Life In 19x19

Hi, friends. I'm Tristan Hodges, a VR game developer that's been playing Go on and off for 10 years. Like so many, I discovered Go watching Hikaru no Go, and in the spirit of that would like to do my part introducing the game to a newer generation. I plan to do so by developing an all inclusive virtual reality Go training/learning/playing software.

My company, FoxLeap, has put out several VR games and I've got a few years of experience in virtual reality coding. Currently this project is called "For the Love of Go", and it aims to give brand new players an extremely interactive experience to learn the game, play against an AI of adjustable skill, and do joseki/tesuji/life-and-death challenges, all while immersed in a serene and peaceful setting.

For the Love of Go is currently about 15% done, with the most basic core functionality nearly finished. You can play a full 19x19, 13x13, or 9x9 game against a GTP engine (Pachi), play on a local board, or pit the AI against itself.

I am currently using Pachi version 12.4 as my engine. I had to learn the Go Text Protocol and how to integrate its input/output stream with my VR GUI, and it certainly took awhile with the documentation as lacking as it is.

I am making this post not just to introduce my project to you, but also to find other Go fans / engine experts, and open dialogue about what you would like out of a project like this, among other specific questions.

My first question that perhaps someone could answer has to due with a trend in Go AI that I'd like to address:
I've noticed Pachi doesn't often pass when the game is clearly over. It tries a bunch of "stupid" testing moves that are easily thwarted before eventually giving up and passing. Is there some way to tweak the pass threshold, or a statistic I should look for in the genmove output that could signify that the AI doesn't have much hope for a move to be a positive score gain?

My second question: What kind of learning experiences would you look for in a guided go learning software? Should beginners, after learning the rules, go straight into doing life and death problems, or learn basic joseki first? Let me know what you think.

Thanks all, I look forward to furthering the influence of Go with you all!

For those of you who would like to see what it's currently looking like at 15% done. Temporary scenery, textures, lots of debug elements just floating around, etc. But pretty much all aspects of an actual game of go work just fine. You physically reach into the bowl, pick up a stone, and place it on the board, and it feels pretty good even in VR.

foxleapgames wrote:My first question that perhaps someone could answer has to due with a trend in Go AI that I'd like to address:
I've noticed Pachi doesn't often pass when the game is clearly over. It tries a bunch of "stupid" testing moves that are easily thwarted before eventually giving up and passing. Is there some way to tweak the pass threshold, or a statistic I should look for in the genmove output that could signify that the AI doesn't have much hope for a move to be a positive score gain?

On the assumption that your targeted audience is that of beginners, I would not worry about requiring the bot to pass when the game is "clearly over". Beginners should keep playing until the game is clearly over to them. They should also try stupid moves that are easily thwarted by experienced players and bots. They do not yet know when moves gain nothing or are "stupid". In fact, most players make mistakes at the end of the game (dame stage) until they are 5 kyu or better, a rank that beginners may take a couple of years to reach, if the actualy manage to do so. That is because some "stupid plays" are not actually stupid. Like the bots, beginners should play by area rules so that stupid plays or not passing is not penalized.

My second question: What kind of learning experiences would you look for in a guided go learning software? Should beginners, after learning the rules, go straight into doing life and death problems, or learn basic joseki first? Let me know what you think.

Forget teaching beginners joseki. That has never been recomended for beginners, anyway. But now the bots have made many joseki obsolete. It has been like turning the clock ahead a couple of centuries. There is a book called 38 basic joseki. Maybe that number is now 10 or so. Beginners will pick up the joseki knowledge they need by playing and watching games.

As for educational material for beginners, I can't say. I don't know of any method that has been tested. As for myself, I made it to about AGA 4 kyu by playing against stronger players and reviewing my own games.

Something about the background makes me expect a Gorn will appear soon. It looks fun. Good luck!