124. Chew (3k) vs Fuego (Bot)

emeraldemon · **#141**

Click Here To Show Diagram Code: [go]$$Bcm71 $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . O . . . . . . . . . | $$ | . . . O . O . . . , . . . . . O . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . X . . . . . . . . . . . . . O . . | $$ | . . . . O . . . 7 . O . . . . X O . . | $$ | . . X . O X . . . . . . O O X O X . . | $$ | . . . X X O O . X . . X X O O O X . . | $$ | . . . X O . . . . , . . . X X X . . . | $$ | . . . . . . . . . . . . . . . . X . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . X X . . . . O . X . | $$ | . . . . X X . X X O X O . . X O . O . | $$ | . . X , X O . O . O O O X X . X O . . | $$ | . . X O . O . O . O . . X O X . O . . | $$ | . . X O . . . O X X X . . . . . . . . | $$ | . . . . . O . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Chew Terr · **#142**

Click Here To Show Diagram Code: [go]$$Bcm71 $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . O . . . . . . . . . | $$ | . . . O . O . . . , . . . . . O . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . X . . . . . . . . . . . . . O . . | $$ | . . . . O . 8 . 7 . O . . . . X O . . | $$ | . . X . O X . . . . . . O O X O X . . | $$ | . . . X X O O . X . . X X O O O X . . | $$ | . . . X O . . . . , . . . X X X . . . | $$ | . . . . . . . . . . . . . . . . X . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . X X . . . . O . X . | $$ | . . . . X X . X X O X O . . X O . O . | $$ | . . X , X O . O . O O O X X . X O . . | $$ | . . X O . O . O . O . . X O X . O . . | $$ | . . X O . . . O X X X . . . . . . . . | $$ | . . . . . O . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Seems reasonable that we're exchanging reductions, but I think my opponent's area was a bit more sealed, so perhaps the exchange is slightly favorable for me. Got to solidify a bit, if I'm to get anything at all, though.

emeraldemon · **#143**

Click Here To Show Diagram Code: [go]$$Bcm71 $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . O . . . . . . . . . | $$ | . . . O . O . . . , . . . . . O . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . X . . . . . . . . . . . . . O . . | $$ | . . . . O . O . X . O . . . . X O . . | $$ | . . X . O X . . . . . . O O X O X . . | $$ | . . . X X O O . X . . X X O O O X . . | $$ | . . . X O . . . . , . . . X X X . . . | $$ | . . . . . . . . . . . . . . . . X . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . X X . . . . O . X . | $$ | . . . . X X . X X O X O . . X O . O . | $$ | . . X , X O 9 O . O O O X X . X O . . | $$ | . . X O . O . O . O . . X O X . O . . | $$ | . . X O . . . O X X X . . . . . . . . | $$ | . . . . . O . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

It's story time, friends of L19. This story begins with me setting up the parameters for Fuego to think. Tree size: 17,000,000 nodes as always. 30 minutes. Fuego has the ability to show the sequence it likes best in real time, which I usually turn on, because it's interesting to watch the sequences shift around. I hit "play move". Its first instinct is the cover at M6. I go make myself a peanut butter & jelly sandwich.

When I come back, sandwich in hand, it's considering a sequence something like this:

Click Here To Show Diagram Code: [go]$$Bcm79 $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . O . . . . . . . . . | $$ | . . . O . O . . . , . . . . . O . . . | $$ | . . . . . . . . . . . . . . . . 4 . . | $$ | . . X . . . . . . . . . . . . . O . . | $$ | . . . . O . O . X . O . . . . X O . . | $$ | . . X . O X . . . . . . O O X O X . . | $$ | . . . X X O O . X . . X X O O O X . . | $$ | . . . X O . . . . , . . . X X X . . . | $$ | . . . . . . . . . . . . . . . . X . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . X X . . . . O . X . | $$ | . . . . X X . X X O X O . . X O . O . | $$ | . . X , X O . O . O O O X X . X O . . | $$ | . . X O 1 O . O . O . . X O X . O . . | $$ | . . X O 2 . 5 O X X X . . . . . . . . | $$ | . . . 3 . O . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

It's quite obvious that white (still!) won't connect at the top. It's just as obvious that if :w82:

is at

, the group lives. Frowning, I set up the problem on my board. As I read it, x, y, and z are miai:

Click Here To Show Diagram Code: [go]$$Bcm71 $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . X X . . . . O . X . | $$ | . . . . X X . X X O X O . . X O . O . | $$ | . . X , X O x O . O O O X X . X O . . | $$ | . . X O . O . O . O . . X O X . O . . | $$ | . . X O . . . O X X X . . . . . . . . | $$ | . . . y . O . z . . . . . . . . . . . | $$ ---------------------------------------[/go]

Black can take any one of them, and the other two become miai for the kill. Note that E3, as Fuego is thinking, doesn't actually remove an eye, and in fact encourages Chew to live.

I finish my sandwich. Fuego still likes E3. The win percentage is dropping, down to 59.2%. My frown deepens. In my head I'm composing a mournful post on Fuego's inability to read. Then, at 23:45, Fuego puts down this sequence:

Click Here To Show Diagram Code: [go]$$Bcm79 $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . O . . . . . . . . . | $$ | . . . O . O . . . , . . . . . O . . . | $$ | . . . . . . . . . . . . . . . . 2 . . | $$ | . . X . . . . . . . . . . . . . O . . | $$ | . . . . O . O . X . O . . . . X O . . | $$ | . . X . O X . . . . . . O O X O X . . | $$ | . . . X X O O . X . . X X O O O X . . | $$ | . . . X O . . . . , . . . X X X . . . | $$ | . . . 5 4 . . . . . . . . . . . X . . | $$ | . . . . . . . . . . . 9 . . . . . . . | $$ | . . . . . . . . . . 0 7 8 . . . . . . | $$ | . . . . . . . . . X X 6 . . 3 O . X . | $$ | . . . . X X . X X O X O . . X O . O . | $$ | . . X , X O 1 O . O O O X X . X O . . | $$ | . . X O . O . O . O . . X O X . O . . | $$ | . . X O . . . O X X X . . . . . . . . | $$ | . . . . . O . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

I won't lie, patient readers. I want Fuego to win. I've been setting it up and watching it think for over a month now, playing its moves, being Hikaru to its Sai. When I saw this, I jumped out of my chair. "Yes! Stop thinking! You've found it!" Fuego kept me in suspense for 4 more minutes (26:40) before committing to this move.

Value: 59.3%
Thinking time: 38% G4, 26% E3, 10% M6, 6% P6.

ez4u · **#144**

emeraldemon wrote:

lightvector · **#145**

emeraldemon and observers:

Chew Terr · **#146**

Click Here To Show Diagram Code: [go]$$Wcm80 $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . O . . . . . . . . . | $$ | . . . O . O . . . , . . . . . O . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . X . . . . . . . . . . . . . O . . | $$ | . . . . O . O . X . O . . . . X O . . | $$ | . . X . O X . . . . . . O O X O X . . | $$ | . . . X X O O . X . . X X O O O X . . | $$ | . . . X O . . . . , . . . X X X . . . | $$ | . . . . . . . . . . . . . . . . X . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . X X . . . . O . X . | $$ | . . . . X X . X X O X O . . X O . O . | $$ | . . X , X O X O . O O O X X . X O . . | $$ | . . X O 2 O 1 O . O . . X O X . O . . | $$ | . . X O 3 . 5 O X X X . . . . . . . . | $$ | . . . 4 . O . 6 . . . . . . . . . . . | $$ ---------------------------------------[/go]

I resign. Thank you, Emerald Demon, for running the bot. I'm interested in looking through and seeing how the bot 'thought'. I was feeling a bit leery of this group, but I kept feeling that if I took gote to defend it, I'd just be too far behind, so I had to hope I was okay and keep an eye out for ways to make a group for M5 to run to, just in case. Earlier, I had convinced myself it was alive, but either that changed with my bad attempts to stick my head out in sente, or it was misreading in the first place.

I'm fairly impressed with Fuego, I imagine that I could actually get some decent practice against it, if I chose to. I probably won't bother (KGS automatch is SO easy, and provides better variety), but it's an intersting option. If I would be in a situation for a long period where internet was unavailable, I think I'd probably give it a go.

And Emdem: we should play a game sometime. Seeing your name in the "Texas Go Players" room, I've been getting tempted to challenge you. =)

Now to read the comments! I suppose I'm the only one who can 'directly' learn from the game-related comments, so I may as well work on doing just that.

topazg · **#147**

Good game!

daniel_the_smith · **#148**

I really enjoyed this thread, thanks to all three players.

If emeraldemon and observers want another CvH match, I'll volunteer to be the human.

Chew Terr · **#149**

Losing publically is embarassing. Losing to a bot publically adds insult to injury. Misreading a tsumego in front of a crowd, too. I'm not that affected by it, but I'm still left shaking my head. Part of the lesson to learn, I suppose, is to always consider long slides when you consider any slides at all. Eyespace is king, and if that would have made the difference, it's a big deal. Thanks to everyone for the interest in the game!

Tonight I'll just have to go beat up on some club members. =D They're more likely to fall for silly tricks!

emeraldemon · **#150**

@Chew Terr:
Thanks for the game! Next time we're both on KGS we definitely play. Or we could play in the L19 tournament: just make sure to win your first 3 games, and I'll do the same!

@daniel_the_smith:
Sure, I'd be happy to start up another game

One thing I'm curious about: how strong is Fuego with these settings? I don't think the game was too lopsided, Fuego might be a bit stronger than 3k, but I don't think too much stronger. How would you rate its play? Maybe we could try 2 stones, or just no komi?

Some thoughts:
There have been arguments in the past about whether or not it's good for learning to play against a computer. I think Fuego is strong enough and "standard" enough (i.e. fuseki isn't too strange) that it could make a reasonable sparring partner, and not give players bad habits. In particular, if you're in the 25k-15k range, I think you might actually get worse habits from playing humans your own level than from playing Fuego! Of course, Fuego can't give advice on how to improve the way a human can. But if you don't have access to strong players, it might be a good way to get some "school of hard knocks" lessons.

@lightvector:
I think what you said is basically right. In the post I made (viewtopic.php?p=69830#p69830) if you follow the first playout towards the end you'll see that black successfully lives on the top. I'm not sure of a better method than just looking at individual playouts.

daniel_the_smith · **#151**

emeraldemon wrote:

@daniel_the_smith:
Sure, I'd be happy to start up another game

One thing I'm curious about: how strong is Fuego with these settings? I don't think the game was too lopsided, Fuego might be a bit stronger than 3k, but I don't think too much stronger. How would you rate its play? Maybe we could try 2 stones, or just no komi?

I actually thought it played pretty well, it's hard to say. Perhaps no komi?

Mike Novack · **#152**

[quote="emeraldemon"

One thing I'm curious about: how strong is Fuego with these settings? I don't think the game was too lopsided, Fuego might be a bit stronger than 3k, but I don't think too much stronger. How would you rate its play? Maybe we could try 2 stones, or just no komi?

You mean time settings equivalent to to normal time settings on a powerful machine? I think it would be more instructive to demonstrate the practical playing strength of the program, realistic time settings on a good, but not super powerful machine. It's not silly that (in this case) the human has close to unlimited time to think. My gut feeling is that it will play weaker but not all that much weaker (the playing strength of the algorithm isn't linear with time). Might I suggest 1 hour basic, then maybe 5 in 60 sec, on a two core machine.

There have been arguments in the past about whether or not it's good for learning to play against a computer.
(without considering that these programs do not all play alike)

I think Fuego is strong enough and "standard" enough (i.e. fuseki isn't too strange) that it could make a reasonable sparring partner, and not give players bad habits. In particular, if you're in the 25k-15k range, I think you might actually get worse habits from playing humans your own level than from playing Fuego!
(I'd say up to 5-6k if feugo is around 3k. I don't think bad habits would be a problem against any of these MCTS programs as long as 3 or more stones stronger. And bad habits works both ways. You aren't going to confuse the program with a "trick" move or "swindle" it and as both players become stronger that sort of thing won't work so don't get into the habit of trying it on)

Of course, Fuego can't give advice on how to improve the way a human can. But if you don't have access to strong players, it might be a good way to get some "school of hard knocks" lessons.
(That is specific to the programs. It might not be the case that none of these programs can be asked to explain "why"; the go purpose behind a move).

124. Chew (3k) vs Fuego (Bot)

Who is online