10-games match at H7 between gnugo 3.8 and LZ_variable-komi 0.3 with weights #157
LZ wins 9-1 (7 times W+R, 1 time B+R, 2 times W+xx.5)
1x1080, twogtp v1.4.10, --visits = 3201 and no pondering for LZvk, komi 7.5 (so, it's more like H6) , no additional commands used (-r, --max-wr , --min-wr ,--mid-wr etc.)
10-games match at H8 between gnugo 3.8 and LZ variable-komi 0.3 with weights #157
LZ wins 6-4 (3 times W+R, 4 times B+R, 3 times W+xx.5)
(same specs)
10-games match at H9 between gnugo 3.8 and LZ variable-komi 0.3 with weights #157
LZ wins 7-3 (7 times W+xx.5 , 3 times B+xx.5)
(same specs, but for the added command -r 0 for LZ_vk)
I should have added the -r 0 command (= never resign !) for the H7 and H8 games, because LZ_vk probably resigned too soon in some of them.
Anyway, very nice wins !
LZ variable komi : handicap games
-
Vargo
- Lives in gote
- Posts: 337
- Joined: Sat Aug 17, 2013 5:28 am
- GD Posts: 0
- Has thanked: 22 times
- Been thanked: 97 times
Re: LZ variable komi : handicap games
10-games match at H6 between LZ_vk_#157 and Pachi 12.10 (quad-core i7-6700K, 1x1080, 5 min/game)
--noponder -r 0 for LZ, no parameter for Pachi.
twogtp(1.4.10) commands : -time 5m -komi 7.5 -handicap 6 -auto
LZ_vk#157 wins 6-4 (6 times W+R, 4 times B+xx.5)
--noponder -r 0 for LZ, no parameter for Pachi.
twogtp(1.4.10) commands : -time 5m -komi 7.5 -handicap 6 -auto
LZ_vk#157 wins 6-4 (6 times W+R, 4 times B+xx.5)
-
Vargo
- Lives in gote
- Posts: 337
- Joined: Sat Aug 17, 2013 5:28 am
- GD Posts: 0
- Has thanked: 22 times
- Been thanked: 97 times
Re: LZ variable komi : handicap games
10-games match at H9 between gnugo (B) and Pachi 12.10 (W) (quad-core i7-6700K, 1x1080)
no parameter for gnugo, -t 5 --dcnn maximize_score for Pachi.
twogtp(1.4.10) commands : -komi 0.5 -handicap 9 -auto
Pachi wins 6-4 (3 times W+R, 3 times W+xx.5, 3 times B+xx.5, 1 time B+R)
Impressive... Pachi seems to deal very well with high handicap.
no parameter for gnugo, -t 5 --dcnn maximize_score for Pachi.
twogtp(1.4.10) commands : -komi 0.5 -handicap 9 -auto
Pachi wins 6-4 (3 times W+R, 3 times W+xx.5, 3 times B+xx.5, 1 time B+R)
Impressive... Pachi seems to deal very well with high handicap.
-
Uberdude
- Judan
- Posts: 6727
- Joined: Thu Nov 24, 2011 11:35 am
- Rank: UK 4 dan
- GD Posts: 0
- KGS: Uberdude 4d
- OGS: Uberdude 7d
- Location: Cambridge, UK
- Has thanked: 436 times
- Been thanked: 3718 times
Re: LZ variable komi : handicap games
How strong is pachi, 3d KGS or so?Vargo wrote:10-games match at H6 between LZ_vk_#157 and Pachi 12.10 (quad-core i7-6700K, 1x1080, 5 min/game)
--noponder -r 0 for LZ, no parameter for Pachi.
-
Vargo
- Lives in gote
- Posts: 337
- Joined: Sat Aug 17, 2013 5:28 am
- GD Posts: 0
- Has thanked: 22 times
- Been thanked: 97 times
Re: LZ variable komi : handicap games
Yes, I think so.Uberdude wrote: 3d KGS or so?
Some Sabaki games against LZ0.15 at 1D, 2D, 3D, 4D and 5D (according to the ranking of LZ)
Laptop with GTX965M, and i7-6700HQ
LZ parameters : --visits=3201 -r 10 --noponder
Pachi parameters : -t =10000:30000 --dcnn
Pachi won against
1D (network 55bf5)
2D (network ded4)
3D (network 714a9)
4D (network e1156)
but lost against
5D (network b3b00)
Pachi seems very strong at high handicap : it won against gnugo at a real H9 (komi 0.5)
Maybe, it uses dynamic komi (I've seen something called dynkomi, somewhere)
There's an Android app too, but it's very slow.