Learning Go / Weiqi / Baduk

[rogerclee]

I want to take a week or two after the Denver nationals to learn Go since my dad has been playing a lot at something like a club level, so for our Christmas gift to him my siblings and I are considering learning the rules of Go so we can play a game or two with him.

 

Is there something like Fred's LTPB software but for Go? What are the best online resources for someone who does not even know the rules? Is there a standard best beginner Go book that all good players recommend?

[billw55]

There's these:

 

http://playgo.to/iwtg/en/

http://www.usgo.org/learn-play

 

For beginner books, Janice Kim's Learn to Play Go series is reasonably well known.

 

Danger, you can get drawn in very deeply. But it's worth it.

[PhantomSac]

 

Danger, you can get drawn in very deeply. But it's worth it.

 

Haha yeah my last friend who got into go ended up quitting bridge. DON'T DO IT CLEE!!!!!!!!!!

[mgoetze]

The way to get good at Go is to lose as many games as you can, i.e. play speedball against the strongest opposition you can find. Thinking hard about your moves as a beginner is much less valuable than just building up a mental database of patterns by actually playing. Also, don't play on the full-size board (19x19) right away, absolutely start with 9x9, as you need a basic grasp of the tactical elements before you can even start thinking about the strategic ones.

 

You may find useful information here: http://senseis.xmp.net/?PagesForBeginners

 

Recommended place to play online: http://www.gokgs.com/

[billw55]

For me KGS only works through the android app. Desktop client (and hence decent screen size) is essentially broken because of java issues. Perhaps you can try OGS, although I don't know how much traffic they get. IGS is still busy, but hardly anybody there speaks english, so you need to know what you are doing first.

 

Agree that playing fast is better. Be prepared to lose every game for a while, unless you can find another raw beginner to play with. It helps to have a teacher at first, even a low level player just to help with mechanics. Otherwise you will likely have difficulty figuring out when the game is over.

 

About small boards I agree to an extent. But I think it is fine to play full board as soon as you are able to play out a complete game without help and correctly count the score. Just keep in mind playing quickly. Most players will be happy to play a beginner sometimes, but it can get irritating if they take too long.

[mgoetze]

About small boards I agree to an extent. But I think it is fine to play full board as soon as you are able to play out a complete game without help and correctly count the score.

That's definitely way, way too early. I have seen such games played by beginners on KGS and I honestly have no clue how on earth that is going to help them get any better at the game.

[cherdano]

Easy and great way to start is atari go: set up a cross-cut in the middle of a 9x9 board, and play "the first one to capture 1 stone wins". Once this seems boring (it actually isn't!) increase 1 to 2, to 3, to 4, ...

[billw55]

Easy and great way to start is atari go: set up a cross-cut in the middle of a 9x9 board, and play "the first one to capture 1 stone wins". Once this seems boring (it actually isn't!) increase 1 to 2, to 3, to 4, ...

I agree.

 

I find this interesting. The atari-go method represents a different philosophy about teaching beginners from what I experienced. It puts capturing as the primary mechanic/goal, rather than surrounding territory, which is the way I was first taught. It was only much later that I asked myself: what is territory really? -and realized that it is only an area where you can capture any stone the opponent places. I wonder how different my game would be if I had learned this way.

[rogerclee]

OK, good tips, thanks everyone!

[Mbodell]

To play many games there is also a free go computer program at igowin (made by a bridge player). The author of this also makes one of the strongest computer go programs (Many Faces of Go, multiple time world computer GO champion). The free versions are a free windows version and also a free tutorial/beginner iOS version. The AI adjusts to your ability, and you can start the 9x9 matches with a 5 stone handicap, and as you learn and win have the handicap shrink. It is relatively easy to understand what an eye is, what a ladder and net is, etc. but once you have some idea of that, try playing games and see what happens. At least that seems reasonable to me, but I have very little go skill myself, so who knows.

[fromageGB]

On linux I use gnugo, with quarry as the GUI. No tutorial but plays a good game. Its rating is about 7 kyu, say average club regular player level, but "the many faces of go" (not on linux) plays at a higher level.

 

Nothing beats F2F for learning, though, and with two new starters learning together and attempting to leapfrog each other, it will be quick to get to a basic standard. I used the Ishi press books, which may be available 2nd hand.

 

I found that play on a 9x9 makes you concentrate too much on captures rather than territory, and territory is key, so I'd say switch to 19x19 as soon as you feel you know what you are doing.

[PassedOut]

Very interesting AI result reported today in Wired: In a Huge Breakthrough, Google’s AI Beats a Top Player at the Game of Go

 

The DeepMind system, dubbed AlphaGo, matched its artificial wits against Fan Hui, Europe’s reigning Go champion, and the AI system went undefeated in five games witnessed by an editor from the journal Nature and an arbiter representing the British Go Federation. “It was one of the most exciting moments in my career, both as a researcher and as an editor,” the Nature editor, Dr. Tanguy Chouard, said during a conference call with reporters on Tuesday.

 

This morning, Nature published a paper describing DeepMind’s system, which makes clever use of, among other techniques, an increasingly important AI technology called deep learning.

Was bound to happen sooner or later, but this is definitely on the sooner side.

[mgoetze]

Very interesting AI result reported today in Wired: In a Huge Breakthrough, Google’s AI Beats a Top Player at the Game of Go

It is very interesting, but the article does a pretty poor job of providing context as to the skill differences between Fan Hui, Norimoto Yoda and Lee Sedol. (The indiscriminate application of the term "grandmaster", which doesn't exist in Go, certainly doesn't help.)

 

Also "oh wow, they had the system play against itself, what a mindblowing development" (my paraphrasing) is rather silly - the same thing was done with backgammon programs, what, 20 years ago? I guess the real news is they managed to develop a system that was capable of profiting from playing against itself, because that was of no use whatsoever for the previous Monte Carlo Tree Search algorithms.

[billw55]

Was bound to happen sooner or later, but this is definitely on the sooner side.

Sigh. So much for saying Go is the last frontier of AI gaming. I feel ... relegated.

[barmar]

It is very interesting, but the article does a pretty poor job of providing context as to the skill differences between Fan Hui, Norimoto Yoda and Lee Sedol. (The indiscriminate application of the term "grandmaster", which doesn't exist in Go, certainly doesn't help.)

They said that Hui is the reigning European champ, and Sodol holds the second most titles ever and the most in the past decade -- that seems like adequate context to me, and they also provide links to relevant Wikipedia pages. Only Yoda is described with the generic term "grandmaster" and no Wiki. How much more detail do you expect them to go into for a general-interest publication? How would you explain the skills of Meckwell, Hamman, or Zia if you were writing a similar article about bridge?

[mgoetze]

They said that Hui is the reigning European champ, and Sodol holds the second most titles ever and the most in the past decade -- that seems like adequate context to me,

The missing context is that there are hundreds of Go titles that are harder to win / demonstrate more skill than "European Champion".

[cherdano]

They said that Hui is the reigning European champ, and Sodol holds the second most titles ever and the most in the past decade -- that seems like adequate context to me, and they also provide links to relevant Wikipedia pages. Only Yoda is described with the generic term "grandmaster" and no Wiki. How much more detail do you expect them to go into for a general-interest publication? How would you explain the skills of Meckwell, Hamman, or Zia if you were writing a similar article about bridge?

You really have a talent of proving other people right while arguing against them. If Fan Hui were to play 50 games against Lee Sodol, I would bet against him winning a single game.

[nige1]

The missing context is that there are hundreds of Go titles that are harder to win / demonstrate more skill than "European Champion".

In the 70s, there was a London Pub, with dozens of Go tables. I was taught by a future European Champion. He took me to an exhibition by a 9-dan Japanese master. My friend challenged his beautiful young daughter. He expected an equal game because he was then 1-dan and the daughter was also 1-dan; but the girl insisted on giving him 4 stones, because she was professional 1-dan. She wiped the floor with him.

 

Nevertheless, on this performance, AlphaGo is a big step forward in AI.

[nige1]

Also "oh wow, they had the system play against itself, what a mindblowing development" (my paraphrasing) is rather silly - the same thing was done with backgammon programs, what, 20 years ago?

From 1952-1959, Arthur Samuels programmed a draughts/checkers-program, with an evaluation-function that was improved by playing the program against itself. It kept a book of previously evaluated positions and used mini-max searches. It was rated as a good beginner but (rather like AlphaGo) beat it's programmer and a Connecticut checkers-master

[barmar]

The missing context is that there are hundreds of Go titles that are harder to win / demonstrate more skill than "European Champion".

Not knowing anything about th Go community, I find this surprising. Is the European field really weak? Or is the European Championship poorly structured, so that it's easier to win by luck rather than pure skill (like the difference between a 4-session bracketed KO and the Spingold or Bermuda Bowl)?

[blackshoe]

In the 70s, there was a London Pub, with dozens of Go tables. I was taught by a future European Champion. He took me to an exhibition by a 9-dan Japanese master. My friend challenged his beautiful young daughter. He expected an equal game because he was then 1-dan and the daughter was also 1-dan; but the girl insisted on giving him 4 stones, because she was professional 1-dan. She wiped the floor with him.

 

Nevertheless, on this performance, AlphaGo is a big step forward in AI.

Perhaps she should have given him more stones. :)

[mgoetze]

Is the European field really weak?

Yes. Because of the "intuitive" nature of the game, it is really hard to get stronger without outside assistance, and typically you are limited by the teachers available to you. All the best players are in Korea, China and Japan and this is self-perpetuating. (Of course due to the massive popularity of the game there they also have a larger talent pool.)

 

A typical pattern of improvement in Go is that you learn things that take you to the next level, but are fundamentally flawed so that you have to unlearn them again to get to the level beyond that. This can be avoided by having a very strong teacher available. Children going to Go schools in Asia will make much more rapid progress than anyone else.

 

Fan Hui chose to be a star in Europe rather than a bottom-tier professional in his native China.