Life In 19x19

Ten matches (10 games each) between Katago (kata1-b18c384nbt-uec) and Gnugo
maxVisits = 1000 for katago (W), handicaps between H10 and H19.
GoGui twogtp, chinese rule, there were no error, no duplicate games, KG and Gnugo always agreed on the score.

H10 : 60% wins for KG
H11 : 40% wins for KG (and ...

Ten matches (10 games each) between Katago (kata1-b18c384nbt-uec) and Gnugo
maxVisits = 10 for katago (W), handicaps between H10 and H19.
GoGui twogtp, chinese rule, there were no error, no duplicate games, KG and Gnugo always agreed on the score.

H10 : 70% wins for KG
H11 : 70% wins for KG
H12 ...

you should be able to play on larger sizes by recompiling: Thank you for your answer, I'm very grateful for GoGUI (twogtp), I like it a lot, and I use it often.
I don't know how to compile or recompile.
I'll wait for future GoGUI or katago updates (or an already compiled, recompiled, or "ready to ...

Thank you very much, it works well :tmbup:
size 25
I'm looking forward to trying handicap games on big boards.

For the moment, I get "size error" messages for sizes bigger than 25, it's probably a katago version problem.
and...I guess you have to be a very special user to play 37x37 games with ...

And wrote: could you make a match on 37x37 boards?

Sorry, no. But if someone is interested in tests on smaller boards (from 4(?) up to 19), I'll do it.

9x9
50 games test between 40b and 18b (b40c256-s11840935168-d2898845681 , b18c384 , 1s/move, GPU: GTX1080 , komi 7.5,
25 times W for each, TwoGtp V1.5.1, all games by resignation, no errors, 3 duplicate games, so, 47 games used for stats.
18b wins 32-15

9x9
50 games test between 60b and 18b ...

50 games test between b60 and b18
(1s/move, GPU: GTX1080 , komi 7.5, b60c320-s6782286336-d3070935549 , b18c384nbt-uec)
25 times W for each, TwoGtp V1.5.1, all games by resignation, no errors, no duplicate games.
b18 wins 43-7

Stats and details

50 games test between b40 and b18 (1s/move, GPU: GTX1080)
25 times W for each, TwoGtp V1.5.1, all games by resignation, no errors, no duplicate games.
b18 wins 36-14

Stats and details

... when do you intend to make a new optimized network 60b?
... When 60b is stronger than 40b at time parity. This result is probably not directly useable, but at time parity (0.5 s/move, laptop with GTX1660Ti)
b40s111d271 wins 80-20 against b60c320s583 (twogtp, 100 games, average length ~180 ...

100 games between KG(40b) and KG(60b) at 1 sec/move (each 50 times B, 50 times W)
Result : 40b wins 77-23 (consistent with the 20 game test at 2sec/move)
Details :
40vs60.jpg
12 games (9x9) between KG(60b) and gtp4zen (KG always W, komi 0)
Result KG wins 10-2 (with ~10 times less thinking time ...

20 game test between katago(40b) and katago(60b) at time parity (2s/move, GTX 1080, twogtp 1.5.1, no error, no duplicate game)
40b wins 15-5
Details :
katago40_katago60.jpg


20 game test between katago(60b) and gnugo , 9x9 , H2, komi 0
(5 duplicate games, katago wins 13-2)
Details :
katago60b ...

10 games between KG 40b-s1003 and KG 60b-s1003 (each 5 times as W, 5 times as B)
time parity : 5 sec/move , gogui-twogtp 1.5.1, GPU: 1x GTX 1660-Ti

average length : 245 moves/game
average time : 628sec/game (for 60b) 617sec/game (for 40b)
no error, no duplicate game, all games by resignation.
Score ...

It was komi 0 and chinese rules, but you're right about LZ and handicap.

According to THIS POST , Leela Zero #125, with 3000-4000 playouts, was already much stronger than 2p.

It usually got to 3000-4000 playouts before hitting the time limit. Baolong also thinks that if he could exploit bugs, he could maybe win one or two games out of a hundred. He thinks this ...

9x9 game, H2, komi 7.5

B: CrazyStoneDL 5d
W: BadukAI, 10s/move, Samsung s20+, network : b40c256-s7087...

Result : W+R
https://lifein19x19.com/download/file.php?id=13908
BadukAI wins more than 50% of these games against CSDL (5d, H2, komi 7.5) but usually loses with H2, komi 0.

What a wonderful ...