Is there any option to remove all dead stones?

[yssaya]

Hi, I run cgf2023 weight on CGOS. But It seems Sayuri passes without removing all dead stones. http://www.yss-aya.com/cgos/viewer.cgi?19x19/SGF/2023/08/24/876080.sgf Is there any option to remove all dead stones?

I use RTX 3090, cuda-11.2, Ubuntu 20.04.2, Ryzen 7 3700X 8-Core. Compile and command are

$ git clone https://github.com/CGLemon/Sayuri
$ cd Sayuri
$ git submodule update --init --recursive
$ cd ..
$ wget https://github.com/CGLemon/Sayuri/archive/refs/tags/cgf2023.tar.gz
$ tar xvf cgf2023.tar.gz
$ cd Sayuri-cgf2023
$ cp -p -r ../Sayuri/third_party ./
$ mkdir build && cd build
$ cmake .. -DBLAS_BACKEND=CUDA -DCMAKE_CUDA_ARCHITECTURES=75
$ make -j 6
$ ./Sayuri -t 4 -p 10000 -w ../network/cgf2023-swa.bin.txt

[CGLemon]

The engine can't play the move in the pass-alive and pass-dead area. So the engine can't remove all dead stones sometimes. I will add this feature into TODO llist.

[CGLemon]

Add support for ---capture-all-dead in the branch remove-dead-stones. Look like it works well. The version is v0.5. Old support for old netowk, such as UEC14 and CGF 2023 network.

[yssaya]

Wow thank you! I will try and run on CGOS.

It has won against CrazyStone stably. Thanks! http://www.yss-aya.com/cgos/19x19/cross/SayuriCGF2023_p10k.html

[CGLemon]

Thanks for testing! Look like I missed some conditions. Already fixed the capture bug in the same branch.

But this option is unsafe in some positions. Did Aya implement the same function? How did Aya implement it?

[yssaya]

Thanks! I have updated binary. Aya is If Tromp-Taylor score is losing, and there is a move except pass, never pass.

[CGLemon]

Fix capture-all-dead mode bug and push the code to the master branch. Now can capture two point eye group.

[yssaya]

Thank you! I updated with latest version, and ran sayuri on CGOS. http://www.yss-aya.com/cgos/19x19/cross/sayuri811_2035_p1k.html Sorry late reply, I can not login github for two-factor authentication fail.

[CGLemon]

Thank!

A little of off-topic, The Elo result shows the sayuri811_2035_p1k almost achieves LZ_05db_ELFv2_p800. But in my local testing, Sayuri is significantly weaker than LZ174 which is also significantly weaker than ELF v2 in the fair match game. I do not think the Elo score of Sayuri can be better than 3600. Is the Elo correct?

[yssaya]

CGOS result are only from around 100 games, so it is not so reliable. It ELO has +- 80 width. Or maybe it is because of chemistory. From CrazyStone-82-15po of view, there is no progress between 2035 and 1785. From LZ_275_b942_v100 of view, 2035 is much stronger than 1785.

Rank  Name                 Elo    +    −   Games
 948  sayuri811_2035_p1k   3623   75   75  199
1081  sayuri811_1785_p1k   3500   70   70  195
1110  SayuriCGF2023_p10k   3471   85   85  126
1116  sayuri811_1570_p1k   3465   82   82  148
1158  sayuri811_1375_p1k   3415   68   68  180
1251  sayuri811_1195_p1k   3326   64   64  173
1263  sayuri811_945_p10k   3315   80   80  113
1278  sayuri811_1275_p1k   3292   85   85  115
1292  sayuriCGF2023_p10k   3269   85   85  104
1340  sayuri011_rs11_p1k   3219   86   86  131
http://www.yss-aya.com/cgos/19x19/bayes.html

Cross-table of results for CrazyStone-82-15po
sayuri811_2035_p1k    3634     3 / 49     6.12
sayuri811_1785_p1k    3508     3 / 47     6.38
SayuriCGF2023_p10k    3472     4 / 31    12.90
sayuri811_1570_p1k    3457     4 / 36    11.11
sayuri811_1375_p1k    3419     9 / 51    17.65
sayuri811_945_p10k    3332    13 / 31    41.94
sayuri811_1195_p1k    3325    17 / 52    32.69
sayuri811_1275_p1k    3292     7 / 36    19.44
sayuriCGF2023_p10k    3269    17 / 33    51.52
sayuri811_945_p1k     3240    27 / 56    48.21
sayuri011_rs11_p1k    3184    15 / 19    78.95

Cross-table of results for LZ_275_b942_v100
sayuri811_2035_p1k    3633     3 / 29    10.34
sayuri811_1785_p1k    3507    25 / 60    41.67
SayuriCGF2023_p10k    3477    14 / 36    38.89
sayuri811_1570_p1k    3453    12 / 31    38.71
sayuri811_1375_p1k    3418    21 / 43    48.84
sayuri811_945_p10k    3327    18 / 34    52.94
sayuri811_1195_p1k    3326    33 / 55    60.00
sayuri811_1275_p1k    3292    23 / 25    92.00
sayuriCGF2023_p10k    3261    17 / 30    56.67
sayuri811_945_p1k     3240    37 / 55    67.27
sayuri011_rs11_p1k    3186     8 / 11    72.73