Sente gains nothing, part 3: Ko calculations

One of the more common mistakes I see in any endgame-related discussion is related to the value of a kō capture. The usual touted value of ½ points only applies to the very last kō of a game; and the more knowledgeable players who have heard of the ⅓-point theory often use it wrongly in relation to other endgame values. This essay seeks to remedy the above mistakes while deepening the reader’s understanding on how endgame calculations work.

Continue reading “Sente gains nothing, part 3: Ko calculations”

The future is here: A professional-level go AI

Thursday, January 28 2016, 7:30 AM in Tokyo, Japan.

I wake up to my alarm clock as usual, preparing to go to the Nihon Ki-in headquarters in Ichigaya for my first day as an official game transcriber. Before my morning shower, I turn on my computer and check the news headlines of the day:

>Game Over? AlphaGo Beats Pro 5-0 in Major AI Advance

“…”

“Maybe it was a low-handicap match?” *reads a bit* “Wait, seems that’s not it, they were even games by Chinese rules.”

“…No, it’s not April 1st yet.”

“Maybe it’s a marketing ruse?”

>Nature: Go players react to computer defeat

“No, that’s not it either, they have too many important people commenting on the match.”

“…”

“Wow, I really am living in the future.”

Just a few weeks ago, I gave a television interview to Yle, Finland’s national public-broadcasting company, confidently saying that go-playing AI would definitely catch up to top human players, but that it might yet take dozens of years. It seems I cannot help but correct my statement.

Fan Hui, a Chinese professional 2 dan now living in France, is perhaps not at the very top of the go world, but he is reasonably high up there. A go-playing AI defeating him 5-0 surely implies a playing strength comparable to professional players. In Fan Hui’s defense, too, the five games played were perhaps not the best show of his skill. Apparently five unofficial matches were also played, which Google’s AlphaGo won 3-2.

In March 2016, AlphaGo is set to play against Lee Sedol, widely considered the current strongest human player. My personal assessment is that Lee Sedol will still win the match, with 4-1 or 5-0 sounding like plausible outcomes. However, AlphaGo’s match with Fan Hui was played in October 2015, and they say the AI is constantly getting stronger, so I might yet have to correct myself about this prediction as well… And even if Lee Sedol wins the match, in a year or two the situation may get reversed anyway.

The news was noticed in Japanese professional circles as well. When I was setting up the computer for transcribing, I overheard two professionals talking along the lines of “Did you read the news?” “Yes, it seems it was a Chinese professional, I wonder how strong he is.”

Finally, in case the reader is interested in what I thought of AlphaGo’s play, here are some comments jotted down by me about the first game in the five-game series! Enjoy!

 

Download SGF

 

PS. Personally, I’m all for go-playing AI getting stronger than humans. Once the AI get strong enough that they don’t copy human tactics anymore, we’re really getting to find out what the game is about!

Sente gains nothing, part 2: Fair exhanges

When estimating the score of an unfinished game, it is useful to know how to value unfinished territories fairly. Many amateur players prefer not to go through the effort, and instead spend their time in locating the largest move on the board. I would however argue that without evaluating the board position at all, the largest move will be challenging to find.

Continue reading “Sente gains nothing, part 2: Fair exhanges”

Sente gains nothing: An endgame view

This text begins a series of blog posts with a common theme: that of explaining what the proverb “sente gains nothing” means. My design is to start the explanation from easier, endgame-technical examples and little by little work towards more difficult concepts. If everything goes according to plan, at the end of the series I should have conveyed the idea of the proverb without (hopefully) having to explain it in abstract terms at all! (Having read essays on the art of human rationality on LessWrong, I wanted to try a similar concept on Go.)

At the same time, through the series I hope to improve on my writing skills. If the blog posts turn out well, I will probably refine them further into a part of an endgame theory book I have been planning for some time.

For readers unfamiliar with the terminology: please bear with me! I will also explain the technical meanings of sente and its opposite, gote in following posts.

Continue reading “Sente gains nothing: An endgame view”

Reaching professional status, and reinstating Go of Ten

Time sure flies! It has been well over a year since I last posted on Go of Ten. My apologies to the people who have asked for updates: I decided I wanted to direct my energy fully to improving on my game, and as a result blogging got left forgotten in the background.

As the Nihon Ki-in officially announced me as a professional go player from 2016 on, I believe we could say that the decision paid out! Starting from now on, I imagine I will have the time and will to write a blog post once in a while again.

My first official professional game is likely to get scheduled for late April or May. Before that, I will go back to Finland for two weeks to spend the Christmas with my family, and between January and April I will be attending professional study meetings, doing teaching activities and possibly writing a theory book in English on endgame calculations.

I would like to extend a big thank-you to all the people who supported me to get this far, and I hope you will keep following my adventures in the future as well!

Glift viewer

Recently I’ve tried to focus my own go studies, and when adding to this my occasional teaching work at the Nordic Go Academy, I’ve unfortunately had little energy to write on my activities in this blog. To try and give myself a motivation boost, I thought of trying a small technical change to the site, namely the Glift javascript viewer which Go Game Guru introduced recently. Below you can test the viewer out on a commented A class game, one of the few that I’ve managed to win.

Leave a comment and share your opinions!

 

Download SGF

 

Level up


The insei leagues of May concluded last weekend. Before the weekend I was comfortably on the first place in the C class with 13 wins and 5 losses, but as the end of the month drew near, pressure crept in. On Saturday I got two losses, which dropped me to shared second place, but on Sunday I was able to score full three wins, which got me back to the first place. As such, I will now for the first time get to play in the B class next weekend!

The insei play in two different rooms at the Nihon Ki-in, on the seventh floor. Classes from E to C play in a larger room, and classes B and A (each with 10 insei) are separated to a different, smaller room. The large training room tends to get noisy; not only because of the larger number of children, but also because of differing schedules between the classes. E class insei play six games a day, D class insei five games a day and C class insei three games a day, and whenever there’s a break between games, it’s not uncommon for the insei to move slightly away from the playing tables and then start discussing their games. Because of the different schedules between the classes, this means that most of the the day there is some background noise.

B and A class insei both play two games a day, save for the first and third Saturdays, when they play three games. The three-game days seem to be fashioned so that exactly 18 games get played each month, which translates to two games against each opponent. When playing three games a day, the time settings are same as in the C class (40 minutes main time and 40 seconds byo-yomi), and on days with two games, the settings are 1 hour of main time and 1 minute byo-yomi. All in all, next weekend I can expect significantly more thinking time, a more quiet playing room and more serious opponents, all of which sound great!

Lately I have been going to the Ichikawa dojo once a week. Last time, while reviewing some insei games from last weekend with Mimura-sensei, we came up with a few interesting tesuji for sabaki, which I would like to share this time in problem format. If the reader is not familiar with sabaki, I might first recommend them to see the lecture I held on sabaki for the Advanced Study Room on KGS not too long ago, accessible on their webpage.

fig1

The initial position was as above, with white just having played the circle-marked move. To amateur players, pushing towards the centre at the right of white’s circle-marked stone might come as an immediate reaction, but there is another move to consider as well…

noname_var1

…that is, black’s direct attachment at 1. If white responded with 7 to black 1, then black 2 would quickly produce resilient shape for black.

In this case, however, white might poke at black’s shape with 2, after which the sequence up to white 10 follows: black’s stones in the upper right corner area remain floating and a little bit weak.

Problem 1: When black plays the hane of 11, how does white make sabaki?

fig2

In the game, black didn’t attach on the white stone towards the right, but instead pushed up towards the centre. Some moves later, the board looked like above.

var2

For black, the kosumi attachment of 1 would be a simple and efficient way of play. White would have to keep himself connected with 2, after which black could force 3-7 in sente and then return to connect at 9, creating a promising moyo position on the lower side.

noname_var2

In the actual game, however, black attached at 1 here, mistakenly thinking that white would have to respond by cutting at j14. White’s counter-hane of 2 instead made it hard for black to figure out how to make sabaki.

Problem 2: It turns out, however, that after white 2, black has a great tesuji for sabaki available—but where?

Answers to these problems will follow later!

Hard mode insei training

The May insei leagues started last weekend. I again scored reasonably well with a 5-1 result, which currently puts me on the shared second place (this time with good promotion prospects, as the two insei sharing the place with me have lower positions in the league). Meanwhile in the D class, Leon from Germany performed a little below average with four wins and six losses. On the plus side for Leon, however, he was the first to score a win against one of April’s new insei, who had gotten something like 44 straight wins last month, and who was winning all of his games in the D class as well. For a quick comparison, two and a half years ago I lost three games in the E class. Because the readers no doubt would find Leon’s game interesting, I’m including it along with the English class teachers’ comments in this blog post. Last weekend, the U20 Globis cup was won by the Japanese Ichiriki Ryō 7 dan, a former participant in the Nihon Ki-in English class, and the second place got taken by Kyo Kagen 2 dan, of Taiwanese origin but playing for Japan. If I remember right, I saw a remark in a Japanese go newspaper or magazine that the last time Japan got a double win in an international professional tournament was over ten years ago. It is no wonder, then, that the Japanese go world is currently in high spirits. Last Monday me and Leon went for a visit to the Ichikawa go dojo, owned by Mimura Tomoyasu 9 dan, who was kind enough to invite both of us in. While Leon is returning to Germany already after June, I’m considering if I should continue my regular training in Ichikawa; not only is the dojo an optimal place for studying and its teacher incredibly nice, but it currently also has two B class insei. The only downside is that it takes me some 40-50 minutes to get there. As for the title of the blog post, that is something that came up recently in a talk with my teacher. As this time my goal is to actually become pro (instead of getting more experience like two years ago), my training is of course to be more rigid than before. Instead of just getting good results, I should strive for the content of my games to be good; after showing a few of my last weekend’s games, though 5-1 could hardly be called a bad result, I got a great deal of criticism for my moves and decisions. Finally, I was warned to not accept any compliments that I might hear from other professionals who see my games, as the Japanese have their culture of not speaking their mind in a direct fashion. It is also interesting to see how many mistakes one can still fit into a game even at this level of play. Below is an example from one of my insei games last weekend, which was a fairly comfortable and uneventful win. As I need to include some variations this time around, for a change I will go back to using move diagrams. I am playing white. Antti-insei_fig1 Up to black 17 we have an opening that at least I haven’t seen before. By the time white gets the cut of 14, he seems to have a comfortable position, but white still needs to be careful with his handling of the two top-left corner stones. White 18, while looking like a tesuji for sabaki, is dangerous. It might well be something that could be found in a bloody Chinese or Korean professional game, but it would have to be backed up by a lot of reading; and still, it would be making things unnecessarily complicated for white. Instead— Antti-insei_dia1 White should probably attach with 1 as in Dia 1, after which the sequence up to 8 could be expected, and white could take sente to for example claim 9 on the right side. The top-side result is in general favourable for white like this, apparent when comparing the white top-right corner with the black top-left corner and the white top-left corner group with the black top-right corner group. Antti-insei_fig2 While black found the most severe way to cut white back in Figure 1, black 23 and 25 in Figure 2 are vulgar. Up to 35, white lives easily on the top side with a good deal of territory. I had thought that cutting with white 36 and 38 then would come as a natural continuation, but it is not an interesting way for white to play at all; white’s top-side group gets confined in, and black can make relatively good shape on the outside with 43. Instead of 36— Antti-insei_dia2 White should move out directly with 1 in Dia 2. This way, white could force black to make bad shape with 4 by playing the atari of white 3, and up to white 13, white would live comfortably on both sides. Antti-insei_fig3 Continuing from Figure 2, white attached at 44 to make sabaki on the left side, similar to white 1 in Dia 2. This time black was forced to go back to capture a white stone with 47, and by the time white gets to extend to 52, he anyway lives comfortably. However, in the process black also got thick shape in the centre, so he wasn’t too badly off. White 64 was a simple reading mistake. Instead— Antti-insei_dia3 It was possible for white to connect with 1; even if black cut with 2 and 4 afterwards, white would keep his shape intact with 3 and 5. This way, white is also bound to get A in sente, which means something of a ten-point advantage over the actual game. Stronger readers might notice that black can prevent white A by forcing black B, white C and black D in sente, but in that case black loses the sente block at E, which is bigger. Back in Figure 3, black played tenuki to 65 too early; he should first block at 66 in sente. Now that white got to exchange 66 for black 67 in sente, white profited by about five points. Antti-insei_fig4 The game continued in a relatively normal fashion, as in Figure 4, up to black 75. Black again went for complications there; when white responded with the sabaki move of 76, black again responded in the most severe way with the cut of 77. Up to 85, white was content to fix his shape in sente, going back to connect at 86. Black then again went to challenge white with the tesuji-ish move of 87, forgetting about the endgame technique mentioned in Dia 3. White then found the time to exchange 88 for black 89, which meant a free profit of three points. Antti-insei_fig5 The rest of the game was fairly straightforward. While white’s play was far from optimal, he was still able to capitalize on black’s problems with timing, and by 142 white had generated a lead of almost 20 points. Upon seeing white 142, black resigned. Below is the Eidogo-plugin version of the game.

[sgfPrepared id=”0″]

Lastly, below is Leon’s game with the prodigy from the E class, along with comments by English class professionals.

[sgfPrepared id=”1″]

Post-Golden week, Globis cup

Insei training resumes tomorrow after a one-week break. The Golden week concluded last Tuesday, but my week continued in an event-rich manner afterwards, as well.

On Wednesday I went to play tennis with Shutō 7 dan, Tajiri 4 dan and a few other friends; it was the second time I played tennis in my life, so the quality of the game wasn’t exceedingly high (at least on my part; Shutō and Tajiri were in fact both quite adept). Then on Thursday, me and Leon went ice skating together with Mitani 7 dan and a few of his friends. It was only the day after that we realized that one of the friends, an incredibly friendly and easy-going Taiwanese person, was in fact Rin Kanketsu 7 dan; he had introduced himself as “Kankan”, and my brain didn’t work fast enough to make the connection on the spot. While I wasn’t good at tennis, I was still able to skate surprisingly well—maybe all those hours of gym in secondary school some ten years ago didn’t go to waste!

The first Globis cup, meant for players under 20 years old is taking place in Tokyo this weekend. On Thursday evening, me and Leon got to attend the tournament’s welcoming party, which in traditional Japanese way meant speeches and good food.

globis-cup
The players (and sponsors) of the tournament; how many can you recognize?

At the time of writing, two rounds have been played in the double knock-out tournament. Unfortunately, the European representative, Lukas Podpera 6 dan is already out with two losses, as is the favourite of the audience, Joanne Missingham 6 dan professional.

joanne-ichiriki
Joanne Missingham 6p and Ichiriki Ryo 7p face off in the first round
pros-analysing
Top professionals, including Yamashiro Hiroshi 9p, Yūki Satoshi 9p and Takemiya Masaki 9p in the front, analysing the games as they are played

This weekend’s insei games are played on the sixth floor (meaning the tatami rooms, instead of the usual rooms on the seventh floor). While my legs aren’t very much looking forward to the ordeal, otherwise I find the tatami rooms and legged boards to provide for a better feeling for the games than regular table boards. Then from next Monday on, I’ll start occasionally visiting the Ichikawa go dojo.

End of April, Golden week, Kamakura

April’s insei games got finished last weekend. My final score was 16 wins to 8 losses (last weekend was five wins and one loss), which just barely wasn’t enough to get promoted to the B class: one of my two competitors losing their last game would have sufficed. As such, I will now start at seat #4 in the C class next month. Meanwhile, Leon was able to keep his place in the D class with about 50% winning ratio. Below is the almost-final results sheet from the C class. Edit 4.5.2014: Apparently showing the results is not allowed, and the sheet is now removed. Right now Japan is celebrating Golden week, a series of national holidays that provides the longest continuous yearly holiday aside from the summer holiday. Consequently we don’t have insei training this week, and many Japanese people take time off to go on holiday trips and sightseeing. Yesterday on Saturday, then, me, Tom from the Nihon Ki-in, Leon and Leon’s German friend Nico went on a day trip to Kamakura, roughly an hour’s train ride away from Tokyo. The day was almost ideal for such a trip, aside for the fact that a huge number of Japanese people had come up with the same idea.

kenchouji
At the Kenchōji temple in Kamakura: Go Seigen and Kitani Minoru played one of their jūbango games in the house barely visible in the photo
crazy-icecream
Sweet potato flavoured gelato: a local speciality, apparently. While it certainly tastes weird, it’s surprisingly edible!
budda
The main focus of the sightseeing trip, the 13-meter-tall world’s second-largest Buddha statue

Other activities of the week included participating in several go meetings organized by my teacher, Kobayashi Chizu-sensei, and participating in a go study meeting at Mitani-sensei’s place, which among others included playing rengo with Leon, Ōba Junya and Matsumoto Takehisa. Finally, below are two of my last weekend’s insei games with professional commentaries.

 

Download SGF

 

[sgfPrepared id=”2″]