigLion80.ini as it came with the download. This contains:

Code: Select all

[Parameter]
CommandLine=
Engine=BigLion80 2.23x
Author=Matthias Gemuh, Cameroon
Debug=False
[Support]
AnalyzeMode=False
PonderMode=True
HasMoveNow=False
CastleAndEp=True
[BigLion80]
NalimovPath=d:\Spiele\ChessBase\tbs
UseTableBases=True
NalimovCache=4
Hash=32
StrangeExtensions=1
CurrentLogNr=9
LoggingON=True
Personality=INTERNAL
[DebuggingAndFun]
bDebuggingAndFun=False
...
bComplexOrdering=True
bTestMoveLegality=False
bRotated_BitBoards=True

If you see anything obviously detrimental there, please let me know.

bDebuggingAndFun=False, so the ini file is OK.

.

[hgm]

Standings after cycle #3 (Gothic Chess):

Code: Select all

    name                       perc.  pnt games  Buchh    S-B

 1. Joker80 n                   92%  38.5 / 42  (777.0, 699.8)
 2. TJchess10x8 0.121           73%  30.5 / 42  (825.0, 467.8)
 3. Smirf 1.75t                 68%  28.5 / 42  (837.0, 437.5)
 4. Fairy-Max 4.8 v             56%  23.5 / 42  (867.0, 336.5)
 5. TSCP Gothic                 49%  20.5 / 42  (885.0, 319.5)
 6. ChessV 0.94                 33%  14.0 / 42  (924.0, 133.5)
 7. ArcBishop80 1.00            19%   8.0 / 42  (960.0, 117.0)
 8. BigLion80 2.23x             11%   4.5 / 42  (981.0,  67.0)

Code: Select all

Cross table, sorted by score percentage, Buchholz, SB

                              Jok TJc Smi Fai TSC Che Arc Big
 1. Joker80 n                 ### 111 111 111 11= 111 110 111
                              ### 110 011 111 111 111 111 111   92%  38.5 (777.0, 699.8)

 2. TJchess10x8 0.121         000 ### 010 =01 1=1 111 111 111
                              001 ### 101 110 11= 111 111 111   73%  30.5 (825.0, 467.8)

 3. Smirf 1.75t               000 101 ### 101 010 111 111 111
                              100 010 ### 001 111 111 =11 111   68%  28.5 (837.0, 437.5)

 4. Fairy-Max 4.8 v           000 =10 010 ### 00= 1=1 111 111
                              000 001 110 ### 001 111 110 111   56%  23.5 (867.0, 336.5)

 5. TSCP Gothic               00= 0=0 101 11= ### 11= 111 110
                              000 00= 000 110 ### 001 110 101   49%  20.5 (885.0, 319.5)

 6. ChessV 0.94               000 000 000 0=0 00= ### 111 =11
                              000 000 000 000 110 ### 111 11=   33%  14.0 (924.0, 133.5)

 7. ArcBishop80 1.00          001 000 000 000 000 000 ### =11
                              000 000 =00 001 001 000 ### 011   19%   8.0 (960.0, 117.0)

 8. BigLion80 2.23x           000 000 000 000 001 =00 =00 ###
                              000 000 000 000 010 00= 100 ###   11%   4.5 (981.0,  67.0)

We now started a round of 'Modern Carrera Chess', i.e. playing from the Carrera opening setup, but with Capablanca castling rules. (Castling was not known in the days of Pietro Carrera.)

[smrf]

Another question: is as "Fairy-Max 4.8 v" really the same engine playing as last year? This time it is earning about 50%.

[Uri Blass]

Another question: is as "Fairy-Max 4.8 v" really the same engine playing as last year? This time it is earning about 50%.

I understood that the only difference is speed when this year Fairy-Max is 1.75 times faster.

H.G.Muller already wrote in this page:

"And for Fairy-Max 4.8v I had a PGO optimized compiled version, which run 1.75 times faster than the compilations I can make myself, so I let that play again.)"

[smrf]

Well, I have seen this, but there also has been the word "again".