Battle of the Goths 10x8 Championship 2009

[beachknight]

Sorry about the confusion; I should probably have posted this on the viewer web-page.

Don't worry. Thanks for this interesting tournament. About Smirf, is there any version available to download? I tried with some old links but they don't work.

I would also be interested in a free x64 version of Smirf.
It was fun watching old free w32 versions with Smirfoglot.

best,

[smrf]

... I would also be interested in a free x64 version of Smirf.
It was fun watching old free w32 versions with Smirfoglot. ...

Smirf and its GUI have been developed using former Turbo C++. It still is not supporting 64 Bit at Windows. Thus the SMIRF engine DLL still will remain 32 Bit. Future Octopus will be running on Mac OS X, thus being a 64 Bit program, probably a kind of a UCI engine, because Winboard still seems to artificially distingish among variants, where merely supporting X-FEN strings setup already would do the full job for several interesting games. But this will still need a lot of time ...

[hgm]

I would also be interested in a free x64 version of Smirf.
It was fun watching old free w32 versions with Smirfoglot.

Why would you want an x64 version of a mailbox engine? That will do no good at all, and very likely will only slow it down.

[Aaron Becker]

I would also be interested in a free x64 version of Smirf.
It was fun watching old free w32 versions with Smirfoglot.

Why would you want an x64 version of a mailbox engine? That will do no good at all, and very likely will only slow it down.

Not necessarily true. Binaries that use the 64 bit instruction set have more general purpose registers available to them, which often lets compilers generate more efficient code. My engine uses an 0x88 board, and gains about 200K nps going from a 32 bit binary to 64.

[hgm]

Cycle #1 (Bird's Chess) has completed. The standings after 14 games are:

Code: Select all

    name                       perc.  pnt games  Buchh    S-B

 1. Joker80 n                  100%  14.0 / 14  ( 84.0,  84.0)
 2. TSCP Gothic                 68%   9.5 / 14  ( 93.0,  42.5)
 3. TJchess10x8 0.121           64%   9.0 / 14  ( 94.0,  40.3)
 4. Smirf 1.75t                 64%   9.0 / 14  ( 94.0,  39.5)
 5. Fairy-Max 4.8 v             57%   8.0 / 14  ( 96.0,  33.5)
 6. ChessV 0.94                 29%   4.0 / 14  (104.0,   8.5)
 7. ArcBishop80 1.00            11%   1.5 / 14  (109.0,   1.5)
 8. BigLion80 2.23x              7%   1.0 / 14  (110.0,   2.8)

The major surprise is TSCP Gothic, which was able to steal a point from Smirf and half a point from TJchess. The latter two also let their guard down against Fairy-Max, which has been doing much better so far than it would have had any right to hope. As a result, TSCP is now in second place, before TJchess and Smirf. Joker80 won all its games, and took a firm lead, despite the fact that in several games it had a quite precarious positon (also according to it own evaluation).

But there are still four more cycles to go, so much can still happen before the tourney is over.

Code: Select all

Cross table, sorted by score percentage, Buchholz, SB

                              Jo TS TJ Sm Fa Ch Ar Bi
 1. Joker80 n                 ## 11 11 11 11 11 11 11  100%  14.0 ( 84.0,  84.0)
 2. TSCP Gothic               00 ## 0= 10 11 11 11 11   68%   9.5 ( 93.0,  42.5)
 3. TJchess10x8 0.121         00 1= ## 01 =0 11 11 11   64%   9.0 ( 94.0,  40.3)
 4. Smirf 1.75t               00 01 10 ## 10 11 11 11   64%   9.0 ( 94.0,  39.5)
 5. Fairy-Max 4.8 v           00 00 =1 01 ## 1= 11 11   57%   8.0 ( 96.0,  33.5)
 6. ChessV 0.94               00 00 00 00 0= ## 11 =1   29%   4.0 (104.0,   8.5)
 7. ArcBishop80 1.00          00 00 00 00 00 00 ## =1   11%   1.5 (109.0,   1.5)
 8. BigLion80 2.23x           00 00 00 00 00 =0 =0 ##    7%   1.0 (110.0,   2.8)

[hgm]

Not necessarily true. Binaries that use the 64 bit instruction set have more general purpose registers available to them, which often lets compilers generate more efficient code. My engine uses an 0x88 board, and gains about 200K nps going from a 32 bit binary to 64.

I guess it depends on how heavily you use pointers. Although SMIRF's source is not public, I always suspected it to use pointers heavily (because its 'fully-informed move generator' suggests it uses n attack table based on linked lists), and in that case the increased cache load due to the larger pointer size will probably outweigh the advantage of the extra registers.

[beachknight]

I would also be interested in a free x64 version of Smirf.
It was fun watching old free w32 versions with Smirfoglot.

Why would you want an x64 version of a mailbox engine? That will do no good at all, and very likely will only slow it down.

Not necessarily true. Binaries that use the 64 bit instruction set have more general purpose registers available to them, which often lets compilers generate more efficient code. My engine uses an 0x88 board, and gains about 200K nps going from a 32 bit binary to 64.

Is there a windows x64 binary of Daydreamer 1.6 available for download?

Best,

[hgm]

The second cycle (Cpablanca Chess) has been completed. Joker80 did not reign as supremely as in the first cycle; it had to conceed 2.5 points. Nevertheless it increased its lead on all of the competitors, although TJchess, which had a very good cycle, only slipped by half a point compared to Joker.

Fairy-Max continues to surprise, with a win aganst Smirf, while it let slip a win against TSCP, due to a perpetual check. TSCP was similarly lucky in its first game against ChessV, but then ChessV took revenge, and won the second game. This allowed Fairy-Max to approach TSCP up to half a point.

Code: Select all

    name                       perc.  pnt games  Buchh    S-B

 1. Joker80 n                   91%  25.5 / 28  (346.0, 315.5)
 2. TJchess10x8 0.121           71%  20.0 / 28  (368.0, 197.8)
 3. Smirf 1.75t                 66%  18.5 / 28  (374.0, 192.5)
 4. TSCP Gothic                 57%  16.0 / 28  (384.0, 162.8)
 5. Fairy-Max 4.8 v             55%  15.5 / 28  (386.0, 136.3)
 6. ChessV 0.94                 34%   9.5 / 28  (410.0,  58.3)
 7. ArcBishop80 1.00            14%   4.0 / 28  (432.0,  42.3)
 8. BigLion80 2.23x             11%   3.0 / 28  (436.0,  26.8)

Code: Select all

Cross table, sorted by score percentage, Buchholz, SB

                              Joke TJch Smir TSCP Fair Ches ArcB BigL
 1. Joker80 n                 #### 1111 1110 11=1 1111 1111 1101 1111   91%  25.5 (346.0, 315.5)
 2. TJchess10x8 0.121         0000 #### 0101 1=11 =011 1111 1111 1111   71%  20.0 (368.0, 197.8)
 3. Smirf 1.75t               0001 1010 #### 0101 1010 1111 111= 1111   66%  18.5 (374.0, 192.5)
 4. TSCP Gothic               00=0 0=00 1010 #### 11=1 11=0 1111 1101   57%  16.0 (384.0, 162.8)
 5. Fairy-Max 4.8 v           0000 =100 0101 00=0 #### 1=11 1111 1111   55%  15.5 (386.0, 136.3)
 6. ChessV 0.94               0000 0000 0000 00=1 0=00 #### 1111 =111   34%   9.5 (410.0,  58.3)
 7. ArcBishop80 1.00          0010 0000 000= 0000 0000 0000 #### =110   14%   4.0 (432.0,  42.3)
 8. BigLion80 2.23x           0000 0000 0000 0010 0000 =000 =001 ####   11%   3.0 (436.0,  26.8)

[beachknight]

ChessV performed much worser than I thought.
Clearly needs heavily improvement.

Best,

[hgm]

It is the first time ChessV can participate. Only a few months ago it was made WinBoard compatible. Gregory is writing a follow-up program (Quadrox), which already plays normal Chess, and participated a few times in the monthly on-line blitz tourneys. From that I know it is already a lot stronger than this ChessV, although bugs still surfaced occasionally. But it does not play any variants yet.

I must say that ChessV is a bit unlucky: from earlier testing I know that it occasionally crashes when the opponent castles. (I just found out that in this case it simply thinks the castling is an illegal move, and ignores it, waiting for ever on a "valid" move. It also does not indicate castling when you pick up the King on its own GUI window in that case.) But this already happened two times now in this tourney.

Another loss on time of ChessV turns out now t be due to a mistake by me: I had it play Bird's Chess, but I had not realized that in Bird's Chess a Chancellor is called an Equerry, and needs promotion character 'e'. So when the opponent promoted to C, it refused the move. I should have renamed the piece, which is possible in ChessV. Probably I will replay the end of that game from the promotion on (against Smirf), with 10 min on each clock.

But I guess this will not alter the general conclusion, that ChessV is in the bottom half of this field, Elo-wise.