Life In 19x19

Kirby wrote:

Kirby wrote:

sorin wrote:This variation appeared publicly the first time I know of in the AlphaGo-KeJie match.

I remember from Ke Jie’s match as well. I’ll see if I can find the Redmond video I saw - I believe it happened after Ke Jie’s game.

I’m having a hard time finding it. It could very well have been Ke Jie’s game, but I seem to recall Redmond reviewing this pattern. He’s done a lot of videos with Chris Garlock, though, so it’s going to take time to find this pattern among those videos...

Here's the exact moment in the video commentary when it was played: https://youtu.be/1U1p4Mwis60?t=3253

And here's the game on the Deepmind website (if you scroll down a bit): https://deepmind.com/alphago-china#match_2

Thanks for looking these up, sorin. I still feel there was something else with Chris Garlock, too. Kind of bugging me, now. I might be hallucinating. Either way, I think you’re right that the Ke Jie game might have been the first with this particular shape and result.

Perhaps slightly off the direct topic, but after a couple year hiatus I'm coming back to go and I found myself reading your study journal because I wanted to make sure your son was still doing ok. Of course in the process I read a lot of the rest of it over the past few years.

You talked a lot about Inseong saying you had too aggressive a style maybe last fall or so, and what he said coincided with my own opinion of your style. I've noticed, though, that you never seem to talk about counting. You have game positions and then say something like "I felt behind" or "black's area looked big" but it's not clear to me what exactly is driving those evaluations. I might recommend whenever you feel one way or the other that you practice taking a count of solid territory on the board, and then calculate potential territory that each side has. Once you have those numbers, you can decide on a calm course of action to keep the game in balance.

When you look at a big moyo on its own, it's easy to think that white's going to get a huge territory there. If you count and say well I have 40 points here plus potential for another 40 in these areas, and they have 10 points plus potential for 80 points over there, it gives you a solid understanding of what you need to do. When you see that all you actually need is to prevent your opponent from getting 10 points in an area, it's much easier to consider a calm reduction that helps them solidify territory, but just not quite enough.

From my own playing, I tend to play much more this way than all out fighting, and perhaps unsurprisingly my study routine was much different, to the point where I think my reading was below par for my level. My studying style was to read and play over pro games, and particularly older Japanese pro games, with the idea of familiarizing myself with what can be considered normal both consciously and unconciously. I think that this made me approach the game with the idea that if we've played an even number of moves, the board should be balanced unless someone has made a grievous mistake. So as long as I am achieving an objective I can feel free to sacrifice on a larger scale or allow my opponent to make points in an area. I also studied counting, and by extension endgame plays. I don't know that this is a magic bullet and don't claim to have ever been as strong as you, but since it seems very much the opposite of the way you seem to spend your go time, you might find it useful.

skydyr wrote:Perhaps slightly off the direct topic, but after a couple year hiatus I'm coming back to go and I found myself reading your study journal because I wanted to make sure your son was still doing ok.

Thank you. That means a lot. My son has been doing well for quite some time, now. I still worry about his health often; it will forever be a concern until one of us dies. In the meantime, I hope to have many good moments with him - I hope the new job doesn't disrupt that too much!

skydyr wrote:
I've noticed, though, that you never seem to talk about counting. You have game positions and then say something like "I felt behind" or "black's area looked big" but it's not clear to me what exactly is driving those evaluations. I might recommend whenever you feel one way or the other that you practice taking a count of solid territory on the board, and then calculate potential territory that each side has. Once you have those numbers, you can decide on a calm course of action to keep the game in balance.

I think your assessment is pretty good. I *don't* often try to take a precise count of the board. As of late, I don't really count at all - my choices are often driven by intuition.

I think practicing counting would be helpful (though, some things are hard to assess, such as thickness or potential).

It's a little hard to change one's style, but maybe when I play another game, I'll try to be conscious about counting. I'll try to remember and post it in a review here. If I post a review without mentioning the count, feel free to remind me that I reneged on my promise

I played another game with Young Seon Yoon. She noted that viewership was a little down on the videos I've been translating for her. So I don't know how the plan will be going forward. Hopefully, she keeps making videos.

Anyway, here is our game:



I won the game, but have kind of weird feelings about it. I felt like she went easy on me compared to last time - I don't really know why, it was just a feeling.

She noted after the game that I didn't fight as much in this game as in our last one, so she didn't have a chance to do anything.

So I don't know if it's more that she went easy on me, or that I fought less, but anyway, the game was somewhat uneventful. We went over a few variations, which can be seen in the file.

Note: I don't know what's up with the second variation from the beginning of the tree. I'll clean up the SGF later. But you can see the main game from the main line.

Whew. What a weird week it's been.

My start date for my new job was last Monday, March 16th. But with all of the COVID-19 stuff, I was told not to go into the office, and that onboarding was virtual. I don't get a physical computer, yet - I have to login to a virtual machine over the internet. It's gonna be that way for a little while.

As I had expected, things are fast paced. I like it, though, I'm a little worried I'll get left behind. I already have a task assigned to me, even though it's my first week, and I have to figure out how everything works. And then there's the added difficulty that since I'm remote, it's hard to chat with people regularly. Then there's the other bit where I live out in the country, and there's a cap on my internet data usage, which I already exceeded - so the internet connection is pretty slow.

Nonetheless, I'm enjoying the work. I have concerns about being fired already, but I'm doing my best to learn as much as possible and to be as productive as possible. I have an office in our basement, and every day I get up at the same time, make coffee, and get dressed in work clothes. It helps me think of work in the right mindset. As a result, I learned more this week than I did in the last several months at my previous job. I just really want to ramp up quickly.

My concern before was, because of the increased workload, I wouldn't be able to play much go. But after experiencing the first week, I think that's BS. With the amount of work I'm thinking about throughout the day, if I don't break up some time to do something extracurricular, I'm fairly certain I'll go crazy - or crazier than I already am.

So I'm back into studying go, when I can, during the evenings. It probably won't work when I'm on-call and stuff, but I want to try to keep some sort of hobby going.

On that note, I played a teaching game with Mark Lee today. He's a really cool guy. Here's the game:


Some general ideas/highlights:
1.) I have no idea what's going on when it comes to potential. I did try to count a few times in the game, and early on, I always counted me as being ahead. I gave no points or account of black's thickness. So I thought I was ahead in a game that I was not ahead in.
2.) I don't really know how to get better at estimating potential, except maybe to get more experience. Maybe when you read well and think about the actual flow of the game and how the stones can be used, you can get a better idea of the potential that's there. Still very mysterious and tricky to me, though.
3.) As usual, I overplay. Mark pointed out that I can't attack my opponent very effectively when my own group is weak.

---

I had a fun time. I'm going to try to be systematic in my hobby time going forward, too. That's the best way to make this work, I think...

I played on OGS today - because I couldn't get a game on KGS. And because the new Mac update doesn't let me run 32 bit apps so all of my wine stuff is broken (read: no Tygem or Fox).

Anyway, I don't really know OGS that well, but I have an account there, so I opened up a game. I played against a 2k. They were stronger than I expected, and I almost lost.

Here's the game:


The big thing to note is probably this one:

Click Here To Show Diagram Code

[go]$$Bc
$$ ---------------------------------------
$$ | . . . O X . X X O O O O O X X . . . . |
$$ | . X . O X . X O O X O X X X X . . . . |
$$ | . X O O O X X X X X O O O X X . X X . |
$$ | . W X O . O O X X O . O X . X X O . . |
$$ | . . X O O . O X O O O O O X X X O . . |
$$ | X . X O . X O X X X X X O X . O X O . |
$$ | . X O . O X O X O O X X X O O O . . . |
$$ | . O O O . X O X O . O X O . . O O O . |
$$ | . . O . X . X O O . . O . . O X X O . |
$$ | . O O , . . X O . , O . O O O X X . . |
$$ | . O X X X X O O X X . . . . O O X . . |
$$ | . . O X . X X X O . O . . . . . X . . |
$$ | . O O X . O X O O . O . . . X . X O . |
$$ | X X X . . O X O O X . . . . . X O X . |
$$ | . . . X X X O O X O O O . . . X O O . |
$$ | O O O O O O X X X X X O . . X , X O . |
$$ | O X X . O X . . . . . X X . . X X O . |
$$ | O X . X O X . . . . . . . . . X O O . |
$$ | . . X . O . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]

I should probably ignore white's threat here. I just felt so sad losing the stones, and I was in byo-yomi, thinking I'd find a way. I hallucinated, and imagined that this would be a threat later:

Click Here To Show Diagram Code

[go]$$Bc
$$ ---------------------------------------
$$ | . . . O X . X X O O O O O X X . . . . |
$$ | . X . O X . X O O X O X X X X . . . . |
$$ | . X O O O X X X X X O O O X X . X X . |
$$ | . O X O . O O X X O . O . W X X O . . |
$$ | . . X O O . O X O O O O O X X X O . . |
$$ | X . X O . X O X X X X X O X . O X O . |
$$ | . X O . O X O X O O X X X O O O . . . |
$$ | . O O O . X O X O . O X O . . O O O . |
$$ | . . O . X . X O O B . O . . O X X O . |
$$ | . O O , . . X O . , O . O O O X X . . |
$$ | . O X X X X O O X X . . . . O O X . . |
$$ | . . O X . X X X O . O . . . . . X . . |
$$ | . O O X . O X O O . O . . . X . X O . |
$$ | X X X . . O X O O X . . . . . X O X . |
$$ | . . . X X X O O X O O O . . . X O O . |
$$ | O O O O O O X X X X X O . . X , X O . |
$$ | O X X . O X . . . . . X X . . X X O . |
$$ | O X . X O X . . . . . . . . . X O O . |
$$ | . . X . O . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]

Obviously, no... So yeah, I underestimated my opponent a little bit and almost lost. Maybe I should try OGS out a bit more.

I have to say that I thought the AI analysis at the end of the game was kind of cool.

---

So anyway, yeah. Back to work. Weekend's over... I'll probably post here again near the end of the week - unless I find some time in the evening. We'll see what happens.

Kirby wrote:Whew. What a weird week it's been.

You too, hey? Glad the new job has got off to a flying start. That should help take your mind off other things happening in the world...

Kirby wrote:I don't really know how to get better at estimating potential, except maybe to get more experience. Maybe when you read well and think about the actual flow of the game and how the stones can be used, you can get a better idea of the potential that's there. Still very mysterious and tricky to me, though.

Yes, I have this problem too. Looking at the board after black 77 (your comment in the game), I see 40-50 points of territory for white (uncertain because maybe there's an invasion at or near L3), and about 30 points of solid territory for black plus a hard-to-estimate area in the left and centre. And I find it hard to imagine the left/centre expanding past 20 points and overtaking white's territory plus komi. But KataGo says no, black is ten points up here. But then you fast forward to move 149 and black has made some territory around K12, which I didn't see coming, White's also made some "extra" territory around M7, but the lower left has shrunk. So I guess that's black's potential in action.

How to see this coming ahead of time? I'll let you know if I ever figure it out. But it's probably something to do with thick/thin positions. Notice all black's forcing (sente?) moves: O5 (move 79), O3 (move 83), B7 (move 93), E18 (move 107) and so on. The fact that he's able to push you around like this suggests that he's outplayed you earlier in the game, but you and are aren't strong enough to see the weaknesses until it's too late.

xela wrote:

Kirby wrote:I don't really know how to get better at estimating potential, except maybe to get more experience. Maybe when you read well and think about the actual flow of the game and how the stones can be used, you can get a better idea of the potential that's there. Still very mysterious and tricky to me, though.

Yes, I have this problem too.

Everybody has this problem. I haven't gotten up to speed yet, but doesn't KataGo project ownership of individual points? Often that will change, I suppose, with who has the move, but averaging the ownership over Black and White to play ought to act like an influence map. eh?

xela wrote:Looking at the board after black 77 (your comment in the game), I see 40-50 points of territory for white (uncertain because maybe there's an invasion at or near L3), and about 30 points of solid territory for black plus a hard-to-estimate area in the left and centre. And I find it hard to imagine the left/centre expanding past 20 points and overtaking white's territory plus komi. But KataGo says no, black is ten points up here. But then you fast forward to move 149 and black has made some territory around K12, which I didn't see coming, White's also made some "extra" territory around M7, but the lower left has shrunk. So I guess that's black's potential in action.

How to see this coming ahead of time? I'll let you know if I ever figure it out.

Bill Spight wrote: Everybody has this problem.

Indeed, it's pretty tricky to measure. What struck me as interesting is that, e.g., around move 77 per xela's comment, Mark had some confidence that black was leading, and he had an idea of by how much (I don't remember if it was the same 10 points that KataGo evaluated, but he had a number if I recall correctly). I suppose it has something to do with intuition about the position. But it's tricky to me to know how to quantify this.

People say to count. So I can practice counting definite points. But this doesn't exercise my ability to evaluate potential. And in fact, I wonder if it even has a minus effect - that is, if I keep practicing counting and telling myself that white is ahead by X points - I'm reinforcing an incorrect evaluation of the board.

So I guess it means that I should practice counting, then adding some amount of points for potential by some sort of fuzzy math, maybe. I'n not really sure how to practice this.

Whatever the case, Mark seems to have it down pretty well.

Kirby wrote: I'n not really sure how to practice this.

We need a website or an app that automates what Bill did earlier.

Shows you random positions from amateur dan and pro games, at all stages of the game (opening, midgame, endgame), evaluates them with a strong bot, and then asks you to judge which side is ahead, and optionally asks you also by approximately how many points.

And ideally with configurable "difficulty level", where on easier difficulties it will give you positions with bigger differences relative to the stage of the game, and/or maybe something about the subsequent moves are expected to have fewer "tactics", etc.

lightvector wrote:

Kirby wrote: I'n not really sure how to practice this.

We need a website or an app that automates what Bill did earlier.

Shows you random positions from amateur dan and pro games, at all stages of the game (opening, midgame, endgame), evaluates them with a strong bot, and then asks you to judge which side is ahead, and optionally asks you also by approximately how many points.

And ideally with configurable "difficulty level", where on easier difficulties it will give you positions with bigger differences relative to the stage of the game, and/or maybe something about the subsequent moves are expected to have fewer "tactics", etc.

I think something like that would be a good start. Something feels like it could be refined, but I'm not sure that I can articulate my feeling very well.

Basically, when we are talking about a human concept like "potential", it's a model for some subset of what a computer might evaluate. So if you're able to practice on a site like this, you can try to estimate who is ahead by some amount of points, but then it'd also be educational to break down the components as to why. For example, if the answer to a given problem is that white is ahead by 10.5, and you guess correctly, it'd be interesting to know if your evaluation of definite territory was working in coordination with your estimation of the effect of potential.

Like if you miscounted definite territory and put white up in definite territory by 10 points, but at the same time underestimated white's potential by 10 points, you might still get the problem correct, but for the wrong reasons - you evaluation wouldn't extend to subsequent problems.

Then again, maybe there's no real problem here. If we had an automated site like that, you could probably just drill a bunch of problems, and eventually your internal model would be refined naturally.

I was able to get a game on KGS somehow.

Here it is:


I played as black against KGS 2d. I was KGS 1d?, so he only had 0.5 komi.

I lost, but there are some big mistakes:

Click Here To Show Diagram Code

[go]$$Bc
$$ ---------------------------------------
$$ | . . . X . X . . . . . . . . . . . . . |
$$ | . O O O X . X . . . . . . . X O X O . |
$$ | . X X X O X . . . . . . . X X O X O . |
$$ | . X O O O X . . . , . . . X O , O O . |
$$ | . O X . . O X X . . . . X O O . . O . |
$$ | . . O O O O O O X . . . X X O . O . . |
$$ | . . . X X O O X X . X X . X X O O . . |
$$ | . . O X . X X O . . X . X O X O . O . |
$$ | . O O O X X . O . O X X X O X O X . . |
$$ | . . O X X . . O X O X O O O O X X X . |
$$ | . O X X O X . O X O O X X B . X . O . |
$$ | . O O O O X . O O O X X . O O X . . . |
$$ | X O O X X X X X . O . X O . O O O . . |
$$ | . X X X . X . . X X X X X O . . . O . |
$$ | . X O X X X O O O . O X O O X X O O X |
$$ | . O O O O O O . O , . O . O O O X X . |
$$ | O . O X O X X O X O O O O X O X . . . |
$$ | . O X X X . . X X X O . X . X X . . . |
$$ | . X . . . . X . . O . . . X . . . . . |
$$ ---------------------------------------[/go]

This is retarded. "Gee, I'm in a capturing race. How about I remove one of my liberties before we get started, eh?"

I should just start taking liberties. In the game, it ended in ko, and he got a lot of my top reduced. So stupid!

Earlier:

Click Here To Show Diagram Code

[go]$$Bc
$$ ---------------------------------------
$$ | . . . . . X . . . . . . . . . . . . . |
$$ | . O O O X . X . . . . . . . . O X O . |
$$ | . X X X O X . . . . . . . X X O X O . |
$$ | . X O O O X . . . , . . . . O X O O . |
$$ | . O X . . O X X . . . . X O . X X O . |
$$ | . . O O O O O O X . . . X X O X O . . |
$$ | . . . X X O O X X . X X . X X O O . . |
$$ | . . O X . X X O . . X . X O X O . O . |
$$ | . O . O X X . O . O X X X O X O X . . |
$$ | . . O , . . . O X O X O O O O X X X . |
$$ | . . . X . X . O . . O X X . . X . O . |
$$ | . . O O . . . . . O . . . . . X . . . |
$$ | . . . . . X . X . . . . . . O O O . . |
$$ | . . . . . . . . . . . . B . . . . . . |
$$ | . . . . . . . . . . . . . . X X . . . |
$$ | . . O O . O . . . , . . . . O , X . . |
$$ | . . O X . X X . X . . O . . O X . . . |
$$ | . O X X . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]

Somehow, I feel the move above is too loose. Seems there should be a way to kill him or at least make profit on one of his 3 local groups.

Maybe just keeping simple:

Click Here To Show Diagram Code

[go]$$Bc
$$ ---------------------------------------
$$ | . . . . . X . . . . . . . . . . . . . |
$$ | . O O O X . X . . . . . . . . O X O . |
$$ | . X X X O X . . . . . . . X X O X O . |
$$ | . X O O O X . . . , . . . . O X O O . |
$$ | . O X . . O X X . . . . X O . X X O . |
$$ | . . O O O O O O X . . . X X O X O . . |
$$ | . . . X X O O X X . X X . X X O O . . |
$$ | . . O X . X X O . . X . X O X O . O . |
$$ | . O . O X X . O . O X X X O X O X . . |
$$ | . . O , . . . O X O X O O O O X X X . |
$$ | . . . X . X . O . . O X X . . X . O . |
$$ | . . O O . . . . . O . . . . . X . . . |
$$ | . . . . . X . X . . . . B . O O O . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . X X . . . |
$$ | . . O O . O . . . , . . . . O , X . . |
$$ | . . O X . X X . X . . O . . O X . . . |
$$ | . O X X . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]

---

Anyway, I'll get this game reviewed on Friday by Mark. He'll probably have some additional insight.

Now, it'll be even harder to get a game on KGS. I have two accounts, Brian and Kirby. Because I lost this game, my 1d? account is down to 4k?. And since I won a couple of games by fluke earlier, somehow Kirby is at 8d?. Hard to get a game against 2d-ish player with either of those accounts.

I guess I should play more games on one of them, but I don't have a ton of time for it. I wish wine still worked on my mac.

Browsed through the game quickly with KataGo - not a lot of play outs, so you can take this with a grain of salt. Either way, at least for entertainment's sake, I have to share this drop:


So yeah, that was a bad move.

A few thoughts on the most recent game:

First, did you actually count?

Second, at move 67, I'd be inclined to just connect the group out directly with H12. White is left with two stranded stones that he will have to figure out what to do with while black has successfully reduced white and has no weak groups and a lot of potential. Black also has a choice of targets for future action, like the white group at the bottom and the stone on the top right in addition to attacking those center stones on a large scale.

Third, at move 149, why not play O9 as white is inviting black to do? I count it as worth at least 28 points between the capture and saving the black stones, plus once that's taken black is connected and white still has to make life for his group. If white plays the fancy cutting move, black can just sacrifice those three stones and connect the left side group and then maybe aim at reducing white further around D6.