It is currently Fri May 03, 2024 11:30 am

View unanswered posts | View active topics View unread posts | View new posts | View your posts


Board index » L19²GO (Malkovich) » 1-on-1 Malkovich games

All times are UTC - 8 hours [ DST ]



Post new topic Reply to topic  Page 6 of 8
  [ 152 posts ]  Go to page Previous  1 ... 3, 4, 5, 6, 7, 8  Next
  Print view Previous topic | Next topic 
Author Message
Offline perceval
 Post subject: Re: 124. Chew (3k) vs Fuego (Bot)
Post #101 Posted: Mon Jul 04, 2011 3:17 am 
Lives in gote
User avatar

Posts: 312
Liked others: 52
Was liked: 41
Rank: 7K KGS
KGS: tictac
Here is what i understand of Fuego process, which might be really wrong:
-it uses the UTC algo ie generate random playouts, with a preferences to explore moves that seems to be better.
ez4u asked why in the number of playout the min seemed to be 23.
my guess is that a minimum value to get a rough estimate of the winning % of each node.

However 23 playouts gives a huge error bar on winning ratio (10%).

Increasing to, say 2500 would take little time (300*2500*=750000 playout, much less than the total number played) and decrease the error bar to 1%

i would like to know if changing this value would change the exploration rates and the final move ?
it should not of course but i wonder how the search tree would be modified.

Esmeralddemon, if you know how to change this value, would you have time to run the 2 search with 23 and 2500 as min playout for next move to see if there is a difference ?

it might be a part of the horizon "effect" ie the fact that increasing the thinking time does not linearly increase the playing strength because a good move might be discarded early because of lack of luck in early explorations. (or more complicated effects such as: its a good move but only following a very narrow tactical path that a few random playouts are unlikely to find).
Woudl you find this experiment interesting ?

_________________
In theory, there is no difference between theory and practice. In practice, there is.

This post by perceval was liked by: ez4u
Top
 Profile  
 
Offline flOvermind
 Post subject: Re: 124. Chew (3k) vs Fuego (Bot)
Post #102 Posted: Mon Jul 04, 2011 8:05 am 
Lives with ko
User avatar

Posts: 295
Location: Linz, Austria
Liked others: 21
Was liked: 44
Rank: EGF 4 kyu
GD Posts: 627
The problem with increasing the minimum playouts of each move is that this involves an exponential decrease in the statistical coverage of the followup moves.

Making 1500 playouts once is not that much, but 1500 playouts for each move at each expanded node is quite a lot. With normal tree traversal that would translate to an exponential increase in time, but the nature of the UCT algorithm's playouts means effectively that the total amount of playouts choosing "good" nodes decreases, while the time remains constant. That means the child nodes will get expanded later (on average), and the maximum tree depth decreases.

Effectively, you're making the search tree more balanced. That makes the horizon effect worse instead of better ;)


Of course it would be interesting to find a good balance between discarding bad nodes early, and increasing the accuracy when deciding what nodes are bad. But 1500 feels a lot too high.

This post by flOvermind was liked by: ez4u
Top
 Profile  
 
Offline emeraldemon
 Post subject: Re: 124. Chew (3k) vs Fuego (Bot)
Post #103 Posted: Mon Jul 04, 2011 10:01 am 
Gosei
User avatar

Posts: 1744
Liked others: 704
Was liked: 288
KGS: greendemon
Tygem: greendemon
DGS: smaragdaemon
OGS: emeraldemon
perceval wrote:

Increasing to, say 2500 would take little time (300*2500*=750000 playout, much less than the total number played) and decrease the error bar to 1%



The trouble is that the UCT algorithm doesn't treat the root node differently from any other node in the search tree. So it wouldn't be +750000 playouts, it would be +750000 playouts per move (which is why fl0vermind said the increase is exponential). The response from Martin Mueller suggested the possibility of changing the algorithm to treat the root node differently, but it would be a nontrivial change to the source code.

That said, there is a specific exploration/exploitation tradeoff parameter I can tweak and play with. I will try some different values for the next move, see what it does.
However, as topazg pointed out. It's easy to see if the program plays differently, but not so easy to tell if it plays better.

This post by emeraldemon was liked by: ez4u
Top
 Profile  
 
Offline perceval
 Post subject: Re: 124. Chew (3k) vs Fuego (Bot)
Post #104 Posted: Mon Jul 04, 2011 1:37 pm 
Lives in gote
User avatar

Posts: 312
Liked others: 52
Was liked: 41
Rank: 7K KGS
KGS: tictac
o ok i though there was something special about the first node because of those 23 playouts for all nodes.

_________________
In theory, there is no difference between theory and practice. In practice, there is.
Top
 Profile  
 
Offline Mike Novack
 Post subject: Re: 124. Chew (3k) vs Fuego (Bot)
Post #105 Posted: Mon Jul 04, 2011 5:29 pm 
Lives in sente

Posts: 1037
Liked others: 0
Was liked: 181
perceval wrote:
ez4u asked why in the number of playout the min seemed to be 23.
my guess is that a minimum value to get a rough estimate of the winning % of each node.

However 23 playouts gives a huge error bar on winning ratio (10%).


Ah, but does an error bar of that size hurt anything? At this point all that is being done is pruning the worst nodes from the tree. Not yet selecting the best. All that is needed is that the right node (the move ultimately decided upon) not get pruned early.

For example, if that first pruning process didn't remove anything with a score over half of the highest score then that possible 10% error wouldn't much.
Top
 Profile  
 
Offline Chew Terr
 Post subject: Re: 124. Chew (3k) vs Fuego (Bot)
Post #106 Posted: Tue Jul 05, 2011 7:39 am 
Gosei
User avatar

Posts: 2060
Location: Texas
Liked others: 546
Was liked: 173
Rank: KGS 3k
GD Posts: 264
KGS: Chew
Click Here To Show Diagram Code
[go]$$Bcm41
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . O . . . . . . . . . |
$$ | . . . O . O . . . , . . . . . O . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . X . . . . . . . . . . . . . O . . |
$$ | . . . . O . . . . . 6 . . . . X O . . |
$$ | . . X . . . . . . . . . O O X O X . . |
$$ | . . . . . . . . . . . X X O O O X . . |
$$ | . . . X . . . . . , . . . X X X . . . |
$$ | . . . . . . . . . . . . . . . . X . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . O . . . |
$$ | . . . . X X . X . . . . . . X O . . . |
$$ | . . X , X O . . . O . . X . . X O . . |
$$ | . . . . . O . O . . . . X O X . O . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]



Click Here To Show Diagram Code
[go]$$Bcm41
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . O . . . . . . . . . |
$$ | . . . O . O . . . , . . . . . O . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . X . . . . . . . . . . . . . O . . |
$$ | . . . . O . . . . . a . . . . X O . . |
$$ | . . X . . . . . . . . . O O X O X . . |
$$ | . . . . . . . . . . . 5 X O O O X . . |
$$ | . . . X . f . . . , . . . X X X . . . |
$$ | . . . . . . . . . . . . . . . . X . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . e . . . . . O . . . |
$$ | . . . . X X . X . . . . . . X O . . . |
$$ | . . X , X O . . . O . . X c . X O . . |
$$ | . . b . . O . O . . . . X O X . O . . |
$$ | . . . d . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]

I wanted all of these points. 'b' is iffy because it hurts my bottom group a bit, but the rest look pretty good. 'c' is slightly less big, now that there's at least some black stones above for these to run to if I steal their base. Essentially though, it seemed TOO good for black to let it turn the curner here in the middle, so I jumped ahead to prevent it. I'll have to think about the biggest next move again in a bit, but I want this now. I like the looks of 'f' actually, not for any threatened followup, but for the safe reduction.

_________________
Someday I want to be strong enough to earn KGS[-].
Top
 Profile  
 
Offline emeraldemon
 Post subject: Re: 124. Chew (3k) vs Fuego (Bot)
Post #107 Posted: Wed Jul 06, 2011 7:47 pm 
Gosei
User avatar

Posts: 1744
Liked others: 704
Was liked: 288
KGS: greendemon
Tygem: greendemon
DGS: smaragdaemon
OGS: emeraldemon
Click Here To Show Diagram Code
[go]$$Bcm41
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . O . . . . . . . . . |
$$ | . . . O . O . . . , . . . . . O . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . X . . . . . . . . . . . . . O . . |
$$ | . . . . O . . . . . O . . . . X O . . |
$$ | . . X . . . . . . . . . O O X O X . . |
$$ | . . . . . . . . . . . X X O O O X . . |
$$ | . . . X . . . . . , . . . X X X . . . |
$$ | . . . . . . . . . . . . . . . . X . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . O . . . |
$$ | . . . . X X . X . . . . . . X O . . . |
$$ | . . X , X O . . . O . . X . . X O . . |
$$ | . . . . . O . O . . . . X O X . O . . |
$$ | . . . . . . . . . 7 . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]


Took 17:03 to decide
value: 51.5%

Click Here To Show Diagram Code
[go]$$Bcm47
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . O . . . . . . . . . |
$$ | . . . O . O . . . , . . . . . O . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . X . . . . . . . . . . . . 2 O . . |
$$ | . . . . O . . . . . O . . . . X O . . |
$$ | . . X . . . . . . 3 . . O O X O X . . |
$$ | . . . . . . . . . . . X X O O O X . . |
$$ | . . . X . . . . . , . . . X X X . . . |
$$ | . . . . . . . . . . . . . . . . X . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . O . . . |
$$ | . . . . X X . X . . . . . . X O . . . |
$$ | . . X , X O . . . O . . X 5 . X O . . |
$$ | . . . . . O . O . 6 9 . X O X . O . . |
$$ | . . . . . . . . 7 1 8 0 . 4 . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]


Still seems to greatly overestimate the aji of its 2 stones in the upper right. For the people who were curious, I ran again w/ a significantly higher search count:
Attachment:
revised.png
revised.png [ 629.88 KiB | Viewed 8604 times ]


It came up with this sequence:

Click Here To Show Diagram Code
[go]$$Bcm47
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . O . . . . . . . . . |
$$ | . . . O . O . . . , . . . . . O . . . |
$$ | . . . . . . . . . . . . . . 8 . . . . |
$$ | . . X . . . . . . . . . . 9 7 4 O . . |
$$ | . . . . O . . . . . O . . 6 5 X O . . |
$$ | . . X . . . . . . . . . O O X O X . . |
$$ | . . . . . . . . . . . X X O O O X . . |
$$ | . . . X . . . . . , . . . X X X . . . |
$$ | . . . . . . . . . . . . . . . . X . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . O . 1 . |
$$ | . . . . X X . X . . . . . . X O . . . |
$$ | . . X , X O . . . O . . X 3 . X O . . |
$$ | . . . . . O . O . . . . X O X . O . . |
$$ | . . . . . . . . . . . . . 2 . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]


seems much worse...
Top
 Profile  
 
Offline Chew Terr
 Post subject: Re: 124. Chew (3k) vs Fuego (Bot)
Post #108 Posted: Thu Jul 07, 2011 8:11 am 
Gosei
User avatar

Posts: 2060
Location: Texas
Liked others: 546
Was liked: 173
Rank: KGS 3k
GD Posts: 264
KGS: Chew
Click Here To Show Diagram Code
[go]$$Bcm41
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . O . . . . . . . . . |
$$ | . . . O . O . . . , . . . . . O . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . X . . . . . . . . . . . . . O . . |
$$ | . . . . O . . . . . O . . . . X O . . |
$$ | . . X . . . . . . . . . O O X O X . . |
$$ | . . . . . . . . . . . X X O O O X . . |
$$ | . . . X . . . . . , . . . X X X . . . |
$$ | . . . . . . . . . . . . . . . . X . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . O . . . |
$$ | . . . . X X . X . . . . . . X O . . . |
$$ | . . X , X O . . . O . . X . . X O . . |
$$ | . . . . . O . O . 8 . . X O X . O . . |
$$ | . . . . . . . . 0 X 9 . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]



Yeah, I've been leaving this group hanging for a bit too long. Really wanted to slide into the corner more for eyespace instead, but didn't like how much white could threaten to sever connections, while undercutting my base.

_________________
Someday I want to be strong enough to earn KGS[-].
Top
 Profile  
 
Offline emeraldemon
 Post subject: Re: 124. Chew (3k) vs Fuego (Bot)
Post #109 Posted: Thu Jul 07, 2011 11:46 am 
Gosei
User avatar

Posts: 1744
Liked others: 704
Was liked: 288
KGS: greendemon
Tygem: greendemon
DGS: smaragdaemon
OGS: emeraldemon
Fuego doesn't like your suggestions...

Click Here To Show Diagram Code
[go]$$Bcm41
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . O . . . . . . . . . |
$$ | . . . O . O . . . , . . . . . O . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . X . . . . . . . . . . . . . O . . |
$$ | . . . . O . . . . . O . . . . X O . . |
$$ | . . X . . . . . . . . . O O X O X . . |
$$ | . . . . . . . . . . . X X O O O X . . |
$$ | . . . X . . . . . , . . . X X X . . . |
$$ | . . . . . . . . . . . . . . . . X . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . O . . . |
$$ | . . . . X X . X . . . . . . X O . . . |
$$ | . . X , X O . . . O . . X . . X O . . |
$$ | . . . . . O . O . O . . X O X . O . . |
$$ | . . . . . . . . 9 X . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]


Value: 51.0%

Click Here To Show Diagram Code
[go]$$Bcm49
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . O . . . . . . . . . |
$$ | . . . O . O . . . , . . . . . O . . . |
$$ | . . . . . . . . . . . . . . . 8 . . . |
$$ | . . X . . . . . . . . . . . . . O . . |
$$ | . . . . O . . . . . O . . . . X O . . |
$$ | . . X . . . . . . . . . O O X O X . . |
$$ | . . . . . . . . . . . X X O O O X . . |
$$ | . . . X . . . . . , . . . X X X . . . |
$$ | . . . . . . . . . . . . . . . . X . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . 6 . . . . . . . . . |
$$ | . . . . . . . . . 9 0 . . . 7 O . . . |
$$ | . . . . X X . X . . . . . . X O . . . |
$$ | . . X , X O . . . O . . X . . X O . . |
$$ | . . . . . O . O . O 4 . X O X . O . . |
$$ | . . . . . . . . 1 X 2 3 . . . . . . . |
$$ | . . . . . . . . . . 5 . . . . . . . . |
$$ ---------------------------------------[/go]


I like the undercutting that fuego is trying, but with the corner open I don't think black can really do much here. All white has to do is slide (which is pretty big). Overall I still prefer white here, but the game's not over yet (I can't help but root for Fuego).
Top
 Profile  
 
Offline Chew Terr
 Post subject: Re: 124. Chew (3k) vs Fuego (Bot)
Post #110 Posted: Thu Jul 07, 2011 12:18 pm 
Gosei
User avatar

Posts: 2060
Location: Texas
Liked others: 546
Was liked: 173
Rank: KGS 3k
GD Posts: 264
KGS: Chew
Click Here To Show Diagram Code
[go]$$Bcm41
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . O . . . . . . . . . |
$$ | . . . O . O . . . , . . . . . O . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . X . . . . . . . . . . . . . O . . |
$$ | . . . . O . . . . . O . . . . X O . . |
$$ | . . X . . . . . . . . . O O X O X . . |
$$ | . . . . . . . . . . . X X O O O X . . |
$$ | . . . X . . . . . , . . . X X X . . . |
$$ | . . . . . . . . . . . . . . . . X . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . O . . . |
$$ | . . . . X X . X . . . . . . X O . . . |
$$ | . . X , X O . . . O . . X . . X O . . |
$$ | . . . . . O . O . O . . X O X . O . . |
$$ | . . . . . . . 0 X X . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]


The cut doesn't work currently, though I'll probably have to fix it in gote soon... This is a lot worse locally than expected. =/

_________________
Someday I want to be strong enough to earn KGS[-].
Top
 Profile  
 
Offline Numsgil
 Post subject: Re: 124. Chew (3k) vs Fuego (Bot)
Post #111 Posted: Thu Jul 07, 2011 3:34 pm 
Lives in gote
User avatar

Posts: 614
Liked others: 28
Was liked: 65
Rank: 1 Kyu KGS
KGS: Numsgil
emeraldemon wrote:
(I can't help but root for Fuego).

I can't help think of John Henry the Steel Driving Man. I'll root for Chew, but it'll be quite sad when, though victorious, the exertion kills him.

_________________
1k KGS

This post by Numsgil was liked by 2 people: emeraldemon, ethanb
Top
 Profile  
 
Offline emeraldemon
 Post subject: Re: 124. Chew (3k) vs Fuego (Bot)
Post #112 Posted: Thu Jul 07, 2011 6:51 pm 
Gosei
User avatar

Posts: 1744
Liked others: 704
Was liked: 288
KGS: greendemon
Tygem: greendemon
DGS: smaragdaemon
OGS: emeraldemon
Click Here To Show Diagram Code
[go]$$Bcm51
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . O . . . . . . . . . |
$$ | . . . O . O . . . , . . . . . O . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . X . . . . . . . . . . . . . O . . |
$$ | . . . . O . . . . . O . . . . X O . . |
$$ | . . X . . . . . . . . . O O X O X . . |
$$ | . . . . . . . . . . . X X O O O X . . |
$$ | . . . X . . . . . , . . . X X X . . . |
$$ | . . . . . . . . . . . . . . . . X . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . O . . . |
$$ | . . . . X X . X . . . . . . X O . . . |
$$ | . . X , X O . . . O . . X . . X O . . |
$$ | . . . . . O . O . O . . X O X . O . . |
$$ | . . . . . . . O X X 1 . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]


Value: 51.9%
This is the most confident Fuego has ever been about the game, I think.

Click Here To Show Diagram Code
[go]$$Bcm51
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . O . . . . . . . . . |
$$ | . . . O . O . . . , . . . . . O . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . X . . . . . . . . . . . . . O . . |
$$ | . . . . O . . . . . O . . . . X O . . |
$$ | . . X . . . . . . . . . O O X O X . . |
$$ | . . . . . . . . . . . X X O O O X . . |
$$ | . . . X . . . . . , . . . X X X . . . |
$$ | . . . . . . . . . . . . . . . . X . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . O . . . |
$$ | . . . . X X . X . . . . . . X O . . . |
$$ | . . X , X O . . . O . . X 3 7 X O . . |
$$ | . . . . . O . O 9 O . . X O X 6 O . . |
$$ | . . 8 . 0 . . O X X 1 . 5 2 4 . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]


Is there anything in particular people are interested in besides the value & the main line?
Top
 Profile  
 
Offline Chew Terr
 Post subject: Re: 124. Chew (3k) vs Fuego (Bot)
Post #113 Posted: Fri Jul 08, 2011 9:46 am 
Gosei
User avatar

Posts: 2060
Location: Texas
Liked others: 546
Was liked: 173
Rank: KGS 3k
GD Posts: 264
KGS: Chew
Click Here To Show Diagram Code
[go]$$Bcm51
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . O . . . . . . . . . |
$$ | . . . O . O . . . , . . . . . O . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . X . . . . . . . . . . . . . O . . |
$$ | . . . . O . . . . . O . . . . X O . . |
$$ | . . X . . . . . . . . . O O X O X . . |
$$ | . . . . . . . . . . . X X O O O X . . |
$$ | . . . X . . . . . , . . . X X X . . . |
$$ | . . . . . . . . . . . . . . . . X . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . O . . . |
$$ | . . . . X X . X . . . . . . X O . . . |
$$ | . . X , X O . 2 . O . . X . . X O . . |
$$ | . . . . . O . O . O . . X O X . O . . |
$$ | . . . . . . . O X X X . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]

Don't want to get cut here. Next, I'll either slide towards the corner or go up and out.

_________________
Someday I want to be strong enough to earn KGS[-].
Top
 Profile  
 
Offline emeraldemon
 Post subject: Re: 124. Chew (3k) vs Fuego (Bot)
Post #114 Posted: Fri Jul 08, 2011 9:43 pm 
Gosei
User avatar

Posts: 1744
Liked others: 704
Was liked: 288
KGS: greendemon
Tygem: greendemon
DGS: smaragdaemon
OGS: emeraldemon
Click Here To Show Diagram Code
[go]$$Bcm51
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . O . . . . . . . . . |
$$ | . . . O . O . . . , . . . . . O . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . X . . . . . . . . . . . . . O . . |
$$ | . . . . O . . . . . O . . . . X O . . |
$$ | . . X . . . . . . . . . O O X O X . . |
$$ | . . . . . . . . . . . X X O O O X . . |
$$ | . . . X . . . . . , . . . X X X . . . |
$$ | . . . . . . . . . . . . . . . . X . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . O . . . |
$$ | . . . . X X . X 3 . . . . . X O . . . |
$$ | . . X , X O . O . O . . X . . X O . . |
$$ | . . . . . O . O . O . . X O X . O . . |
$$ | . . . . . . . O X X X . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]


value: 51.5%

Click Here To Show Diagram Code
[go]$$Bcm53
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . 5 . . . . . O . . . . . . . . . |
$$ | . . 3 O . O . . . , . . . . . O . . . |
$$ | . . . . . . . . . . . . . . . 6 . . . |
$$ | . . X . . . . . . . . . . . . . O . . |
$$ | . . . . O . . . . . O . . . . X O . . |
$$ | . . X . . . . . . . . . O O X O X . . |
$$ | . . . . . . . . . . . X X O O O X . . |
$$ | . . . X . . . . . , . . . X X X . . . |
$$ | . . . . . . . . . . . . . . . . X . . |
$$ | . . . . . . . . . . . . . . . . . 9 . |
$$ | . . . . . . . . . . . . . . . . 8 0 . |
$$ | . . . . . . . . . . . . . . . O . 7 . |
$$ | . . . . X X . X 1 . . . . . X O . . . |
$$ | . . X , X O . O . O . . X . . X O . . |
$$ | . . . . . O . O . O . . X O X . O . . |
$$ | . . 2 . 4 . . O X X X . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]


What's the deal with :w56: and :w58: ? I know I said Fuego evaluates both sides the same way, but somehow white seems to take all the pointless gote moves in this sequence...
Top
 Profile  
 
Offline Loons
 Post subject: Re: 124. Chew (3k) vs Fuego (Bot)
Post #115 Posted: Sat Jul 09, 2011 2:10 am 
Gosei
User avatar

Posts: 1378
Location: wHam!lton, Aotearoa
Liked others: 253
Was liked: 105
Wait, I'm confused about how Fuego thinks, again.

So the numbers on the board are the quantity of MC game-play-outs it started there, and every victorious play-out adds weight to that play-out's starting move, and high placing moves (2nd, 3rd place etc) are tried earlier in later play-outs (as they become established)?

Could you give us some samples of the finished MC play-outs Fuego reaches ? I am curious how crazy they look. (Sorry if I just overlooked them earlier). Would it be useful to post a white's-move-values board for the first move after Fuego's chosen move, or can we just infer that from Fuego's predicted sequences?
PS: The board doesn't look *so* terrible for black, though I guess I keep imagining in some kind of life on top. Where would other people play their next few moves as black, after Chew slides for life? (To contrast Fuego's tenuki sequences).
PPS: Sorry if I come across as confused, I am.

_________________
Revisiting Go - Study Journal
My Programming Blog - About the evolution of my go bot.
Top
 Profile  
 
Offline emeraldemon
 Post subject: Re: 124. Chew (3k) vs Fuego (Bot)
Post #116 Posted: Sat Jul 09, 2011 8:46 am 
Gosei
User avatar

Posts: 1744
Liked others: 704
Was liked: 288
KGS: greendemon
Tygem: greendemon
DGS: smaragdaemon
OGS: emeraldemon
I will try to explain in a bit more detail how Fuego works. There are two parts: the tree search and the playouts.

The tree starts completely empty (just the root node). If the same move is ever tried 3 times in a position in playouts, that node is added to the tree, so the tree of variations grows over time (that number 3 is a parameter than can be tweaked). Periodically Fuego goes through and prunes nodes in the tree with few visits, so the tree is constantly being reshaped over time.

Inside the tree, moves are chosen with basically a minimax policy: black chooses the move with the highest win %, then white chooses the response with the best win % for white, and so on down the tree. When the end of the tree is reached, the pseudo-random playout begins, and Fuego plays until a win or a loss. At that point, the win % for every move in that playout is adjusted.

The thing I think you're talking about is something I explained a bit of before, called RAVE. In the tree, instead of just looking at a win% for a move at a particular position, Fuego also considers the win% for other times that move was played, elsewhere in the tree. The idea is that good moves tend to be good moves even when played at different times, so Fuego can use that information to evaluate moves more quickly (RAVE stands for Rapid Action Value Estimation).

Of course, we all now order really does matter sometimes, so RAVE is actually on a kind of clock, where as the search goes on, less and less weight is given to the move's plays at other times, until eventually only the plays from that exact node are counted.

Now the interesting part: what playouts look like! This sgf has 4 sample playouts from the last move (they take a few moves to diverge):



Notice they play until everything is single-point eyes. Also remember it's area scoring, not territory scoring. As explained earlier in the thread, these aren't really "random", rather they're somewhat random with some simple heuristics (never fill your own eyes, capture stones, extend from atari, etc). You can judge from the sgf about how they look.


Attachments:
games2.sgf [11.18 KiB]
Downloaded 697 times
This post by emeraldemon was liked by 2 people: Loons, perceval
Top
 Profile  
 
Offline Chew Terr
 Post subject: Re: 124. Chew (3k) vs Fuego (Bot)
Post #117 Posted: Mon Jul 11, 2011 11:34 am 
Gosei
User avatar

Posts: 2060
Location: Texas
Liked others: 546
Was liked: 173
Rank: KGS 3k
GD Posts: 264
KGS: Chew
Click Here To Show Diagram Code
[go]$$Bcm51
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . O . . . . . . . . . |
$$ | . . . O . O . . . , . . . . . O . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . X . . . . . . . . . . . . . O . . |
$$ | . . . . O . . . . . O . . . . X O . . |
$$ | . . X . . . . . . . . . O O X O X . . |
$$ | . . . . . . . . . . . X X O O O X . . |
$$ | . . . X . . . . . , . . . X X X . . . |
$$ | . . . . . . . . . . . . . . . . X . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . O . . . |
$$ | . . . . X X . X X . . . . . X O . . . |
$$ | . . X , X O . O . O . . X . . X O . . |
$$ | . . . . . O . O . O . . X O X . O . . |
$$ | . . . 4 . . . O X X X . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]



As painful as this is, let's see about getting some eyespace.

_________________
Someday I want to be strong enough to earn KGS[-].
Top
 Profile  
 
Offline emeraldemon
 Post subject: Re: 124. Chew (3k) vs Fuego (Bot)
Post #118 Posted: Mon Jul 11, 2011 5:27 pm 
Gosei
User avatar

Posts: 1744
Liked others: 704
Was liked: 288
KGS: greendemon
Tygem: greendemon
DGS: smaragdaemon
OGS: emeraldemon
Click Here To Show Diagram Code
[go]$$Bcm51
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . O . . . . . . . . . |
$$ | . . . O . O . . . , . . . . . O . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . X . . . . . . . . . . . . . O . . |
$$ | . . . . O . . . . . O . . . . X O . . |
$$ | . . X . . . . . . . . . O O X O X . . |
$$ | . . . . . . . . . . . X X O O O X . . |
$$ | . . . X . . . . . , . . . X X X . . . |
$$ | . . . . . . . . . . . . . . . . X . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . O . . . |
$$ | . . . . X X . X X . . . . . X O . . . |
$$ | . . X , X O . O . O . . X . . X O . . |
$$ | . . . . . O . O . O . . X O X . O . . |
$$ | . . 5 O . . . O X X X . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]


Value: 51.2%

Click Here To Show Diagram Code
[go]$$Bcm55
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . O . . . . . . . . . |
$$ | . . . O . O . . . , . . . . . O . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . X . . . . . . . . . . . . . O . . |
$$ | . . . . O . . . . . O . . . . X O . . |
$$ | . . X . . . . . . . . . O O X O X . . |
$$ | . . . . . . . . . . . X X O O O X . . |
$$ | . . . X . . . . . , . . . X X X . . . |
$$ | . . . . . . . . . . . . . . . . X . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . 5 0 . . 9 8 . . . . |
$$ | . . . . . . . . . . 6 . . . 7 O . . . |
$$ | . . . . X X . X X 4 . . . . X O . . . |
$$ | . . X , X O . O . O . . X . . X O . . |
$$ | . . 3 2 . O . O . O . . X O X . O . . |
$$ | . . 1 O . . . O X X X . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]


This one took Fuego a bit longer than usual, there some other points it was considering:

Click Here To Show Diagram Code
[go]$$Bcm51
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . O . . . . . . . . . |
$$ | . . b O . O . . . , . . . . . O . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . X . d . . . . . . . . . . . O . . |
$$ | . . . . O . . . . . O . . . . X O . . |
$$ | . . X . . . . . . c . . O O X O X . . |
$$ | . . . . . . . . . . . X X O O O X . . |
$$ | . . . X . . . . . , . . . X X X . . . |
$$ | . . . . . . . . . . . . . . . . X . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . O . . . |
$$ | . . . . X X . X X . . . . . X O . . . |
$$ | . . X , X O . O . O . . X . . X O . . |
$$ | . . . . . O . O . O . . X O X . O . . |
$$ | . . a O . . . O X X X . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]


a: 66%
b: 11%
c: 8%
d: 4%

Considering many of the moves have been 85-95% on a particular move, there were a lot of choices in Fuego's mind for this one. I like the forcing moves here, if black can keep white weak on the bottom that middle wall may be worth something after all. That said, white may just have enough space to live outright, so maybe it doesn't matter.

I also agree that b in the diagram above is a big point, I think Chew feels that way also, it'll be really interesting to see who gets there first.
Top
 Profile  
 
Offline Chew Terr
 Post subject: Re: 124. Chew (3k) vs Fuego (Bot)
Post #119 Posted: Tue Jul 12, 2011 7:42 am 
Gosei
User avatar

Posts: 2060
Location: Texas
Liked others: 546
Was liked: 173
Rank: KGS 3k
GD Posts: 264
KGS: Chew
Click Here To Show Diagram Code
[go]$$Bcm51
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . O . . . . . . . . . |
$$ | . . . O . O . . . , . . . . . O . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . X . . . . . . . . . . . . . O . . |
$$ | . . . . O . . . . . O . . . . X O . . |
$$ | . . X . . . . . . . . . O O X O X . . |
$$ | . . . . . . . . . . . X X O O O X . . |
$$ | . . . X . . . . . , . . . X X X . . . |
$$ | . . . . . . . . . . . . . . . . X . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . O . . . |
$$ | . . . . X X . X X . . . . . X O . . . |
$$ | . . X , X O . O . O . . X . . X O . . |
$$ | . . . 6 . O . O . O . . X O X . O . . |
$$ | . . X O . . . O X X X . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]


Protects my connection in sente against a threat to live. I'll still have to play to escape or possibly try to make eyes, regardless, so this changes little... There's a REALLY intimidating amount of black strength where I've got to run, though... I need to decide if F1 will live.

_________________
Someday I want to be strong enough to earn KGS[-].
Top
 Profile  
 
Offline Loons
 Post subject: Re: 124. Chew (3k) vs Fuego (Bot)
Post #120 Posted: Tue Jul 12, 2011 9:07 pm 
Gosei
User avatar

Posts: 1378
Location: wHam!lton, Aotearoa
Liked others: 253
Was liked: 105
I'm not personally under the impression that Chew has the option of living locally after C3. Against eg. the placement F1, you could (spoiler)
reduce eyespace with G4 or D1. There's a good reason "reduce eyespace" is the first fundamental of L&D.

I think the shape resembles a popular tsumego.

I guess Fuego's predicted sequence shows it noticed this also... Though I find it funny that in its previous prediction, it also felt white needed to add a move to live but decided to make two moves in a row elsewhere (contrasted to chasing white, now). I guess the situations are quite different.

I am curious what the proper joseki will be for this corner when the time comes.
Click Here To Show Diagram Code
[go]$$Bc Springs to mind for :b1: (resembles a large knight's joseki).
$$ +---------------------------
$$ | . . . . . . . . . . . . .
$$ | . . 8 . . . . . . . . . .
$$ | . 3 2 . . . . . . O . . .
$$ | . 5 1 O . O . . . , . . .
$$ | . 7 4 6 . . . . . . . . .
$$ | . . X . . . . . . . . . .
$$ | . . . . O . . . . . O . .
$$ | . . X . . . . . . . . . O
$$ | . . . . . . . . . . . X X
$$ | . . . X . . . . . , . . .[/go]

I may have considered the 3-3 first.

_________________
Revisiting Go - Study Journal
My Programming Blog - About the evolution of my go bot.
Top
 Profile  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  Page 6 of 8
  [ 152 posts ]  Go to page Previous  1 ... 3, 4, 5, 6, 7, 8  Next

Board index » L19²GO (Malkovich) » 1-on-1 Malkovich games

All times are UTC - 8 hours [ DST ]

Who is online

Users browsing this forum: No registered users and 1 guest

You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group