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

Pachi UCT problems
http://www.lifein19x19.com/viewtopic.php?f=18&t=14533		 Page 1 of 1
Author:  garde [ Mon Sep 11, 2017 8:53 pm ]
Post subject:  Pachi UCT problems
I'm using the GUI that came with Fuego, and started to use Pachi UCT. I see that it somehow has some analysis mode, and a command to make it stronger. But, I can't find any way to put in the code. I opened the pachi.exe (pachi-11.00-win32 is the one im using) but anything i type comes up as unknown. So I try to go into the Shell commands, and same result. I see a whole bunch of stuff in the Analyze commands from where I downloaded pachi-pachi-11.00-retsugen-9f8c498 (Like a directory?). So here are the issues:

- I want it to play faster
- I want to enter in the commands
- I want to use analysis mode

The README file that comes with Pachi UCT itself says that I can make it faster with the command /pachi -t _1200 threads=8,max_tree_size=3072 , and by changing it. Could I possibly go higher? Or will that just make it slower? And this is what it says about analysis:

Analysis
--------

Pachi can also help you analyze your games by being able to provide
its opinion on various positions. The user interface is very rudimentary,
but the ability is certainly there.

There are currently several Pachi interfaces provided for this purpose.

Winrate Development
~~~~~~~~~~~~~~~~~~~

Pachi can evaluate all moves within a given game and show how
the winrates for both players evolved - i.e. who was winning at which
game stage. This is implemented using the `tools/sgf-analyse.pl` script.
See the comment on top of the script about its usage.

Move Ranking
~~~~~~~~~~~~

Pachi can evaluate all available moves in a given situation
and for each give a value between 0 and 1 representing perceived
likelihood of winning the game if one would play that move. I.e. it can
suggest which moves would be good and bad in a single given situation.

To achieve the latter, note the number of move at the situation you
want to evaluate and run the `tools/sgf-ratemove.sh` script.
See the comment on top of the script about its usage.

Pattern Move Hinting
~~~~~~~~~~~~~~~~~~~~

Pachi can show instantenous pattern-based move suggestions very much
like for example Moyo Go Studio (though of course without a GUI).
You can use the Move Ranking method above (tools/sgf-ratemove.sh),
but pass it an extra parameter '-e patternplay'.

If you don't have Pachi, but think you could help, here are the links I have:

http://repo.or.cz/w/pachi.git/snapshot/ ... gen.tar.gz (Directory)

http://pachi.or.cz/pachi-11.00-win32.7z (PachiUCT itself)

Sorry, I'm very new to this.
Is there also anyway to create a book of games that the engine can use? I want to make PachiUCT aggressive, if possible. (Maybe I'm asking WAY to much for one post.)
Author:  emeraldemon [ Tue Sep 12, 2017 9:57 am ]
Post subject:  Re: Pachi UCT problems
In gogui click Tools -> GTP Shell. This pops up a text shell where you can enter commands directly to pachi.
In the GTP prompt try
time_settings 60 5 1
(60 seconds plus one 5-second byo-yomi)

The gtp output will already show you in text what pachi thinks of its winrate and most likely sequence.
You can try live "thinking graphics" with tools -> analyze commands -> Live gfx best squence (or the other choices).

I haven't tried the analysis scripts you mentioned, but the extension makes me think perl scripts, which you will need to install perl and run from a terminal, not gogui, I'm guessing.
Author:  garde [ Mon Sep 18, 2017 6:58 pm ]
Post subject:  Re: Pachi UCT problems
Thanks. I've now just downloaded Perl. How exactly can I use it?
Author:  as0770 [ Mon Sep 25, 2017 7:41 am ]
Post subject:  Re: Pachi UCT problems
garde wrote:
- I want it to play faster


You can setup a timelimit within GoGui under Game -> Game Info (I have the german version so I am not sure that the menu entries are named correct by me...)

garde wrote:
- I want to enter in the commands


This "pachi -t _1200 threads=8,max_tree_size=3072" are command line arguments, under windows you have to start the pachi like this: pachi.exe -t _1200 threads=8,max_tree_size=3072

garde wrote:
- I want to use analysis mode


Open the GTP window of GoGui with F9 or by menu. Here you find the engineoutput you want to see:

Code:
Move:   3  Komi: 6.5  Handicap: 0  Captures B: 0 W: 0
      A B C D E F G H J K L M N O P Q R S T        A B C D E F G H J K L M N O P Q R S T 
    +---------------------------------------+    +---------------------------------------+
19 | . . . . . . . . . . . . . . . . . . . | 19 | : : : : : : : : : : : : : : : : : : : |
18 | . . . . . . . . . . . . . . . . . . . | 18 | : : : : : : : : : : : : : : : : : : : |
17 | . . . . . . . . . . . . . . . . . . . | 17 | : : : : : : : : : : : : : : : : : : : |
16 | . . . . . . . . . . . . . . . . . . . | 16 | : : : : : : : : : : : : : : : : : : : |
15 | . . . . . . . . . . . . . . . . . . . | 15 | : : : : : : : : : : : : : : : : : : : |
14 | . . . . . . . . . . . . . . . . . . . | 14 | : : : : : : : : : : : : : : : : : : : |
13 | . . . . . . . . . . . . . . . . . . . | 13 | : : : : : : : : : : : : : : : : : : : |
12 | . . . . . . . . . . . . . . . . . . . | 12 | : : : : : : : : : : : : : : : : : : : |
11 | . . . . . . . . . . . . . . . . . . . | 11 | : : : : : : : : : : : : : : : : : : : |
10 | . . . . . . . . . . . . . . . . . . . | 10 | : : : : : : : : : : : : : : : : : : : |
  9 | . . . . . . . . . . . . . X). . . . . |  9 | : : : : : : : : : : : : : : : : : : : |
  8 | . . . . . . . . . . . . . . . . . . . |  8 | : : : : : : : : : : : : : : : : : : : |
  7 | . . . . . . . . . . . . . . . . . . . |  7 | : : : : : : : : : : : : : : : : : : : |
  6 | . . . . . . . . . . . . . . . . . . . |  6 | : : : : : : : : : : : : : : : : : : : |
  5 | . . . . . . . . . . . . . . . . . . . |  5 | : : : : : : : : : : : : : : : : : : : |
  4 | . . . O . . . . . . . . . . . X . . . |  4 | : : : : : : : : : : : : : : : : : : : |
  3 | . . . . . . . . . . . . . . . . . . . |  3 | : : : : : : : : : : : : : : : : : : : |
  2 | . . . . . . . . . . . . . . . . . . . |  2 | : : : : : : : : : : : : : : : : : : : |
  1 | . . . . . . . . . . . . . . . . . . . |  1 | : : : : : : : : : : : : : : : : : : : |
    +---------------------------------------+    +---------------------------------------+

desired 107.10, worst 119.00, clock [1] 59.00 + 60.00/1*1, lag 17.85
prior_dcnn(pass) = [ D16 D17 Q16 R16 O3 ]      [ 0.29 0.16 0.12 0.07 0.06 ]
[10000] best 0.526565 xkomi 1.0 | seq D16  R6  Q7  R7 | can w D16(0.527) D17(0.516) Q16(0.533)  O3(0.541)
[20000] best 0.523203 xkomi 1.0 | seq D16  R6  Q7  R7 | can w D16(0.523) D17(0.520) Q16(0.520)  O3(0.529)
[30000] best 0.518452 xkomi 1.0 | seq D16 C16 D15 C15 | can w D16(0.518) D17(0.520)  O3(0.523) Q16(0.515)
[40000] best 0.524808 xkomi 1.0 | seq D16 C16 C17 P16 | can w D16(0.525) D17(0.518) Q16(0.519)  O3(0.521)
[50000] best 0.523357 xkomi 1.0 | seq D16 C16 C17 P16 | can w D16(0.523) D17(0.519)  O3(0.524) Q16(0.516)
[60000] best 0.523286 xkomi 1.0 | seq D16 C16 C17 P16 | can w D16(0.523) D17(0.518) Q16(0.519) R16(0.524)
[70000] best 0.522423 xkomi 1.0 | seq D16 B12  B8  B9 | can w D16(0.522) D17(0.519) Q16(0.517) R16(0.522)
Best2 ratio 1.312776 < threshold 2.500000
Best2 ratio 1.302733 < threshold 2.500000
Best2 ratio 1.298231 < threshold 2.500000
Best2 ratio 1.298135 < threshold 2.500000
Best2 ratio 1.298135 < threshold 2.500000
Best2 ratio 1.298135 < threshold 2.500000
Best2 ratio 1.298135 < threshold 2.500000
Best2 ratio 1.298283 < threshold 2.500000
Best2 ratio 1.295998 < threshold 2.500000
Best2 ratio 1.294183 < threshold 2.500000
Best2 ratio 1.293303 < threshold 2.500000
Best2 ratio 1.293355 < threshold 2.500000
Best2 ratio 1.293355 < threshold 2.500000
Best2 ratio 1.293355 < threshold 2.500000
Best2 ratio 1.293355 < threshold 2.500000
Best2 ratio 1.293355 < threshold 2.500000
Best2 ratio 1.293355 < threshold 2.500000
Best2 ratio 1.293355 < threshold 2.500000
Best2 ratio 1.293355 < threshold 2.500000
Best2 ratio 1.293355 < threshold 2.500000
Best2 ratio 1.293355 < threshold 2.500000
Best2 ratio 1.293355 < threshold 2.500000
Best2 ratio 1.293355 < threshold 2.500000
Best2 ratio 1.293355 < threshold 2.500000
Best2 ratio 1.293355 < threshold 2.500000
Best2 ratio 1.291642 < threshold 2.500000
Best2 ratio 1.291526 < threshold 2.500000
[80000] best 0.521671 xkomi 1.0 | seq D16 B12  C8  B9 | can w D16(0.522) D17(0.522) R16(0.522) Q16(0.516)
Best2 ratio 1.291526 < threshold 2.500000
Best2 ratio 1.290344 < threshold 2.500000
Best2 ratio 1.287411 < threshold 2.500000
Best2 ratio 1.284116 < threshold 2.500000
Best2 ratio 1.281960 < threshold 2.500000
Best2 ratio 1.279065 < threshold 2.500000
Best2 ratio 1.276619 < threshold 2.500000
Best2 ratio 1.272549 < threshold 2.500000
Best2 ratio 1.271444 < threshold 2.500000
Best2 ratio 1.271182 < threshold 2.500000
Best2 ratio 1.271162 < threshold 2.500000
Best2 ratio 1.272070 < threshold 2.500000
Best2 ratio 1.270725 < threshold 2.500000
Best2 ratio 1.271172 < threshold 2.500000
Best2 ratio 1.271605 < threshold 2.500000
Best2 ratio 1.275070 < threshold 2.500000
Best2 ratio 1.277453 < threshold 2.500000
Best2 ratio 1.279835 < threshold 2.500000
Best2 ratio 1.282434 < threshold 2.500000
Best2 ratio 1.283705 < threshold 2.500000
Best2 ratio 1.287291 < threshold 2.500000
Best2 ratio 1.286023 < threshold 2.500000
Best2 ratio 1.283712 < threshold 2.500000
Best2 ratio 1.285324 < threshold 2.500000
Best2 ratio 1.285519 < threshold 2.500000
Best2 ratio 1.279359 < threshold 2.500000
Best2 ratio 1.279431 < threshold 2.500000
Best2 ratio 1.279339 < threshold 2.500000
Best2 ratio 1.279339 < threshold 2.500000
Best2 ratio 1.279339 < threshold 2.500000
Best2 ratio 1.279339 < threshold 2.500000
Best2 ratio 1.279339 < threshold 2.500000
Best2 ratio 1.279339 < threshold 2.500000
Best2 ratio 1.279339 < threshold 2.500000
Best2 ratio 1.279339 < threshold 2.500000
Best2 ratio 1.279339 < threshold 2.500000
Best2 ratio 1.279339 < threshold 2.500000
Best2 ratio 1.279339 < threshold 2.500000
Best2 ratio 1.279339 < threshold 2.500000
Best2 ratio 1.279339 < threshold 2.500000
Best2 ratio 1.279339 < threshold 2.500000
Best2 ratio 1.279339 < threshold 2.500000
Best2 ratio 1.279339 < threshold 2.500000
Best2 ratio 1.279339 < threshold 2.500000
Best2 ratio 1.279339 < threshold 2.500000
Best2 ratio 1.279339 < threshold 2.500000
Best2 ratio 1.280913 < threshold 2.500000
Best2 ratio 1.281127 < threshold 2.500000
Best2 ratio 1.281127 < threshold 2.500000
Best2 ratio 1.281270 < threshold 2.500000
Best2 ratio 1.283499 < threshold 2.500000
Best2 ratio 1.289430 < threshold 2.500000
Best2 ratio 1.292146 < threshold 2.500000
Best2 ratio 1.292247 < threshold 2.500000
Best2 ratio 1.292247 < threshold 2.500000
Best2 ratio 1.292247 < threshold 2.500000
Early stop, result cannot change: best 18085, best2 13995, estimated 4025.896918 simulations to go (84196/97.640910=862.302491 pps)
(UCT tree; root black; extra komi 0.970000; max depth 9)
[pass] 0.520/84201 [prior 1.000/0 amaf 1.000/0 crit -0.520 vloss 0] h=0 c#=359 <2630df4600000006>
[D16] 0.521/18085 [prior 0.987/749 amaf 0.520/126758 crit 0.033 vloss 0] h=0 c#=358 <2630e06700000127>
  [B12] 0.513/5816 [prior 0.500/20 amaf 0.507/30711 crit 0.043 vloss 0] h=0 c#=357 <2630e18000000240>
   [C8] 0.531/1038 [prior 0.863/73 amaf 0.522/8494 crit 0.020 vloss 0] h=0 c#=356 <2633a5050002c5c5>
   [B8] 0.526/993 [prior 0.500/20 amaf 0.529/8598 crit 0.035 vloss 0] h=0 c#=356 <2633a5040002c5c4>
  [C16] 0.513/3470 [prior 0.123/81 amaf 0.514/23087 crit -0.004 vloss 0] h=0 c#=357 <2630e1cd0000028d>
   [C17] 0.529/1162 [prior 0.947/190 amaf 0.527/6318 crit 0.006 vloss 0] h=0 c#=356 <2633c2df0002e39f>
  [R6] 0.511/1558 [prior 0.244/41 amaf 0.513/22670 crit 0.022 vloss 0] h=0 c#=357 <2630e11e000001de>
  [C8] 0.517/1402 [prior 0.137/73 amaf 0.511/21707 crit 0.030 vloss 0] h=0 c#=357 <2630e136000001f6>
  [M3] 0.517/924 [prior 0.137/73 amaf 0.512/21537 crit 0.013 vloss 0] h=0 c#=357 <2630e0e2000001a2>
[D17] 0.522/13995 [prior 0.983/591 amaf 0.522/122794 crit 0.024 vloss 0] h=0 c#=358 <2630e07a0000013a>
  [F17] 0.518/3949 [prior 0.111/90 amaf 0.521/23448 crit 0.025 vloss 0] h=0 c#=357 <2630e34800000408>
   [F3] 0.552/1124 [prior 0.756/41 amaf 0.549/6308 crit 0.012 vloss 0] h=0 c#=356 <26320d6800012e28>
  [B9] 0.499/1834 [prior 0.500/20 amaf 0.504/18603 crit 0.046 vloss 0] h=0 c#=357 <2630e2ae0000036e>
  [M18] 0.507/1223 [prior 0.500/20 amaf 0.506/17529 crit 0.029 vloss 0] h=0 c#=357 <2630e36100000421>
[Q16] 0.516/9297 [prior 0.980/506 amaf 0.519/111625 crit 0.035 vloss 0] h=0 c#=358 <2630e07300000133>
  [D5] 0.501/1410 [prior 0.500/20 amaf 0.503/10934 crit 0.026 vloss 0] h=0 c#=357 <2630e3ca0000048a>
   [E5] 0.535/865 [prior 0.938/162 amaf 0.522/3057 crit 0.037 vloss 0] h=0 c#=356 <26359cc10004bd81>
  [R10] 0.494/1287 [prior 0.217/46 amaf 0.502/12552 crit 0.012 vloss 0] h=0 c#=357 <2630e435000004f5>
  [D16] 0.506/1182 [prior 0.204/49 amaf 0.505/12047 crit 0.034 vloss 0] h=0 c#=357 <2630e49a0000055a>
  [C6] 0.501/960 [prior 0.172/58 amaf 0.509/12058 crit 0.056 vloss 0] h=0 c#=357 <2630e3dc0000049c>
[R16] 0.521/9169 [prior 0.976/409 amaf 0.522/111753 crit 0.027 vloss 0] h=0 c#=358 <2630e07400000134>
  [H3] 0.504/1874 [prior 0.137/73 amaf 0.512/13481 crit 0.048 vloss 0] h=0 c#=357 <2630ed6e00000e2e>
  [C8] 0.511/1261 [prior 0.137/73 amaf 0.511/12230 crit 0.029 vloss 0] h=0 c#=357 <2630edc600000e86>
  [N17] 0.513/954 [prior 0.137/73 amaf 0.509/11812 crit 0.023 vloss 0] h=0 c#=357 <2630ee7900000f39>
[O3] 0.522/8945 [prior 0.973/367 amaf 0.521/109433 crit 0.033 vloss 0] h=0 c#=358 <2630df7d0000003d>
  [N17] 0.503/1719 [prior 0.137/73 amaf 0.509/13083 crit 0.025 vloss 0] h=0 c#=357 <2632156a0001362a>
  [F3] 0.510/1401 [prior 0.172/58 amaf 0.510/12395 crit 0.044 vloss 0] h=0 c#=357 <2632145d0001351d>
  [C11] 0.518/1089 [prior 0.137/73 amaf 0.508/11805 crit 0.025 vloss 0] h=0 c#=357 <263214ee000135ae>
[R15] 0.524/7473 [prior 0.968/312 amaf 0.521/108077 crit 0.027 vloss 0] h=0 c#=358 <2630e06100000121>
  [J3] 0.512/4071 [prior 0.137/73 amaf 0.514/15837 crit 0.029 vloss 0] h=0 c#=357 <2630ec0900000cc9>
   [K17] 0.532/988 [prior 0.863/73 amaf 0.528/6455 crit 0.012 vloss 0] h=0 c#=356 <26318e890000af49>
[K4] 0.522/6148 [prior 0.961/256 amaf 0.522/106982 crit 0.033 vloss 0] h=0 c#=358 <2630df8b0000004b>
  [N17] 0.507/1886 [prior 0.137/73 amaf 0.513/10252 crit 0.024 vloss 0] h=0 c#=357 <263630a400055164>
  [H17] 0.514/845 [prior 0.137/73 amaf 0.507/8326 crit 0.003 vloss 0] h=0 c#=357 <2636309f0005515f>
[C16] 0.521/5133 [prior 0.960/250 amaf 0.522/99416 crit 0.029 vloss 0] h=0 c#=358 <2630e06600000126>
  [J17] 0.510/1072 [prior 0.137/73 amaf 0.501/7491 crit 0.022 vloss 0] h=0 c#=357 <2635e724000507e4>
  [D16] 0.508/965 [prior 0.133/75 amaf 0.509/6359 crit 0.030 vloss 0] h=0 c#=357 <2635e70c000507cc>
  [D9] 0.491/959 [prior 0.204/49 amaf 0.499/7502 crit 0.043 vloss 0] h=0 c#=357 <2635e68900050749>
[Q17] 0.508/1654 [prior 0.964/277 amaf 0.522/95423 crit 0.023 vloss 0] h=0 c#=358 <2630e08600000146>
[Q15] 0.513/1437 [prior 0.963/268 amaf 0.519/98606 crit 0.036 vloss 0] h=0 c#=358 <2630e06000000120>


garde wrote:
Is there also anyway to create a book of games that the engine can use? I want to make PachiUCT aggressive, if possible.


You can use Fuego-books by adding the commandline: "-f pachibook.dat"

I don't know how to create a fuego book but you will probably find the answer in the fuego documentation.

garde wrote:
The README file that comes with Pachi UCT itself says that I can make it faster with the command /pachi -t _1200 threads=8,max_tree_size=3072 , and by changing it. Could I possibly go higher? Or will that just make it slower?


-t is just the time setting in seconds. It will be overwritten by what you enter in GoGui. The more time pachi has, the better will it play. Threads can be expanded to the number of cores you like to give pachi, this will make it stronger. The size is the RAM you want to give pachi, this will make it slightly stronger, but don't give pachi more than what is available.

If you need a stronger oponent and better analysis tool try Leela, it is easy to setup and has an own UI.

garde wrote:
(Maybe I'm asking WAY to much for one post.)


Keep on asking if you like :)

Best,
Alex
Page 1 of 1 All times are UTC - 8 hours [ DST ]
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group
http://www.phpbb.com/