How to play a 19x19 20k game?

Hello, as a newbie (25k) how do I play 19x19 against the computer please to practice and see if I can improve my rank? The lowest to play against the computer seems to be 11k, or am I doing something wrong please? Thank you.

2 Likes

PS Forgot to say (or wasn’t very clear), I have been playing the 11k computer and learning but I have to play them unranked as it says I can’t play them ranked because the difference is more than 9 so how do I ever get above 25k when I have to play the computer privately (unranked)?

Hello, welcome to OGS. It looks like the 11k bot is the lowest ranked bot to play against at the moment. This can change, because the bots are run by individual users, not by OGS itself, and I recall seeing a lower-ranked bot there before. Until then, you have a couple of options for increasing your rank:

  1. Play the 16k bot on a 9x9 board with handicap.
  2. Play humans (it’s not that scary, and there’s a lot of new players here who aren’t confident either).

If you’re a little computer savvy, you can download https://sourceforge.net/projects/amigogtp/ and have a beginner bot to play against on your own device (not OGS). Of course, that won’t increase your rank here. I hope you find a good solution.

6 Likes

Thank you for your help. I have played a few real games (real people) but the lower ranked games don’t come up much so I have only played 2. I’ll have to start my own game and see who wants to play a newbie :grinning:

2 Likes

This may help: https://github.com/online-go/online-go.com/wiki/Finding-(starting)-a-game#custom-game
Creating custom games is typically a quicker way to get a game.

2 Likes

Look up amybot maybe… But try and play real players. Bad bots are SO bad they will teach you bad things.

3 Likes

My recommendation would be to play against the 11k bot but give yourself a handicap of as many stones as you need. Say 9 stones. It should take only a game or two until you learn what’s going on and start beating it with such odds.

Also, if you’re having a hard time finding bot games on OGS with your current rating, you may want to practice offline with the following resources:

Leela: https://www.sjeng.org/leela.html
This is a free program that is rated 4 Dan out of the box. It also has a very helpful option to turn on a heatmap that will show you what Leela thinks are the next best moves. VERY useful teaching aid for beginners.

Alternately, you can download something like Sabaki: https://sabaki.yichuanshen.de/
and load it up with various go AI engines like AQ
https://github.com/ymgaq/AQ
(note: this requires a bit more advanced installing and configuring, whereas Leela is just install-and-go)

If you want to play in a browser, there’s this AI which is rated about a 2 Dan, and also has an option to enable a heat-map of the best next moves
https://chrisc36.github.io/deep-go/

Lastly, if all of these bots are too high-ranked, there’s always Gnugo which is rated between 14 kyu and 7 kyu depending on the settings
https://www.gnu.org/software/gnugo/

If you want to run Gnugo, you can install it in an SGF editor like Sabaki, or (if you want something simpler) there’s also Panda glGo
https://www.pandanet.co.jp/English/glgo/

good luck!
tonyb

1 Like

I think that already happened to me. I was doing OK against the bots but then played a real person and was like “OMG” lol.

1 Like

Thanks Alex, I have been doing that but because it will only let me play those 11k games privately, even if I win I stay at 25k :grinning:

1 Like

Thank you :slight_smile:

Thank you, the Leela one sounds good. I shall have a look at that tomorrow. I could do with some handy hints :smile:

1 Like

Just a word of warning; Leela (Zero) is a bot that makes just about anybody go “OMG”, including professionals :wink:

1 Like

Yeah no disrespect to tonybe, but considering you post was about finding easier AI… Maybe ignore everything he said.

All the bots he listed are some of the stronger bots available… So if you though the 11k bot was too much, he gave you almost exclusively bots that are WAY harder.

Iirc even gnu go is rated like 12k?

On top of that all those bots will play moves IMMEDIATELY after you place, and will probably encourage you to think less.

1 Like

With regard to beginner players playing high-ranked bots - yes, of course you folks all have excellent points. If one is stymied by an 11 kyu bot, trying to play a 4 dan bot will be very confusing and discouraging - complete agreement.

What I was trying to point out to the O.P. was that some of the AI I had listed had these very handy heat-maps that give the beginning player a glimpse of how an experienced player sees the game from a strategic and tactical point of view.

Of course it might take months or years for that beginning player to make sense of those strategic frameworks but - in the absence of a human teacher to help us see these things in our own good time - there are very few other resources which will help ease the pain of learning it the hard way (by “losing your first 1000 games quickly,” etc…)

And yes, there is also a lot of conventional wisdom that says that learning specific tactics from Leela can teach beginners bad habits, or that the high-level strategies are so far over even advanced player’s heads that beginners shouldn’t even go there.

To this point of view I say - again - for the beginning player, even a small taste of something that’s pointing them in the right strategic direction is better than nothing.

And yes, in order for that learning to happen, watching a high-level AI work out the strategy needs to be paired with playing actual humans on one’s own level, watching YouTube videos which break down elements of go strategy, and all the other well-worn steps that one should take in order to become better at go. I absolutely agree that just learning from AI heatmaps is not enough.

For myself, the thing that really worked was pitting the Leela AI against the 2 dan AI in the browser (this one: https://chrisc36.github.io/deep-go/) because both of them had a heat-map of the next moves, I could see where they matched and when they were drastically different.

The thing that was interesting for me is - neither of these 2 AI ever made any obviously BAD moves - they played a balanced opening, they always made shape, they knew when to tenuki. And yet - the 4 dan one always beat the 2 dan one - often by quite a large margin.

It took me a long time of studying and watching those games to figure out the Secret Sauce - the 2 dan AI was myopic and focused on local fights, while the 4 dan one was able to take advantage of opportunities created by whole-board play, and was better at stealing eye-space and capturing big dragons.

Of course - I know that’s not some final Answer At The Back Of The Book - it’s only one possible strategy out of a thick, twisty jungle of available go strategies. But again - it taught me how to look for opportunities on the board, and think about

  • who has sente
  • how to use direction of play
  • how to know when a particular group is settled, etc

in a way that would have taken me much longer without that teaching aide

Am I going to try to copy some corner-stealing-fuseki the way Leela does? Probably not! Most of those are still too advanced and over my head. But at least I’ve learned to invade when my opponent threatens to make large frameworks in the opening, and it would have probably taken me years to learn that on my own, just playing the occasional correspondence game with another 15-18 kyu and wondering why I’m always behind by move 70.

That’s my 2 cents, anyway.
cheers
tonyb

2 Likes