dfan's quest for competence

dfan · Post by **dfan** » Wed Aug 24, 2016 3:51 pm

After a years-long lull (during which I finally achieved my big lifetime goal in chess), the AlphaGo - Lee Sedol match piqued my interest in go again, and it turns out the local scene is very vibrant and friendly, increasing my motivation. Plus the Go Congress was just held in Boston, which was really fun, although I didn't have the time to play in the Open. So here I am again.

I'm not going to set any rank goals here, since I think those are counter-productive (for me, at least), but I am trying to become more competent at the game. If I am successful at that, great, my rank will go up too. I have some ideas, based on my successful chess improvement, largely having to do with spaced repetition. We'll see what happens!

One resource that wasn't very available the last time I devoted energy to the game is streamers. I think it's incredibly useful to see how strong players think on their feet, and it's enjoyable too. I have to be careful not to spend too much time watching at the expense of playing, though.

I've been playing long games when I can and then going over them carefully with the help of Crazy Stone. It's not the same as a human teacher but I think I am still learning a lot. It will be interesting to see how much it helps.

On the study front, I started by going through my old standbys, Graded Go Problems for Beginners volumes 2 and 3 and Get Strong at Tesuji. Now I'm working on GGPB4 and 1001L&D, with some All About Life & Death thrown in (this is one place where I am hoping spaced repetition will help).

I had a couple of questions about tsumego problems that spurred me to finally create this thread, but as I was writing them up I came up with my own answers

Hopefully this thread will justify itself soon.

[I should add for reference, in case my rank goes up (or down!): I'm currently 4k AGA, though I last played in a tournament in 2008, and am bouncing around between 3k and 5k on KGS (few games means lots of volatility).]

Kirby · Post by **Kirby** » Wed Aug 24, 2016 4:13 pm

Glad to see your journal!

By the way, what was your lifetime goal in chess?

dfan · Post by **dfan** » Wed Aug 24, 2016 4:25 pm

Reaching a 2000 USCF tournament rating. I think of it as being essentially equivalent to 1 dan; when you get to that rank, even a strong player thinks you are basically competent. Which maybe gives away the true rank goal implied by the title of this journal

dfan · Post by **dfan** » Wed Sep 07, 2016 2:59 pm

Finally finished 1001 Life and Death Problems! I've started a few times before but never had the willpower to see it through. I don't know why many people seem to consider this a 10k book. The five-move problems were definitely a workout, and I didn't even get all the one-move-to-live problems right.

What I feel I got out of it:

Confidence in reading. I had to bear down and really see everything, especially since a lot of problems involve shortage of liberties of future groups. When playing chess I routinely calculate variations with 5+ moves for each side, but somehow even 5 moves (total) in go is a real brain-buster for me. But now I have confidence I can do it when necessary.
Hopefully it actually was good for my reading ability as well
Concentration. This is related to the first point; for each problem, I honestly thought I should be able to solve it (after all, it's only 5 moves), so if I couldn't see the answer right away, I kept at it until I came up with a solution, rather than just giving up after a few minutes and guessing at a vital point or something.
I definitely got some patterns drilled into me that I am now more familiar with than before.
I already knew that the number one question in life and death problems is What points will be the eyes? but this really drove it home. Every problem, the first thing I did was figure out where the two eyes could possibly be.

I didn't keep stats until I did the last 101 (in the five moves to kill section), of which I got 79 totally right, 9 right except for some minor detail (e.g., White's try in the solution wasn't the most testing, but it's true that I didn't explicitly look at it), and 13 either wrong or right by accident. So depending on your strictness, I got 78% to 87% right. Average time 2:20 per problem, though it's 2:11 if you throw out the 15 minutes I spent on one problem that turned out not to have a solution (according to the Errata page on Sensei's Library). Median time was only 1:18 because there were a few outliers.

I think next up will be the first Maeda Tsumego volume. I'll try to give it the same level of dedication I gave the last 101 problems of this.

Meanwhile:

Made flashcards for Get Strong at Tesuji
Making flashcards for All About Life and Death. I am getting a lot of these wrong, which is great; it means that there is a lot to get better at. We'll see how far I get with this.
Watching some livestreamed games. I enjoy this a lot and also learn a bunch from seeing how strong players think. I love watching Haylee's games but once she got to 9d Tygem I feel like the games are over my head a bit. Recently I've been watching the games of Longstride, an AGA 5d who plays slow games and who has a nice chill manner that I enjoy more than some of his more animated counterparts.
Actually playing games. As with many people I should probably study less and play more, but I usually get in at least one game a week at the local go club and one or two online. I go over all of them really carefully with Crazy Stone, which eats up a lot of my game-playing time, but I think it's time well spent.

Two local tournaments are coming up in October so hopefully I'll have the gumption to enter at least one of them.

jorden · Post by **jorden** » Thu Sep 08, 2016 7:57 am

Hi dfan, interesting reading. I was wondering if you could elaborate a bit on how you use Crazy Stone while going over your games? Do you use it to try out alternative moves and see how CS responds?

Jorden

dfan · Post by **dfan** » Thu Sep 08, 2016 11:30 am

Crazy Stone provides a couple of things in analysis mode:

An assessment of the current position with "Situation" (e.g., +5.5) and "Dispersion" (e.g., +-14), which I assume are roughly the median and standard deviation of the scores of the playouts from this point, as well as the "win rate" (fraction of the playouts that Black wins)
A list of moves it is considering for the side to play, with the number of playouts and win rate for each one

When I analyze a game with it, I go through the game move by move, looking at the moves it recommends to see if there's something I just didn't consider, and looking at the Situation value to see if I or my opponent played what it considers to be a clear mistake.

Then, as you say, I often try out alternative moves and see what variations result, which is the one way I can get some insight from its analysis since it can't do things like tell me "White's thickness more than makes up for Black's territory."

One problem, which comes up in computer-aided chess analysis as well, is that it might tell me that, say, it is clearly better for Black to play L8 than L7, but in fact L8 is only better because of a very tricky sequence that I would never find over the board, and in an actual game it would be much better to play the safer move instead. Sometimes I can get a sense of this by playing out a variation and seeing how crazy it looks.

While I do all this I annotate the game in a human way in a separate SGF file, as people do when they post their games here for comments.

dfan · Post by **dfan** » Fri Sep 09, 2016 4:03 pm

All right, I'm going through the first Maeda volume and I'm already confused. This is problem 26.

Click Here To Show Diagram Code: [go]$$B Black to play $$ --------------+ $$ . . . . . . . | $$ . . . . X X . | $$ . . O O X O X | $$ . . . X O O . | $$ . X . X O . O | $$ . . . X X O . | $$ . . . O O X . | $$ . . . . . X . | $$ . . . . O O . | $$ . . . . . . . |[/go]

I'll hide the rest of the comment in case you want to try it yourself first.

dfan · Post by **dfan** » Sat Sep 10, 2016 5:11 pm

Here's an example of the way I use Crazy Stone to analyze. This was a casual game (but we both spent plenty of time thinking) played live against someone approximately 5k.

I try to come up with some takeaways for every game. Usually some of them are things that I didn't know yet and hope will stick, and some are things that I should have known already and need execute in practice as well as in theory.

Up to problem 36 in Maeda volume 1. It's tough that some of the problems end in ko, but I think that's a good thing, since in games you never know whether ko is the best you can do. I also find it interesting that a bunch of these problems so far I can solve almost instantly (spot likely correct move, read out main variation, verify, all within 20 seconds) while some take me a few minutes (or I get them wrong!). I suspect that for a stronger player they would almost fall into the first category.

dfan · Post by **dfan** » Mon Sep 26, 2016 2:23 pm

I'm continuing to slowly work through Maeda v. 1, and am confused by problem 63:

Click Here To Show Diagram Code: [go]$$Bc Black to play $$ . . . . . . . . . | $$ . X X X . X X X . | $$ . X O . . O X O O | $$ . X O . . O X O . | $$ . . O . . . O . . | $$ ------------------|[/go]

The book's solution:

My solution, and question:

sparky314 · Post by **sparky314** » Mon Sep 26, 2016 2:35 pm

My thoughts:

1. In a ko fight (elsewhere on the board), black could let white live with one less point in the book's variation.
2. It protects against the peep/wedge at P4, in case black is weak on the outside.

dfan · Post by **dfan** » Sun Oct 09, 2016 5:38 pm

Just got back from my first live tournament since 2008! My AGA rating is currently 4k so I wasn't sure whether to enter at that rank or be aspirational and enter at 3k. In the end I decided to go with 3k and went 3-1, so I guess it was the right decision. I'm not sure yet what the effect will be on my rating.

Game 1: Black against a 2k (0.5 komi). I killed a big group and then let my foot off the gas pedal, letting him back in it, until I took advantage of some reduced liberties in the endgame to make a capture that clinched it, and he resigned.

Game 2: Black with 4 stones against a 2d. He never really tried to make things complicated and I cruised to a 30 point win.

Game 3: Black with 3 stones against a 1d. Had a nice opening, again relaxed a bit too much, got a big group cut off and even though I was able to use some threats to do a lot of reduction elsewhere, I still lost by 5.

Game 4: White against a 4k (0.5 komi). I made a huge moyo and when she tried to live in it, I wanted to take territory by attacking but instead got into a gigantic fight. Luckily, I won it.

All the tsumego I've been doing has definitely helped; there were multiple times that I explicitly used tricks that I've learned recently.

Independently of my actual result (which was good, obviously), I actually felt like a 3k, which was great.

[Edit: the tournament was just rated, and I am indeed officially an AGA 3k (-3.58).]

shoryuu · Post by **shoryuu** » Thu Oct 13, 2016 12:40 pm

dfan wrote:Crazy Stone provides a couple of things in analysis mode:
An assessment of the current position with "Situation" (e.g., +5.5) and "Dispersion" (e.g., +-14), which I assume are roughly the median and standard deviation of the scores of the playouts from this point, as well as the "win rate" (fraction of the playouts that Black wins)
A list of moves it is considering for the side to play, with the number of playouts and win rate for each one
When I analyze a game with it, I go through the game move by move, looking at the moves it recommends to see if there's something I just didn't consider, and looking at the Situation value to see if I or my opponent played what it considers to be a clear mistake.

Then, as you say, I often try out alternative moves and see what variations result, which is the one way I can get some insight from its analysis since it can't do things like tell me "White's thickness more than makes up for Black's territory."

One problem, which comes up in computer-aided chess analysis as well, is that it might tell me that, say, it is clearly better for Black to play L8 than L7, but in fact L8 is only better because of a very tricky sequence that I would never find over the board, and in an actual game it would be much better to play the safer move instead. Sometimes I can get a sense of this by playing out a variation and seeing how crazy it looks.

Do you have to find that tricky sequence on your own or will Crazystone tell you? What I mean is other than the breadth it covers, does it have depth once you click on one of the recommendations, will it carry on the analysis?

dfan · Post by **dfan** » Thu Oct 13, 2016 12:52 pm

shoryuu wrote:
dfan wrote:One problem, which comes up in computer-aided chess analysis as well, is that it might tell me that, say, it is clearly better for Black to play L8 than L7, but in fact L8 is only better because of a very tricky sequence that I would never find over the board, and in an actual game it would be much better to play the safer move instead. Sometimes I can get a sense of this by playing out a variation and seeing how crazy it looks.

Do you have to find that tricky sequence on your own or will Crazystone tell you? What I mean is other than the breadth it covers, does it have depth once you click on one of the recommendations, will it carry on the analysis?

I have to uncover the tricky sequence one move at a time by repeatedly executing Crazy Stone's top move and seeing what it thinks should happen next. It does not display a principal variation (PV) like chess analysis programs (and Leela) do.

shoryuu · Post by **shoryuu** » Sun Oct 16, 2016 6:42 am

dfan wrote:I have to uncover the tricky sequence one move at a time by repeatedly executing Crazy Stone's top move and seeing what it thinks should happen next. It does not display a principal variation (PV) like chess analysis programs (and Leela) do.

So you have to repeatedly click the analyze button everytime you click one of it's follow up moves? Boy is that a pain. Crazystone is just lucky they don't have competition. Would jump ship to a more efficient one in a heartbeat.

dfan · Post by **dfan** » Sun Oct 16, 2016 7:30 am

shoryuu wrote:
dfan wrote:I have to uncover the tricky sequence one move at a time by repeatedly executing Crazy Stone's top move and seeing what it thinks should happen next. It does not display a principal variation (PV) like chess analysis programs (and Leela) do.

So you have to repeatedly click the analyze button everytime you click one of it's follow up moves? Boy is that a pain. Crazystone is just lucky they don't have competition. Would jump ship to a more efficient one in a heartbeat.

No, I just have to click the follow-up move. In version 1.00 you did have to click on Analyze again every time, which was indeed a massive pain.

I'd say that version 1.01 upgrades the usefulness of Crazy Stone as a tool for analysis from 2/10 to 5/10. It is still far from being a pleasure to use, which is really too bad since the engine is apparently very strong. I don't think it is a priority for Unbalance.

Life In 19x19

dfan's quest for competence

dfan's quest for competence

Re: dfan's quest for competence

Re: dfan's quest for competence

Re: dfan's quest for competence

Re: dfan's quest for competence

Re: dfan's quest for competence

Re: dfan's quest for competence

Re: dfan's quest for competence

Re: dfan's quest for competence

Re: dfan's quest for competence

Re: dfan's quest for competence

Re: dfan's quest for competence

Re: dfan's quest for competence

Re: dfan's quest for competence

Re: dfan's quest for competence