Engine Tournament

Cyan · **#81**

Please add DarkGo! It's a neural network at around 1d level.
https://pjreddie.com/darknet/darkgo-go-in-darknet/
How to run in GTP: https://groups.google.com/forum/#!topic ... BNZTkMmUHI

Also you can try Hiratuka 10.38B since you're using GPU anyway.

Pachi can use neural networks as well, if you're not using them already.
https://github.com/pasky/pachi

as0770 · **#82**

Cyan wrote:

Please add DarkGo! It's a neural network at around 1d level.
https://pjreddie.com/darknet/darkgo-go-in-darknet/
How to run in GTP: https://groups.google.com/forum/#!topic ... BNZTkMmUHI

Also you can try Hiratuka 10.38B since you're using GPU anyway.

Pachi can use neural networks as well, if you're not using them already.
https://github.com/pasky/pachi

Thanks for the hints, I had Pachi with neural networks on my list but I wasn't aware of DarkGo.

Unfortunately Hiratuka 10.38B doesn't run on my Linux machine because of a cuda error...

Cyan · **#83**

as0770 wrote:

Thanks for the hints, I had Pachi with neural networks on my list but I wasn't aware of DarkGo.

Unfortunately Hiratuka 10.38B doesn't run on my Linux machine because of a cuda error...

You're welcome! Also are you making sure that the programs are using as close to two hours as possible? For example you can add the argument -t _7200 to pachi to tell it to use two hours. I've noticed that in gogui-twogtp you're giving the engines two hours (i.e. -time 120) but you're only giving ray one hour (i.e. --time 3600). FYI rn-4.6.1 has just been released.

as0770 · **#84**

Cyan wrote:

Also are you making sure that the programs are using as close to two hours as possible? For example you can add the argument -t _7200 to pachi to tell it to use two hours. I've noticed that in gogui-twogtp you're giving the engines two hours (i.e. -time 120) but you're only giving ray one hour (i.e. --time 3600). FYI rn-4.6.1 has just been released.

I use Ray without the --time 3600 argument, it was a typo in the configuration section and I just corrected it. Ray and Pachi seem to work well with the gtp time command, I observe all games also in matter of time consumption. The engines with fixed playouts use in average 48 seconds per move after 50-100 moves. For the games against DarkForrest I measure how much each engine is slowed down by the CPU usage of DarkForrest and give it more time. So I think it is as fair as possible

as0770 · **#85**

Update League A with DarkGo, it won 1 against Pachi and Hiratuka, and 3 against DarkForrest, next Update will be Pachi DCNN.

League A:with

Code:

    1. Leela 0.10.0                 22/24
    2. Rayon 4.6.0                  19/24
    3. Oakfoam 0.2.1 NG-06          18/24
    4. Hiratuka 10.37B (CPU)        10/24
    5. DarkForest v2 MCTS 1.0        7/24
    6  DarkGo 1.0                    5/24
    7. Pachi 11.99                   3/24

Configuration:

League B:

Code:

    1. Pachi 11.99                  23/24
    2. MoGo 4.86                    19/24
    3. deltaGo 1.0.0                14/24
    4. Fuego 1.1                    13/24
    5. Michi C-2 1.4.2               8/24
    6. Orego 7.08                    5/24
    7. GNU Go 3.8                    2/24

Configuration:

League C:

Code:

    1. GNU Go 3.8                   24/28
    2. Hara 0.9                     18/28
    3. Dariush 3.1.5.7              16/28
    4. Indigo 2009                  15/28
    5. Matilda 1.24                 15/28
    6. Aya 6.34                     11/28
    7. Fudo Go 3.0                  11/28
    8. JrefBot 081016-2022           2/28

Configuration:

League D:

Code:

    1. JrefBot 081016-2022          25/28
    2. Iomrascálaí 0.3.2            21/28
    3. Crazy Patterns 0008-13       19/28
    4. Marcos Go 1.0                18/28
    5. AmiGo 1.8                    16/28
    6. Stop 0.9-005                  6/28
    7. CopyBot 0.1                   4/28
    8. Brown 1.0                     3/28

Configuration:

Links:

Best,
Alex

as0770 · **#86**

Update League A with Pachi DCNN.

League A:

Code:

    1. Leela 0.10.0                 22/24
    2. Rayon 4.6.0                  19/24
    3. Oakfoam 0.2.1 NG-06          18/24
    4. Hiratuka 10.37B (CPU)         9/24
    5. DarkForest v2 MCTS 1.0        7/24
    6  DarkGo 1.0                    5/24
    7. Pachi DCNN 11.99              4/24

Configuration:

League B:

Code:

    1. Pachi 11.99                  23/24
    2. MoGo 4.86                    19/24
    3. deltaGo 1.0.0                14/24
    4. Fuego 1.1                    13/24
    5. Michi C-2 1.4.2               8/24
    6. Orego 7.08                    5/24
    7. GNU Go 3.8                    2/24

Configuration:

League C:

Code:

    1. GNU Go 3.8                   24/28
    2. Hara 0.9                     18/28
    3. Dariush 3.1.5.7              16/28
    4. Indigo 2009                  15/28
    5. Matilda 1.24                 15/28
    6. Aya 6.34                     11/28
    7. Fudo Go 3.0                  11/28
    8. JrefBot 081016-2022           2/28

Configuration:

League D:

Code:

    1. JrefBot 081016-2022          25/28
    2. Iomrascálaí 0.3.2            21/28
    3. Crazy Patterns 0008-13       19/28
    4. Marcos Go 1.0                18/28
    5. AmiGo 1.8                    16/28
    6. Stop 0.9-005                  6/28
    7. CopyBot 0.1                   4/28
    8. Brown 1.0                     3/28

Configuration:

Links:

Best,
Alex

Shoreline · **#87**

How do you guys gets the Orego 7.08 running?

I downloaded Jar file, unziped, or downloaded the 7.14 version release and javac compiled with no erros. However,
I can not get the software running in linux Ubuntu 14.04. Erros:
java -ea -Xmx1024M edu.lclark.orego.ui.Orego
Error: Could not find or load main class edu.lclark.orego.ui.Orego

No matter how I switched to src folder, or change to localhost.Orego or just Orego, it does not run

as0770 · **#88**

Shoreline wrote:

How do you guys gets the Orego 7.08 running?

I downloaded Jar file, unziped, or downloaded the 7.14 version release and javac compiled with no erros. However,
I can not get the software running in linux Ubuntu 14.04. Erros:
java -ea -Xmx1024M edu.lclark.orego.ui.Orego
Error: Could not find or load main class edu.lclark.orego.ui.Orego

No matter how I switched to src folder, or change to localhost.Orego or just Orego, it does not run

I had problems running later versions because of similar problems. v7.08 works by adding:

Main-Class: orego.ui.Orego

to /META-INF/MANIFEST.MF inside the .jar file.

Then execute the .jar file by:

java -jar /path to/orego-7.08.jar threads=1 grace

If you get later versions to work let me know

Best,
Alex

Shoreline · **#89**

as0770 wrote:

Shoreline wrote:

How do you guys gets the Orego 7.08 running?

I downloaded Jar file, unziped, or downloaded the 7.14 version release and javac compiled with no erros. However,
I can not get the software running in linux Ubuntu 14.04. Erros:
java -ea -Xmx1024M edu.lclark.orego.ui.Orego
Error: Could not find or load main class edu.lclark.orego.ui.Orego

No matter how I switched to src folder, or change to localhost.Orego or just Orego, it does not run

I had problems running later versions because of similar problems. v7.08 works by adding:

Main-Class: orego.ui.Orego

to /META-INF/MANIFEST.MF inside the .jar file.

Then execute the .jar file by:

java -jar /path to/orego-7.08.jar threads=1 grace

If you get later versions to work let me know

Best,
Alex

Thanks, I got it running like you said. However, I found it is very weak bot at 19x19. How come this Orego 7.08 ranks so high against gnugo?
It appears that it is several stones weaker than gnugo.

as0770 · **#90**

Shoreline wrote:

Thanks, I got it running like you said. However, I found it is very weak bot at 19x19. How come this Orego 7.08 ranks so high against gnugo?
It appears that it is several stones weaker than gnugo.

Here it won 4-0 against GnuGo 3.8, the games where all close. I think one has to play much more games to see which one is stronger.

Shoreline · **#91**

I wonder if the java version or OS made the difference. I am using a linux machine,
dual core, 8G of memory, quite low end machine. The java verison I am using is:
$ java -version
java version "1.7.0_45"
Java(TM) SE Runtime Environment (build 1.7.0_45-b18)
Java HotSpot(TM) 64-Bit Server VM (build 24.45-b08, mixed mode)

I would say in my setting, the Orego version 7.08 is probably more than 9 stones weaker than gnugo, it probably ranks below 20 kyu in DGS.

Cyan · **#92**

Here are estimates of goratings elo of different bots (from various sources and also my own testing). GPU is assumed to be high-end (e.g. 1080 Ti). Good to put things into perspective.

Code:

AlphaGo       ~4800 elo
Ke Jie        ~3700 elo
DeepZenGo     ~3600 elo
Average pro   =3000 elo
Ray GPU       ~2700 elo
Leela GPU     ~2500 elo
Hira GPU      ~2300 elo
Leela         ~2100 elo
Zen 6         ~2100 elo
Hira          ~1900 elo
DarkGo        ~1600 elo
Pachi         ~1300 elo
1k player     ~1000 elo

as0770 · **#93**

Shoreline wrote:

I wonder if the java version or OS made the difference. I am using a linux machine,
dual core, 8G of memory, quite low end machine. The java verison I am using is:
$ java -version
java version "1.7.0_45"
Java(TM) SE Runtime Environment (build 1.7.0_45-b18)
Java HotSpot(TM) 64-Bit Server VM (build 24.45-b08, mixed mode)

I would say in my setting, the Orego version 7.08 is probably more than 9 stones weaker than gnugo, it probably ranks below 20 kyu in DGS.

This is my Java version:

java -version
openjdk version "1.8.0_131"
OpenJDK Runtime Environment (build 1.8.0_131-8u131-b11-2ubuntu1.16.04.3-b11)
OpenJDK 64-Bit Server VM (build 25.131-b11, mixed mode)

Does Orego use all the available time? Maybe you use a time setting that Orego doesn't support and it moves faster.

Shoreline · **#94**

as0770 wrote:

Shoreline wrote:

I wonder if the java version or OS made the difference. I am using a linux machine,
dual core, 8G of memory, quite low end machine. The java verison I am using is:
$ java -version
java version "1.7.0_45"
Java(TM) SE Runtime Environment (build 1.7.0_45-b18)
Java HotSpot(TM) 64-Bit Server VM (build 24.45-b08, mixed mode)

I would say in my setting, the Orego version 7.08 is probably more than 9 stones weaker than gnugo, it probably ranks below 20 kyu in DGS.

This is my Java version:

java -version
openjdk version "1.8.0_131"
OpenJDK Runtime Environment (build 1.8.0_131-8u131-b11-2ubuntu1.16.04.3-b11)
OpenJDK 64-Bit Server VM (build 25.131-b11, mixed mode)

Does Orego use all the available time? Maybe you use a time setting that Orego doesn't support and it moves faster.

Ｔhe Orego in my machine moves bleedingly fast, may be 1 second per move. It can play 5 or 10 different games within 10 seconds.

How can I change the time setting in GTP command? Maybe that is the issue.

Checked GTP manual and list_commands of Orego, found 2 commands is related "time_settings" "time_left". Not sure how to use them in games.

as0770 · **#95**

Shoreline wrote:

as0770 wrote:

Shoreline wrote:

I wonder if the java version or OS made the difference. I am using a linux machine,
dual core, 8G of memory, quite low end machine. The java verison I am using is:
$ java -version
java version "1.7.0_45"
Java(TM) SE Runtime Environment (build 1.7.0_45-b18)
Java HotSpot(TM) 64-Bit Server VM (build 24.45-b08, mixed mode)

I would say in my setting, the Orego version 7.08 is probably more than 9 stones weaker than gnugo, it probably ranks below 20 kyu in DGS.

This is my Java version:

java -version
openjdk version "1.8.0_131"
OpenJDK Runtime Environment (build 1.8.0_131-8u131-b11-2ubuntu1.16.04.3-b11)
OpenJDK 64-Bit Server VM (build 25.131-b11, mixed mode)

Does Orego use all the available time? Maybe you use a time setting that Orego doesn't support and it moves faster.

Ｔhe Orego in my machine moves bleedingly fast, may be 1 second per move. It can play 5 or 10 different games within 10 seconds.

How can I change the time setting in GTP command? Maybe that is the issue.

I use game in x for my testing because this works with most engines. The gtp command is:

time_settings 3600 0 0

for a game in 60 minutes. But Orego uses some more time as allowed.

Shoreline · **#96**

as0770 wrote:

I use game in x for my testing because this works with most engines. The gtp command is:

time_settings 3600 0 0

for a game in 60 minutes. But Orego uses some more time as allowed.

THanks for the information. I think you nailed the issue already correctly. I tried your setting and found Orego runs
about 40 seconds per move and it is much stronger player. Now I settled down to
time_settings 600 60 0 roughly at 10 to 20 seconds per move in my machine.

Here is the DGS bot "oregobot" with same setting:

http://www.dragongoserver.net/ratinggraph.php?uid=99178

We will see how strong this bot can be. My internal testing says that this bot should at least get teen kyu level.

as0770 · **#97**

Update League B with Pachi DCNN, DarkGo and Ray 9.0.1

League A:

Code:

    1. Leela 0.10.0                 22/24
    2. Rayon 4.6.0                  19/24
    3. Oakfoam 0.2.1 NG-06          18/24
    4. Hiratuka 10.37B (CPU)         9/24
    5. DarkForest v2 MCTS 1.0        7/24
    6  DarkGo 1.0                    5/24
    7. Pachi DCNN 11.99              4/24

Configuration:

League B:

Code:

    1. DarkGo 1.0                   29/32
    2. Pachi DCNN 11.99             27/32
    3. Ray 9.0.1                    25/32
    4. MoGo 4.86                    19/32
    5. deltaGo 1.0.0                16/32
    6. Fuego 1.1                    13/32
    7. Michi C-2 1.4.2               8/32
    8. Orego 7.08                    5/32
    9. GNU Go 3.8                    2/32

Configuration:

League C:

Code:

    1. GNU Go 3.8                   24/28
    2. Hara 0.9                     18/28
    3. Dariush 3.1.5.7              16/28
    4. Indigo 2009                  15/28
    5. Matilda 1.24                 15/28
    6. Aya 6.34                     11/28
    7. Fudo Go 3.0                  11/28
    8. JrefBot 081016-2022           2/28

Configuration:

League D:

Code:

    1. JrefBot 081016-2022          25/28
    2. Iomrascálaí 0.3.2            21/28
    3. Crazy Patterns 0008-13       19/28
    4. Marcos Go 1.0                18/28
    5. AmiGo 1.8                    16/28
    6. Stop 0.9-005                  6/28
    7. CopyBot 0.1                   4/28
    8. Brown 1.0                     3/28

Configuration:

Links:

Best,
Alex

anpd · **#98**

Is there a version of Pachi with neural networks available for Windows somewhere? Do you use patterns with that build?

q30 · **#99**

Cyan wrote:

Please add DarkGo! It's a neural network at around 1d level.
https://pjreddie.com/darknet/darkgo-go-in-darknet/
How to run in GTP: https://groups.google.com/forum/#!topic ... BNZTkMmUHI

I had downloaded DarkGo.
I had downloaded also CUDA. Unfortunately CUDA requires NVIDIA driver v.375 or higher, but v.375 driver tells, that with my video card can work drivers v.340 or lower (card has 32 CUDA cores and 256 MB memory)...
I had downloaded also go.train and go.weights. And there is a lot of cfg files, but it isn't any description. Can anybody tell me, how is it possible to run darknet in gtp mode with maximum strong (including pondering, neural network), using 2 CPU cores, up to 8 GB memory and up to 2 min / move time?

as0770 · **#100**

Unfortunately AQ can't play my tournament because of some issues: AQ always ponders and it uses the GPU very effectively so that other engines get less GPU power. Beside that Rayon and Oakfoam don't work together with AQ because of a cuda memory error. It was possible to set up a match between Leela and AQ with ponder on for both engines where Leela reaches around 35% of it's playouts with full power. So I gave AQ 2h/game and Leela 4h/game. Still I can't control what is happening on the GPU during the games, but I think the conditions are as fair as possible. BTW DarkGo was able to win one of four games against AQ during my tests.

Update with the next Leela version will follow

Leela vs. AQ

Code:

    1. AQ 2.0.2                      5/8
    2. Leela 0.10.0                  3/8

Configuration:

League A:

Code:

    1. Leela 0.10.0                 22/24
    2. Rayon 4.6.0                  19/24
    3. Oakfoam 0.2.1 NG-06          18/24
    4. Hiratuka 10.37B (CPU)         9/24
    5. DarkForest v2 MCTS 1.0        7/24
    6  DarkGo 1.0                    5/24
    7. Pachi DCNN 11.99              4/24

Configuration:

League B:

Code:

    1. Pachi DCNN 11.99             25/28
    2. Ray 9.0.1                    25/28
    3. MoGo 4.86                    18/28
    4. deltaGo 1.0.0                16/28
    5. Fuego 1.1                    13/28
    6. Michi C-2 1.4.2               8/28
    7. Orego 7.08                    5/28
    8. GNU Go 3.8                    2/28

Configuration:

League C:

Code:

    1. GNU Go 3.8                   24/28
    2. Hara 0.9                     18/28
    3. Dariush 3.1.5.7              16/28
    4. Indigo 2009                  15/28
    5. Matilda 1.24                 15/28
    6. Aya 6.34                     11/28
    7. Fudo Go 3.0                  11/28
    8. JrefBot 081016-2022           2/28

Configuration:

League D:

Code:

    1. JrefBot 081016-2022          25/28
    2. Iomrascálaí 0.3.2            21/28
    3. Crazy Patterns 0008-13       19/28
    4. Marcos Go 1.0                18/28
    5. AmiGo 1.8                    16/28
    6. Stop 0.9-005                  6/28
    7. CopyBot 0.1                   4/28
    8. Brown 1.0                     3/28

Configuration:

Links:

Best,
Alex

Engine Tournament

Who is online