CT800 V1.40 released

Discussion of anything and everything relating to chess playing software and machines.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Ras
Posts: 1373
Joined: Tue Aug 30, 2016 6:19 pm
Full name: Rasmus Althoff
Contact:

CT800 V1.40 released

Post by Ras » Sun Apr 05, 2020 4:38 pm

A new release V1.40 of the CT800 engine is available! The engine should go well into the CCRL 2300 range.

The changes are:
  • More aggressive pruning in search.
  • Increased value of rook vs. minor and pawns if both sides have pawns.
  • Improved handling of king and queen vs. king and rook.
  • Compressed the bitbase for king and pawn vs. king from 32 kB to 24 kB.
  • The mate searching mode is about 30% faster.
  • More opening book knowledge.
  • Improved protection against false opening book hits.
  • Added option for aggregate castling rights in the position editor of the microcontroller version.
Download: https://www.ct800.net/download.htm


Here are two games of V1.40 (microcontroller version) vs. Mephisto Berlin Pro London (68020, 24 MHz), both at 30 minutes per game. The CT800 without overclocking.

The CT800 with Black opts for Modern Defence, and the BLP goes for the f2-f4 attack. The CT800 holds its ground and undermines White's pawn centre as intended. White doesn't get around to castling, but Black's king position isn't exactly a low-energy house, either. At the end of a wild game where both sides overlook possible advantages, the BLP heads for draw by threefold repetition through perpetual check.

In the rematch, the CT800 has White and chooses the pretty rare f2-f3 side variant of the Trompovsky attack, still from the embedded opening book. White castles long, Black short. White unleashes a pawn storm. The BLP, losing ground, tries a centre pawn sacrifice to get some counterplay, but White's attack is already too strong. A queen exchange lets Black fend off mate, but not the white passed pawns, and the BLP resigns.

That were the first two test games of V1.40, and not only that the CT800 has reached its first draw against the BLP, but even a win.



Rasmus Althoff
https://www.ct800.net

User avatar
Scally
Posts: 122
Joined: Thu Sep 28, 2017 7:34 pm
Location: Bermondsey, London
Full name: Alan Cooper
Contact:

Re: CT800 V1.40 released

Post by Scally » Sun Apr 05, 2020 5:23 pm

Hi Rasmus,

Thanks for the update, unfortunately your Raspberry Pi compile is missing a library:

Al@PicoChess:~/ct800-v1.40/binaries/Raspberry-Pi $ ./CT800_V1.40_x32
./CT800_V1.40_x32: /lib/arm-linux-gnueabihf/libc.so.6: version `GLIBC_2.28' not found (required by ./CT800_V1.40_x32)

However I was able to compile it myself using your code.


Thanks,

Al.
Alan Cooper
My Chess Computers

Ras
Posts: 1373
Joined: Tue Aug 30, 2016 6:19 pm
Full name: Rasmus Althoff
Contact:

Re: CT800 V1.40 released

Post by Ras » Sun Apr 05, 2020 5:46 pm

Scally wrote:
Sun Apr 05, 2020 5:23 pm
Thanks for the update, unfortunately your Raspberry Pi compile is missing a library:
Yes, I updated to the GCC 8 toolchain, targeting the current Raspbian Buster distro version from October 2019 because that supports also Raspi 3 and later. The previous compiles used GCC 4 targeting the Raspbian Wheezy distro version from July 2013.

Which distro version are you using? But good that at least the build script works.
Rasmus Althoff
https://www.ct800.net

User avatar
Scally
Posts: 122
Joined: Thu Sep 28, 2017 7:34 pm
Location: Bermondsey, London
Full name: Alan Cooper
Contact:

Re: CT800 V1.40 released

Post by Scally » Sun Apr 05, 2020 8:07 pm

Thanks Rasmus,

Ah yes that makes sense, I didn’t try it on Buster. Most of our users are still using Raspbian Stretch, we are encouraging them to switch over, but I’m still compiling Engines on Stretch because of this.

Al.
Alan Cooper
My Chess Computers

Ras
Posts: 1373
Joined: Tue Aug 30, 2016 6:19 pm
Full name: Rasmus Althoff
Contact:

Re: CT800 V1.40 released

Post by Ras » Sun Apr 05, 2020 8:13 pm

Scally wrote:
Sun Apr 05, 2020 8:07 pm
Ah yes that makes sense, I didn’t try it on Buster.
Ah that's why. I used the opportunity to add a readme file to the Raspi binary folder in the download archive that describes the problem and the two possible solutions (distro update or build script), which will help people in troubleshooting. Thanks for the feedback! :D

Distributing Linux binaries... Torvalds already ranted about that at the DebConf all back in 2014.
Rasmus Althoff
https://www.ct800.net

User avatar
CMCanavessi
Posts: 886
Joined: Thu Dec 28, 2017 3:06 pm
Location: Argentina

Re: CT800 V1.40 released

Post by CMCanavessi » Sun Apr 05, 2020 9:31 pm

Nice! I'll run a gauntlet with 1.34 and 1.40 and compare.
Follow my tournament and some Leela gauntlets live at http://twitch.tv/ccls

User avatar
Graham Banks
Posts: 33959
Joined: Sun Feb 26, 2006 9:52 am
Location: Auckland, NZ

Re: CT800 V1.40 released

Post by Graham Banks » Sun Apr 05, 2020 10:30 pm

Nice surprise. :)
My email addresses:
gbanksnz at gmail.com
gbanksnz at yahoo.co.nz

Jamal Bubker
Posts: 290
Joined: Mon May 24, 2010 2:32 pm

Re: CT800 V1.40 released

Post by Jamal Bubker » Mon Apr 06, 2020 9:12 am

Thanks Rasmus for this new release ! 8-)

Gabor Szots
Posts: 463
Joined: Sat Jul 21, 2018 5:43 am
Location: Szentendre, Hungary
Full name: Gabor Szots

Re: CT800 V1.40 released

Post by Gabor Szots » Mon Apr 06, 2020 2:50 pm

I have started a gauntlet and first results are very promising.
Gabor Szots
CCRL testing group

User avatar
CMCanavessi
Posts: 886
Joined: Thu Dec 28, 2017 3:06 pm
Location: Argentina

Re: CT800 V1.40 released

Post by CMCanavessi » Mon Apr 06, 2020 4:37 pm

Some results with 1.32, 1.34 and 1.40. Same gauntlet for all, improvement for every version, but huge jump from 1.34 to 1.40 :mrgreen:

Code: Select all

    Engine                         Score          CT
01: CT800 V1.32 x64                99.0/248 ········ 
02: RattateChess 1.0 Nosferatu x32 7.0/8    11110111 
02: KnightX 1.92 x32               7.0/8    10111111 
02: Fruit 1.0 x32                  7.0/8    11011111 
02: Fruit 1.5 x32                  7.0/8    11011111 
02: Betsabe II 1.75 x32            7.0/8    11111101 
07: Absolute Zero 2.4.7.2 x64      6.5/8    11=10111 
07: RubiChess 0.8.1 x64            6.5/8    1110=111 
07: Betsabe II 1.84 x32            6.5/8    01=11111 
10: Dimitri 3.93 x32               6.0/8    =011=111 
10: Chispa 4.0.3 x32               6.0/8    10111110 
12: RubiChess 0.7 x64              5.5/8    110=0111 
13: Isa 2.0.61 x64                 5.0/8    0101=11= 
13: Popochin v4.0 x64              5.0/8    =011011= 
13: CPW-Engine 1.1 x64             5.0/8    0=0=1111 
13: Popochin v4.1 x64              5.0/8    11==0110 
13: EveAnn 1.72 x32                5.0/8    011=011= 
18: Baislicka v1.0 x64             4.5/8    =1101001 
18: Zevra v1.8.6 r672 x64          4.5/8    0111010= 
18: GopherCheck 0.2.3 x64          4.5/8    01010=11 
21: Zeus 1.29 x32                  4.0/8    11=000=1 
21: FORTRESS 1.6.2 x32             4.0/8    =10110=0 
21: Isa 2.0.64 x64                 4.0/8    01100110 
21: Paladin 0.1                    4.0/8    110==001 
25: Dumb 1.0 x64                   3.5/8    10000=11 
25: CmcChess 2.6 x64               3.5/8    =1001001 
27: Buzz 0.08 x64                  3.0/8    0=00=1== 
27: Kurt 0.9.2.2 x64               3.0/8    00000111 
27: FireFly v2.7.2 x64             3.0/8    0=0=1010 
30: Beowulf v2.4a x32              2.5/8    00100=01 
30: Dumb 1.1 x64                   2.5/8    010=0=0= 
32: Sungorus 1.4 x64               1.5/8    0000=001 

Code: Select all

    Engine                         Score           CT
01: CT800 V1.34 x64                104.0/248 ········ 
02: RubiChess 0.8.1 x64            7.5/8     1=111111 
02: Betsabe II 1.75 x32            7.5/8     1111111= 
04: Dimitri 3.93 x32               7.0/8     11111=1= 
04: Isa 2.0.61 x64                 7.0/8     11110111 
04: Betsabe II 1.84 x32            7.0/8     11=111=1 
07: Fruit 1.5 x32                  6.5/8     1110=111 
08: Paladin 0.1                    6.0/8     01111110 
09: FireFly v2.7.2 x64             5.5/8     =11=0=11 
10: Fruit 1.0 x32                  5.0/8     10110101 
10: Sungorus 1.4 x64               5.0/8     001111== 
10: Isa 2.0.64 x64                 5.0/8     01011101 
10: Zeus 1.29 x32                  5.0/8     00==1111 
10: Zevra v1.8.6 r672 x64          5.0/8     011=011= 
15: RubiChess 0.7 x64              4.5/8     =1110100 
15: RattateChess 1.0 Nosferatu x32 4.5/8     =11=0=10 
15: GopherCheck 0.2.3 x64          4.5/8     =1011100 
15: FORTRESS 1.6.2 x32             4.5/8     ==01110= 
15: EveAnn 1.72 x32                4.5/8     0101=110 
15: Popochin v4.0 x64              4.5/8     0=11=0=1 
21: Kurt 0.9.2.2 x64               4.0/8     011=00=1 
21: Absolute Zero 2.4.7.2 x64      4.0/8     0=111=00 
21: Dumb 1.0 x64                   4.0/8     10000111 
21: CmcChess 2.6 x64               4.0/8     =0====1= 
25: Baislicka v1.0 x64             3.5/8     =1010010 
25: Popochin v4.1 x64              3.5/8     00=10011 
25: Chispa 4.0.3 x32               3.5/8     0100110= 
25: CPW-Engine 1.1 x64             3.5/8     0=010011 
29: Beowulf v2.4a x32              3.0/8     =00100=1 
30: Dumb 1.1 x64                   2.0/8     01000010 
30: Buzz 0.08 x64                  2.0/8     11000000 
32: KnightX 1.92 x32               1.0/8     0000==00 

Code: Select all

    Engine                         Score           CT   
01: CT800 V1.40 x64                148.5/248 ········ 
02: Fruit 1.5 x32                  8.0/8     11111111 
03: RubiChess 0.8.1 x64            6.5/8     =0111111 
03: Betsabe II 1.84 x32            6.5/8     111=1101 
05: Zeus 1.29 x32                  5.0/8     01011==1 
06: Fruit 1.0 x32                  4.5/8     1110100= 
06: CPW-Engine 1.1 x64             4.5/8     01001=11 
06: Dumb 1.0 x64                   4.5/8     0111010= 
06: Betsabe II 1.75 x32            4.5/8     00011=11 
06: Sungorus 1.4 x64               4.5/8     100===11 
11: Kurt 0.9.2.2 x64               3.5/8     100=0011 
11: CmcChess 2.6 x64               3.5/8     100=0101 
11: Isa 2.0.61 x64                 3.5/8     11=00100 
14: Chispa 4.0.3 x32               3.0/8     00100101 
14: Popochin v4.0 x64              3.0/8     10010010 
14: Paladin 0.1                    3.0/8     01=0001= 
17: RubiChess 0.7 x64              2.5/8     010=0=0= 
17: Buzz 0.08 x64                  2.5/8     100=0100 
17: Absolute Zero 2.4.7.2 x64      2.5/8     0=100100 
17: EveAnn 1.72 x32                2.5/8     0=000101 
17: FORTRESS 1.6.2 x32             2.5/8     0110000= 
17: RattateChess 1.0 Nosferatu x32 2.5/8     000=0101 
17: Popochin v4.1 x64              2.5/8     010=1000 
24: Dumb 1.1 x64                   2.0/8     00100001 
24: KnightX 1.92 x32               2.0/8     00100100 
24: Isa 2.0.64 x64                 2.0/8     =00000=1 
24: Beowulf v2.4a x32              2.0/8     0=0=0010 
28: Zevra v1.8.6 r672 x64          1.5/8     10000=00 
28: FireFly v2.7.2 x64             1.5/8     00000=10 
30: GopherCheck 0.2.3 x64          1.0/8     10000000 
30: Dimitri 3.93 x32               1.0/8     01000000 
30: Baislicka v1.0 x64             1.0/8     00000001 

Ratings on my list:

Code: Select all

   # PLAYER                                   :  RATING  PLAYED    W    D    L   (%)  D(%)  OppAvg  OppN  OppDiv
 291 CT800 V1.40 x64                          :  2333.8     248  133   31   84    60    13  2260.7    31    31.0
 322 CT800 V1.34 x64                          :  2207.2     266   93   49  124    44    18  2252.3    34    33.9
 333 CT800 V1.30 x64                          :  2170.0     242   96   53   93    51    22  2165.1    37    33.7
 334 CT800 V1.32 x64                          :  2167.3     286   93   44  149    40    15  2242.9    39    38.2
 340 CT800 V1.20 x64                          :  2148.9     326   98   46  182    37    14  2304.5   116   100.1
 346 CT800 V1.12 x64                          :  2131.7     248   88   35  125    43    14  2213.1    86    75.0
Follow my tournament and some Leela gauntlets live at http://twitch.tv/ccls

Post Reply