Life In 19x19
http://lifein19x19.com/

LZ's progression
http://lifein19x19.com/viewtopic.php?f=18&t=15718		 Page 12 of 21
Author:  jlt [ Wed Dec 19, 2018 10:12 am ]
Post subject:  Re: LZ's progression
Oops, I indeed made a confusion between two names. The game rn195_lz195_2sec_c_LZ_B-5.sgf was unfinished, so it's impossible to count.
Author:  zakki [ Thu Dec 20, 2018 7:17 pm ]
Post subject:  Re: LZ's progression
Sometime RLO plays PASS at losing game, because they cannot find any good move, but win rate is bigger than a resign threshold.
Author:  And [ Sat Dec 22, 2018 1:57 am ]
Post subject:  Re: LZ's progression
new network 15x192 from bjiyxo is stronger lz157: 92297ff2 VS d351f06e 260 : 144 (64.36%)
http://zero.sjeng.org
Author:  Vargo [ Sat Dec 22, 2018 3:39 am ]
Post subject:  Re: LZ's progression
100 game match between #196 and #157 at short time parity
1 min per side per game, komi 7.5 no pondering, 1x1080, LZ0.16, twogtp V1.5.0
0 duplicate game, 0 error, all games by resignation.
#196 wins 55:45
196 is now clearly stronger than 157, even at short time settings (both networks spending 0.5-0.6 sec/move)

The 100 games :
Attachment:
157_196_1minute.zip [81.87 KiB]
Downloaded 380 times
Detailed stats :
Attachment:
Pas de nom.jpg
Pas de nom.jpg [ 215.24 KiB | Viewed 7964 times ]

And wrote:
new network 15x192 from bjiyxo is stronger lz157: 92297ff2 VS d351f06e 260 : 144 (64.36%)
Very interesting !

I've launched a 100 game match at 1 min per game between #196 and this new network 92297ff2, result later.
Author:  Vargo [ Sat Dec 22, 2018 6:36 am ]
Post subject:  Re: LZ's progression
Here we are :
#196 (40x256) vs 92297ff2 (15x192) at short time parity : 1 min per side per game, komi 7.5, no pondering, 1x1080, LZ0.16, twogtp V1.5.0
0 duplicate game, 0 error, all games by resignation.

This time, #196 is not the winner :

92297ff2 v 196 : 63:37

It was to be expected, because 92297ff is stronger than #157.

The 100 games :
Attachment:
196_15x192black_1m.zip [82.36 KiB]
Downloaded 387 times
The stats :
Attachment:
Pas de nom.jpg
Pas de nom.jpg [ 685.42 KiB | Viewed 7915 times ]
Author:  And [ Sat Dec 22, 2018 10:14 am ]
Post subject:  Re: LZ's progression
I wonder if 92297ff2 is stronger than elf? (time parity)
Author:  splee99 [ Sat Dec 22, 2018 12:26 pm ]
Post subject:  Re: LZ's progression
Vargo wrote:
Here we are :
#196 (40x256) vs 92297ff2 (15x192) at short time parity : 1 min per side per game, komi 7.5, no pondering, 1x1080, LZ0.16, twogtp V1.5.0
0 duplicate game, 0 error, all games by resignation.

This time, #196 is not the winner :



What is (maybe) unexpected, is that #196 is not as strong as #195 in some aspects. I had run some handicap games on my computer and got this feeling.
Author:  splee99 [ Sat Dec 22, 2018 11:23 pm ]
Post subject:  Re: LZ's progression
Just run two time parity games with 20 seconds per move between ELF and #196 and #196 won both games. It looks like to me that the 40b weight is indeed superior in handling large groups and taking advantage of a ko fight. However, I don't understand why #196 lost to the new 15b weight in the above tests. Maybe the game time is too short ( 1 minute )?
Author:  Vargo [ Sat Dec 22, 2018 11:47 pm ]
Post subject:  Re: LZ's progression
splee99 wrote:
I don't understand why #196 lost to the new 15b weight in the above tests. Maybe the game time is too short ( 1 minute )?
I thinnk you're right.

I've run two more matches #196 v 92297ff at time parity :
the three results are :

1min/game (corresponding to visits=~300 and visits=~1100) #196 wins 37% (100 games, 1x1080)
5min/game (corresponding to visits=~1000 and visits=~3600) #196 wins 60% (20 games, 1x 1080)
5min/game with 2 GPU (corresponding to visits=~3000 and visits=~10800) #196 wins 65% (20 games, 2x 1080Ti)

I've done several tests to determine time-visits ratio, the results vary a lot, these correspondences are approximate.

games and stats for 2 gpus :
Attachment:
196_v_92297_5min_2gpu.zip [19.19 KiB]
Downloaded 372 times
Attachment:
2gpu.jpg
2gpu.jpg [ 247.26 KiB | Viewed 7796 times ]
Author:  And [ Sun Dec 23, 2018 2:36 am ]
Post subject:  Re: LZ's progression
20 games is enough for a confident conclusion? I doubt, especially considering the slight advantage
Author:  Vargo [ Sun Dec 23, 2018 8:29 am ]
Post subject:  Re: LZ's progression
And wrote:
20 games is enough for a confident conclusion?
20 games is not nearly enough, but each of these 20 game match is almost 3 hours...
After several such matches, we can hopefully see a trend.

Ideally, this kind of match should be... 400 games, for example ;-)
To do this, one possiblity would be to find 20 people OK to run a 20 game match with the same parameters including -v 1000 for #196 and -v 3600 for #92297ff, so, we would have 400 games at time parity, and at the same strength, I think -v 1000 and -v 3600 should be at near time parity on any machine (?)
But it would probably be more useful for people to donate their GPU time to autogtp.

Anyway, I ran 40 more games on each machine, and after 60 games, #196's winrate has gone down.

5min/game (corresponding to visits=~1000 and visits=~3600) #196 wins 52% (31:29, 1x 1080)
5min/game with 2 GPU (corresponding to visits=~3000 and visits=~10800) #196 wins 52% too (31:29, 2x 1080Ti)

I'll upload stats and/or games if someone wants them.
Author:  And [ Sun Dec 23, 2018 8:41 am ]
Post subject:  Re: LZ's progression
thanks! after such a result (52%) is not the fact that #196 is stronger
Author:  And [ Mon Dec 24, 2018 3:03 am ]
Post subject:  Re: LZ's progression
Strange, why there is no new network 92297ff2 in the table "Best Network Hash"? (http://zero.sjeng.org) for example, under the number 197. it is logical and convenient
Author:  Tryss [ Mon Dec 24, 2018 3:43 am ]
Post subject:  Re: LZ's progression
Because this network is not trained the same way.
Author:  dfan [ Mon Dec 24, 2018 5:28 am ]
Post subject:  Re: LZ's progression
And wrote:
Strange, why there is no new network 92297ff2 in the table "Best Network Hash"? (http://zero.sjeng.org) for example, under the number 197. it is logical and convenient

It's a 15-block network, which played a test match against the previous best 15-block network, not against the current best 40-block network.
Author:  And [ Mon Dec 24, 2018 5:53 am ]
Post subject:  Re: LZ's progression
In the table there are other networks 15x192 (weaker) and from bjiyxo. where then to look for this network, if not in this table? the strongest 15x192!
Author:  dfan [ Mon Dec 24, 2018 6:02 am ]
Post subject:  Re: LZ's progression
Each entry in the table is stronger than the entry before it (reading from bottom to top). To put this latest 15-block network one in, they'd have to either break that rule or change all the numbers or call this network 157a or something. Maybe they should have used a 2-dimensional numbering scheme.

I agree that it would be nice to have a place to easily find the best 15-block network.
Author:  And [ Mon Dec 24, 2018 6:12 am ]
Post subject:  Re: LZ's progression
the table is made for the convenience of people, not for the report! besides, it has a Hash, Size, Elo, Games, Training. if a person came to this site, then this is not a random visitor, and he will understand what's what
Author:  dfan [ Mon Dec 24, 2018 6:52 am ]
Post subject:  Re: LZ's progression
I was just trying to answer your original question of why it was not in the table. If you would like the format of the table to be changed, you could try opening an issue on the Leela Zero GitHub page: https://github.com/gcp/leela-zero/issues
Author:  And [ Mon Dec 24, 2018 7:06 am ]
Post subject:  Re: LZ's progression
Formal reasons are clear to me, but thanks anyway for your answers!
Page 12 of 21 All times are UTC - 8 hours [ DST ]
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group
http://www.phpbb.com/