Chess2u
Would you like to react to this message? Create an account in a few clicks or log in to continue.

Chess2uLog in

Mars tests

descriptionMars tests - Page 3 EmptyRe: Mars tests

more_horiz
Thank you, hafaba!

descriptionMars tests - Page 3 EmptyRe: Mars tests

more_horiz
Currently running 1000 games of Gull and Mars 3.31 with TC 1m / 0 increment. Will post the result once the tournament finish. Hope this will help increase the strength of mars

descriptionMars tests - Page 3 EmptyGull3 vs MARS3.31 1000 round-robin for 1m / 0 increment

more_horiz
If you go through the games, many times MARS lost on time

Code:


Round-Robin
Gull3_vs_Mars331

#   Engine            Score                 Result
================================================================
1.   Gull 3x64       1447 / 2000             +1184 - 289 = 527
2.   MARS3.31_TMOFF   786 / 2000             +378 -806 =816
3.   MARS3.31_TMON    767 / 2000             +365 -832 =803
================================================================
GUI: Crafty Chess             Crosstable: SCID vs PC
Operating: Windows 7 64Bit
Hardware: Intel core -i7 [4 cores]
TB: No
TC: 1 minutes / 0 increment
Number of Round: 1000
Total Number of Games: 3000
Opening Book: No

PGN: GULL3_VS_MARS331.PGN_1000R

descriptionMars tests - Page 3 EmptyRe: Mars tests

more_horiz
@hafaba wrote:
If you go through the games, many times MARS lost on time

Yes, Mars isn't good at bullet without an increment.
Thanx for the tests, hafaba!

descriptionMars tests - Page 3 EmptyMars 3.32 against Elektro 1.2

more_horiz
RankEngineScore%MaElS-B
1Mars_3.32_x64    64.0/10064.0· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·0110======110=0==11=1==11=1==1=1=1====1===1==11001111========01=11=====1==1=1=1=1=11=1====1===1===1=2304.00
2Elektro_12_popcnt_w6436.0/10036.01001======001=1==00=0==00=0==0=0=0====0===0==00110000========10=00=====0==0=0=0=0=00=0====0===0===0=· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·2304.00
100 games played / Tournament is finished
Tournament start: 2014.12.01, 00:38:46
Latest update: 2014.12.01, 05:00:16

Level: Blitz 1/0
Hardware: Intel(R) Core(TM) i7-2630QM CPU @ 2.00GHz with 7.9 GB Memory
Operating system: Windows 7 Home Premium Home Edition Service Pack 1 (Build 7601) 64 bit
PGN-File: Hafaba_TEST_on_MARS.pgn
Opening: Randomly chosen from ECO BXX which contain 1-0 and 0-1 of 36377 games [first 4 moves]
Table created with: Arena 3.5

descriptionMars tests - Page 3 EmptyMars 3.32 against Gull 3

more_horiz
RankEngineScore%GuMaS-B
1Gull 3 x64  66.5/10066.5· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·11110===11=10=1110111=111=1==1==1=11=0=11=1010111==1=0=01=1=11==1=1=0=1011====111===1==1==1=1=0=01=02227.75
2Mars_3.32_x6433.5/10033.500001===00=01=0001000=000=0==0==0=00=1=00=0101000==0=1=10=0=00==0=0=1=0100====000===0==0==0=0=1=10=1· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·2227.75
100 games played / Tournament is finished
Tournament start: 2014.12.01, 15:17:12
Latest update: 2014.12.01, 19:12:26

Level: Blitz 1/0
Hardware: Intel(R) Core(TM) i7-2630QM CPU @ 2.00GHz with 7.9 GB Memory
Operating system: Windows 7 Home Premium Home Edition Service Pack 1 (Build 7601) 64 bit
PGN-File: Hafaba_TEST_on_MARS_GULL.pgn
Opening: Randomly chosen from ECO BXX which contain 1-0 and 0-1 of 36377 games [first 4 moves]
Table created with: Arena 3.5

descriptionMars tests - Page 3 EmptyMars 3.33 against Gull 3 1m/0inc

more_horiz
RankEngineScore%GuMaS-B
1Gull 3 x64  66.0/10066.0· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·1==1==1111====0011=1===0=0=101=0==1111=11111111011=11==1=1101010==1======1=11===11=1=01====1=11===102244.00
2Mars_3.33_x6434.0/10034.00==0==0000====1100=0===1=1=010=1==0000=00000000100=00==0=0010101==0======0=00===00=0=10====0=00===01· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·2244.00
100 games played / Tournament is finished
Tournament start: 2014.12.11, 15:51:07
Latest update: 2014.12.12, 14:19:43

Level: Blitz 1/0
Hardware: Intel(R) Core(TM) i7-2630QM CPU @ 2.00GHz with 7.9 GB Memory [4 cores]
Operating system: Windows 7 Home Premium Home Edition Service Pack 1 (Build 7601) 64 bit
Opening: NoomenTestsuite2014.pgn [First 4 moves only]
PGN-File: MARS3.33_VS_GULL3.pgn
Table created with: Arena 3.5


Compare with the result above between Mars 3.32 against Gull 3, Mars 3.33 shows an increment. Keep up the good work. I will always support you clap

descriptionMars tests - Page 3 EmptyRe: Mars tests

more_horiz
@hafaba wrote:
... Keep up the good work. I will always support you clap


Thank you, hafaba!

descriptionMars tests - Page 3 EmptyMars tests

more_horiz
TC 70"+0.7"

Code:

  Program                   Elo    +    -   Games   Score   Av.Op.  Draws
   1 Stockfish 150105 x64    : 3230    6    6  6000    65.2 %   3117   39.2 %
   2 Komodo 8 x64            : 3184    6    6  6000    58.0 %   3125   38.9 %
   3 Houdini 4 x64           : 3180    5    5  6000    57.6 %   3125   38.0 %
   4 Gull 3 x64              : 3102    5    5  6000    45.1 %   3138   43.9 %
   5 Fire 4 x64              : 3093    6    6  6000    43.7 %   3140   44.4 %

   6 Mars 3.35 x64           : 3073    5    5  6000    40.5 %   3143   40.2 %

   7 Equinox 3.3 x64         : 3070    5    5  6000    40.0 %   3144   42.9 %


Spoiler :

Code:

Games        : 21000 (finished)
White Wins   : 7230 (34.4 %)
Black Wins   : 5145 (24.5 %)
Draws        : 8625 (41.1 %)
Unfinished   : 0
 
White Score  : 55.0 %
Black Score  : 45.0 %
 

Individual statistics:
 
1 Stockfish 150105 x64  : 3230 6000 (+2734,=2350,-916), 65.2 %
Mars 3.35 x64             : 1000 (+516,=368,-116), 70.0 %
Komodo 8 x64              : 1000 (+363,=413,-224), 57.0 %
Houdini 4 x64             : 1000 (+415,=371,-214), 60.0 %
Gull 3 x64                : 1000 (+475,=401,-124), 67.5 %
Fire 4 x64                : 1000 (+467,=398,-135), 66.6 %
Equinox 3.3 x64           : 1000 (+498,=399,-103), 69.8 %
 
2 Komodo 8 x64          : 3184 6000 (+2316,=2334,-1350), 58.0 %
Mars 3.35 x64             : 1000 (+479,=364,-157), 66.1 %
Stockfish 150105 x64      : 1000 (+224,=413,-363), 43.0 %
Houdini 4 x64             : 1000 (+327,=368,-305), 51.1 %
Gull 3 x64                : 1000 (+388,=408,-204), 59.2 %
Fire 4 x64                : 1000 (+431,=390,-179), 62.6 %
Equinox 3.3 x64           : 1000 (+467,=391,-142), 66.3 %
 
3 Houdini 4 x64         : 3180 6000 (+2313,=2283,-1404), 57.6 %
Mars 3.35 x64             : 1000 (+472,=385,-143), 66.5 %
Stockfish 150105 x64      : 1000 (+214,=371,-415), 40.0 %
Komodo 8 x64              : 1000 (+305,=368,-327), 48.9 %
Gull 3 x64                : 1000 (+412,=395,-193), 61.0 %
Fire 4 x64                : 1000 (+437,=387,-176), 63.0 %
Equinox 3.3 x64           : 1000 (+473,=377,-150), 66.2 %
 
4 Gull 3 x64            : 3102 6000 (+1389,=2634,-1977), 45.1 %
Mars 3.35 x64             : 1000 (+326,=377,-297), 51.5 %
Stockfish 150105 x64      : 1000 (+124,=401,-475), 32.5 %
Komodo 8 x64              : 1000 (+204,=408,-388), 40.8 %
Houdini 4 x64             : 1000 (+193,=395,-412), 39.0 %
Fire 4 x64                : 1000 (+214,=609,-177), 51.9 %
Equinox 3.3 x64           : 1000 (+328,=444,-228), 55.0 %
 
5 Fire 4 x64            : 3093 6000 (+1289,=2661,-2050), 43.7 %
Mars 3.35 x64             : 1000 (+327,=415,-258), 53.5 %
Stockfish 150105 x64      : 1000 (+135,=398,-467), 33.4 %
Komodo 8 x64              : 1000 (+179,=390,-431), 37.4 %
Houdini 4 x64             : 1000 (+176,=387,-437), 37.0 %
Gull 3 x64                : 1000 (+177,=609,-214), 48.1 %
Equinox 3.3 x64           : 1000 (+295,=462,-243), 52.6 %
 
6 Mars 3.35 x64         : 3073 6000 (+1221,=2412,-2367), 40.5 %
Stockfish 150105 x64      : 1000 (+116,=368,-516), 30.0 %
Komodo 8 x64              : 1000 (+157,=364,-479), 33.9 %
Houdini 4 x64             : 1000 (+143,=385,-472), 33.5 %
Gull 3 x64                : 1000 (+297,=377,-326), 48.5 %
Fire 4 x64                : 1000 (+258,=415,-327), 46.5 %
Equinox 3.3 x64           : 1000 (+250,=503,-247), 50.1 %
 
7 Equinox 3.3 x64       : 3070 6000 (+1113,=2576,-2311), 40.0 %
Mars 3.35 x64             : 1000 (+247,=503,-250), 49.9 %
Stockfish 150105 x64      : 1000 (+103,=399,-498), 30.3 %
Komodo 8 x64              : 1000 (+142,=391,-467), 33.8 %
Houdini 4 x64             : 1000 (+150,=377,-473), 33.9 %
Gull 3 x64                : 1000 (+228,=444,-328), 45.0 %
Fire 4 x64                : 1000 (+243,=462,-295), 47.4 %


http://spcc.beepworld.de/top-bullet-list.htm

descriptionMars tests - Page 3 EmptyRe: Mars tests

more_horiz
TC=600"+3"
Hash=128MB
Cores=1
Ponder=off
GUI=cutechess-cli 0.6.0

Code:

# PLAYER                   : RATING  ERROR   POINTS  PLAYED    (%)

1 Stockfish 6 64 POPCNT    : 3210.8   26.4    757.5    1000   75.8%
2 Mars 3.36 x64            : 3022.0   35.9    127.0     500   25.4%
3 Mars 1 AVX x64           : 3000.0   35.9    115.5     500   23.1%


Code:

Individual statistics:

1 Stockfish 6 64 POPCNT     : 3210.8  1000 (+553,=409,- 38), 75.8 %

Mars 3.36 x64                 : 500 (+268,=210,- 22), 74.6 %
Mars 1 AVX x64                : 500 (+285,=199,- 16), 76.9 %

2 Mars 3.36 x64             : 3022  500 (+ 22,=210,-268), 25.4 %

Stockfish 6 64 POPCNT         : 500 (+ 22,=210,-268), 25.4 %

3 Mars 1 AVX x64            : 3000  500 (+ 16,=199,-285), 23.1 %

Stockfish 6 64 POPCNT         : 500 (+ 16,=199,-285), 23.1 %


Games

descriptionMars tests - Page 3 EmptyMars 3.38

more_horiz
Ran a quick test with Mars 3.38 against Deep Junior Yokohama. 3' + 1" bonus per move.

20 games. Mars 17.0 DJ 3.0 +14 =6 -0

i7 6 CPUs ponder off. So one sided I stopped the test.

Impressive start for Mars 3.38! clap

descriptionMars tests - Page 3 EmptyMars 3.38 vs. 3.37

more_horiz
First 50 games.

Mars 3.37 25.5 - Mars 3.38 24.5 +6 =39 -5 in favor of 3.37.

2' + 1" bonus
i7 running 6 CPUs
no ponder

More games to be added as they finish.

descriptionMars tests - Page 3 EmptyMars 3.38 vs. 3.37

more_horiz
OK! After 100 games, Mars 3.38 and Mars 3.37 were in a dead heat. 50 points each. +9 =82 -9

Same computer and config. 2' + 1"

more games are needed on engines this close in strength!

descriptionMars tests - Page 3 EmptyRe: Mars tests

more_horiz
Small test at blitz TC:

Code:

    Program                          Elo    +   -   Games   Score   Av.Op.  Draws

  1 Mars 3.42 x64                  : 3033   19  19   600    52.8 %   3013   54.8 %
  2 Critter 1.6a 64-bit            : 3019   23  23   397    48.0 %   3033   55.7 %
  3 Equinox 3.30 x64mp             : 3003   33  33   203    45.8 %   3033   53.2 %


Individual statistics:

1 Mars 3.42 x64             : 3033  600 (+152,=329,-119), 52.8 %

Equinox 3.30 x64mp            : 203 (+ 56,=108,- 39), 54.2 %
Critter 1.6a 64-bit           : 397 (+ 96,=221,- 80), 52.0 %

2 Critter 1.6a 64-bit       : 3019  397 (+ 80,=221,- 96), 48.0 %

Mars 3.42 x64                 : 397 (+ 80,=221,- 96), 48.0 %

3 Equinox 3.30 x64mp        : 3003  203 (+ 39,=108,- 56), 45.8 %

Mars 3.42 x64                 : 203 (+ 39,=108,- 56), 45.8 %

descriptionMars tests - Page 3 EmptyRe: Mars tests

more_horiz
Permissions in this forum:
You cannot reply to topics in this forum