Page 4 of 8
Re: 135. daniel_the_smith (2d) vs Fuego (bot)
Posted: Wed Aug 24, 2011 3:11 am
by hyperpape
Every time Fuego evaluates a position, it redoes the tree. As it goes deeper in the tree, it applies less time to each node*. As a result, sometimes a continuation looks right because Fuego hasn't evaluated it very carefully. But once the actual move is on the board, Fuego can accurately look a little further into the future. In that regard, it's quite like a person's reading.
Re: 135. daniel_the_smith (2d) vs Fuego (bot)
Posted: Wed Aug 24, 2011 7:37 pm
by emeraldemon
@daal:
Basically what hyperpape said. As Fuego spends more time thinking about a position, additional playouts can cause other moves to appear better. Here's an explanation of search:
viewtopic.php?p=69830#p69830 @snorri:
I asked the same question to the Fuego development list, here's the response:
viewtopic.php?p=68939#p68939 One thing to keep in mind is that the developers are mostly interested in much shorter time limits, say 10 or 20 seconds per move. In that time frame, it's important for the search to narrow down quickly and make sure a move doesn't have a terrible consequence deeper in the tree.
If these don't make sense, ask me and I will try to explain in more depth.
$$Bcm31 No Prisoners
Click Here To Show Diagram Code [go]$$Bcm31 No Prisoners
Value: 50.8%, Fuego's optimism has been dropping steadily over the last few moves.
Sorry everyone, I accidentally deleted my notes with the sequence and moves chosen. The sequence looked something like this (from memory):
$$Bcm31 No Prisoners
Click Here To Show Diagram Code [go]$$Bcm31 No Prisoners
Re: 135. daniel_the_smith (2d) vs Fuego (bot)
Posted: Fri Aug 26, 2011 2:29 pm
by daniel_the_smith
$$Bcm31 No Prisoners
Click Here To Show Diagram Code [go]$$Bcm31 No PrisonersI debated between this move and the hane (b2).
$$Wc I don't like having gote after this.
Click Here To Show Diagram Code [go]$$Wc I don't like having gote after this.$$Wc I think this might be black's best play? I'm not sure where I'll go next, but I don't think I'll want to play b2 here.
Click Here To Show Diagram Code [go]$$Wc I think this might be black's best play? I'm not sure where I'll go next, but I don't think I'll want to play b2 here.$$Wc This is the other way I think it might go. I don't think Fuego will opt for this, though. Seems good for white.
Click Here To Show Diagram Code [go]$$Wc This is the other way I think it might go. I don't think Fuego will opt for this, though. Seems good for white.$$Wc I think this would make me happy. Black is too smart for this, though.
Click Here To Show Diagram Code [go]$$Wc I think this would make me happy. Black is too smart for this, though.
Re: 135. daniel_the_smith (2d) vs Fuego (bot)
Posted: Fri Aug 26, 2011 6:38 pm
by emeraldemon
$$Bcm31 No Prisoners
Click Here To Show Diagram Code [go]$$Bcm31 No Prisoners
Value: 50.8%
90% of playouts at the chosen move.
$$Bcm35 No Prisoners
Click Here To Show Diagram Code [go]$$Bcm35 No Prisonersinteresting play. The aji at
was pretty big, but this feels a tad slow to me. Fuego really likes that thick, influence style though, so maybe it can find a way to make this work. One thing I've noticed in these games: Fuego tennukis more than I do
Re: 135. daniel_the_smith (2d) vs Fuego (bot)
Posted: Sat Aug 27, 2011 6:42 am
by daniel_the_smith
Man, that is not a tenuki I would have made as black.
$$Bcm31 No Prisoners
Click Here To Show Diagram Code [go]$$Bcm31 No PrisonersI think black has to push from behind or play on the 2nd line. I was going to play up one point, but I don't like this:
$$Wc
Click Here To Show Diagram Code [go]$$WcI can't take the stone on the edge and protect my cutting point at the same time.
Re: 135. daniel_the_smith (2d) vs Fuego (bot)
Posted: Mon Aug 29, 2011 12:33 pm
by emeraldemon
$$Bcm31 No Prisoners
Click Here To Show Diagram Code [go]$$Bcm31 No Prisoners
Value: 50.8%
Fuego's opinion is holding basically steady.
$$Bcm37 No Prisoners
Click Here To Show Diagram Code [go]$$Bcm37 No PrisonersNo way Daniel will let Fuego off with that
. I think Fuego may be in a tighter spot than it realizes...
Re: 135. daniel_the_smith (2d) vs Fuego (bot)
Posted: Mon Aug 29, 2011 7:21 pm
by daniel_the_smith
$$Bcm31 No Prisoners
Click Here To Show Diagram Code [go]$$Bcm31 No PrisonersI wanted to double hane but it doesn't seem to quite work.
I'm fairly happy with the situation. I think fuego gave up a large local advantage with that tenuki. I plan to push along the 4th line here for a bit. Then I have to decide if b2 is big enough (probably not, but it looks so huge). I also have to figure out: reduce/destroy, or solidify? I think I will have sente after the next few moves. Do I defend in the upper right? Destroy black's moyo in the center-right? destroy black's area at the top? I'm not sure yet.
Re: 135. daniel_the_smith (2d) vs Fuego (bot)
Posted: Tue Aug 30, 2011 4:20 am
by flOvermind
$$Wc No Prisoners
Click Here To Show Diagram Code [go]$$Wc No PrisonersAm I the only one who thinks the marked black stone was an excellent move?
Granted, white will build influence with the bottom left group, pushing black low. But the white wall will face a very strong black position. K3 and J8 do an excellent job reducing the influence. Black has not a single weak group, so what will white do with the wall?
Now compare this with white having cut. That way, white could chase a weak group. And that just has to be worth more than building a strong position that's already limited in scope... Especially since black then has a weak group, there is a bigger chance the influence in the bottom left will come to good use, even though it will be a bit weaker because of the missing tenuki.
Re: 135. daniel_the_smith (2d) vs Fuego (bot)
Posted: Tue Aug 30, 2011 6:04 am
by ez4u
flOvermind wrote:
$$Wc No Prisoners
Click Here To Show Diagram Code [go]$$Wc No PrisonersAm I the only one who thinks the marked black stone was an excellent move?
Granted, white will build influence with the bottom left group, pushing black low. But the white wall will face a very strong black position. K3 and J8 do an excellent job reducing the influence. Black has not a single weak group, so what will white do with the wall?
Now compare this with white having cut. That way, white could chase a weak group. And that just has to be worth more than building a strong position that's already limited in scope... Especially since black then has a weak group, there is a bigger chance the influence in the bottom left will come to good use, even though it will be a bit weaker because of the missing tenuki.
I think there are a couple of problems locally with
. First, it makes the R11 stone over-concentrated. If
were played first, Black would not choose to extend to R11. At the very least, I think Black would prefer to hit White's shoulder at Q12. Second, imagine that White captures the single stone in the bottom right as below. Normally Black would like to use the hanging connection of
. Here though
has the threat to cut on either side of the P6 stone. If Black connects at "a" instead, the White hane at "b" becomes sente yose.
clings too tightly to the R8 stone, which plays no strategic role in the game since the White corner is completely alive (Black would have to play both Q2
and S8 before T3 would threaten White's eye shape).
$$Wc No Prisoners
Click Here To Show Diagram Code [go]$$Wc No Prisoners
Re: 135. daniel_the_smith (2d) vs Fuego (bot)
Posted: Tue Aug 30, 2011 7:38 am
by emeraldemon
$$Bcm31 No Prisoners
Click Here To Show Diagram Code [go]$$Bcm31 No Prisoners
value: 50.6%
Sequence:
$$Bcm39 No Prisoners
Click Here To Show Diagram Code [go]$$Bcm39 No PrisonersFuego changed its mind again. From this sequence it almost seems that black wants to play on top as a kind of kikashi in sente before returning to the lower left. It did also consider this variation:
$$Bcm39 6% of total playouts
Click Here To Show Diagram Code [go]$$Bcm39 6% of total playouts@ez4u: thanks for pointing out that cutting potential, I missed that.
Re: 135. daniel_the_smith (2d) vs Fuego (bot)
Posted: Tue Aug 30, 2011 10:22 am
by daniel_the_smith
Does b6 kill that corner? I think it might. Must read.
Re: 135. daniel_the_smith (2d) vs Fuego (bot)
Posted: Tue Aug 30, 2011 12:59 pm
by daniel_the_smith
$$Bcm31 No Prisoners
Click Here To Show Diagram Code [go]$$Bcm31 No PrisonersLetting black have two moves up there is huge, but so is this corner. My stones on top are flexible still. I will be fine.
Plus:
Re: 135. daniel_the_smith (2d) vs Fuego (bot)
Posted: Tue Aug 30, 2011 2:50 pm
by emeraldemon
$$Bcm41 No Prisoners
Click Here To Show Diagram Code [go]$$Bcm41 No Prisoners
Value: 50.2%, slowly but steadily dropping
Sequence:
$$Bcm41 No Prisoners
Click Here To Show Diagram Code [go]$$Bcm41 No PrisonersI don't really like black's choice, I think white will get too strong this way. I think white will block the other way, protecting the corner.
Re: 135. daniel_the_smith (2d) vs Fuego (bot)
Posted: Tue Aug 30, 2011 3:50 pm
by daniel_the_smith
$$Bcm41 No Prisoners
Click Here To Show Diagram Code [go]$$Bcm41 No PrisonersMoves like this should always make white feel good. I was thinking black would play Q18, which would give me more trouble, I think.
Re: 135. daniel_the_smith (2d) vs Fuego (bot)
Posted: Thu Sep 01, 2011 6:47 am
by emeraldemon
$$Bcm41 No Prisoners
Click Here To Show Diagram Code [go]$$Bcm41 No Prisoners
Value: 50.1%
Sequence:
$$Bcm43 No Prisoners
Click Here To Show Diagram Code [go]$$Bcm43 No PrisonersI like black taking the side in sente, but I think white can play at C7 instead and threaten the hane underneath, making a weaker shape for black.
When the move chosen gets less than 90% of playouts, I try to post the other moves considered:
$$Bcm41 No Prisoners
Click Here To Show Diagram Code [go]$$Bcm41 No Prisonersa - 58%
b - 17%
c - 14%
was pretty big, but this feels a tad slow to me. Fuego really likes that thick, influence style though, so maybe it can find a way to make this work. One thing I've noticed in these games: Fuego tennukis more than I do 
. I think Fuego may be in a tighter spot than it realizes...
. Here though 
has the threat to cut on either side of the P6 stone. If Black connects at "a" instead, the White hane at "b" becomes sente yose. 
