How many chess programmers are turning to Go?

[Rémi Coulom]

Johan de Koning, Peter McKenzie, and Chrilly Donninger also tried Go.

Go is really a lot more exciting than chess. Monte Carlo was a major breakthrough, but it feels like we still need to find a few more big ideas. Currently, the top programs have almost stopped improving for 1-2 years. It is difficult to predict, but I feel that reaching pro level may take a while.

Rémi

[PK]

I play Go mostly for recreation, having reached the level of European 6 kyu. As for programming, a half-hearted attempt convined me that entry barrier is much higher than for computer chess. Beside that there seems to be an aesthetic problem with computer Go. Even top or near-programs play relatively ugly, one-sided game, going for the center of the board and easily entering do-or-die situations.

[lucasart]

Johan de Koning, Peter McKenzie, and Chrilly Donninger also tried Go.

Go is really a lot more exciting than chess. Monte Carlo was a major breakthrough, but it feels like we still need to find a few more big ideas. Currently, the top programs have almost stopped improving for 1-2 years. It is difficult to predict, but I feel that reaching pro level may take a while.

Rémi

I just learnt to play Go today. It's fascinating: the rules are simple, yet the possibilities and complexity that arises just immense.

What strikes me most, is that it's never clear who is winning until you get within "calculation distance to the end". Well, it's not clear to my patzer eyes anyway. Perhaps it's obvious for advanced players much before, of course.

PS: I just installed a basic Go app on my phone, and managed to beat the computer in level 1 on the 9x9 board, on my third game (lost the first two games, mostly because I hadn't understood the rules). That's probably the Go equivalent of -500 elo or so, but it feels good to beat the machine (unlike in chess)

[Rémi Coulom]

I just learnt to play Go today. It's fascinating: the rules are simple, yet the possibilities and complexity that arises just immense.

Beware it is a hard drug. Stronger than chess. IIRC, Chrilly once said that computer chess and computer go are like cocaine and heroin.

Rémi

[duncan]

Johan de Koning, Peter McKenzie, and Chrilly Donninger also tried Go.

Go is really a lot more exciting than chess. Monte Carlo was a major breakthrough, but it feels like we still need to find a few more big ideas. Currently, the top programs have almost stopped improving for 1-2 years. It is difficult to predict, but I feel that reaching pro level may take a while.

Rémi

I once read that a european (dutch?) im computer chess programmer said he was got some new ideas that should bring go to the top, but does not have the time to implement it. I forgot his name.

[Laskos]

Johan de Koning, Peter McKenzie, and Chrilly Donninger also tried Go.

Go is really a lot more exciting than chess. Monte Carlo was a major breakthrough, but it feels like we still need to find a few more big ideas. Currently, the top programs have almost stopped improving for 1-2 years. It is difficult to predict, but I feel that reaching pro level may take a while.

Rémi

I just learnt to play Go today. It's fascinating: the rules are simple, yet the possibilities and complexity that arises just immense.

What strikes me most, is that it's never clear who is winning until you get within "calculation distance to the end". Well, it's not clear to my patzer eyes anyway. Perhaps it's obvious for advanced players much before, of course.

PS: I just installed a basic Go app on my phone, and managed to beat the computer in level 1 on the 9x9 board, on my third game (lost the first two games, mostly because I hadn't understood the rules). That's probably the Go equivalent of -500 elo or so, but it feels good to beat the machine (unlike in chess)

There are many books on Go (19x19) theory. It's probably richer in theory than chess. Pros are talking an unintelligible to me language. I played Go for several years, to maybe some 5-7 kyu level. To me, Crazy Stone and Zenith are monsters already. Go seems even worse to master than Chess, if one is not educated with it from childhood. On KGS I often see 7-9 dan level games of very strong amateurs, and the games are completely obscure to me. They are worse in understanding than GM games in Chess. Simply unintelligible for me.

[duncan]

Johan de Koning, Peter McKenzie, and Chrilly Donninger also tried Go.

Go is really a lot more exciting than chess. Monte Carlo was a major breakthrough, but it feels like we still need to find a few more big ideas. Currently, the top programs have almost stopped improving for 1-2 years. It is difficult to predict, but I feel that reaching pro level may take a while.

Rémi

I once read that a european (dutch?) im computer chess programmer said he was got some new ideas that should bring go to the top, but does not have the time to implement it. I forgot his name.

I remember it was Vincent Diepeveen.

[lucasart]

I just learnt to play Go today. It's fascinating: the rules are simple, yet the possibilities and complexity that arises just immense.

Beware it is a hard drug. Stronger than chess. IIRC, Chrilly once said that computer chess and computer go are like cocaine and heroin.

Rémi

I know. That's why I will not attempt to write a Go engine. Writing a strong chess engine is hard enough, and I've done that already. But I will *play* Go every now and then, perhaps.

[CRoberson]

I learned Go in grad school. Learned it, because I heard it was harder for computers than Chess. I wrote a Go program 2 years before I started NoonianChess. I wanted to do Chess first, but there was a guy at work that was a 4th dan Go player. I had it working under the Cgoban GUI and the goal was to be better than Godummy. I achieved that goal and switched to Chess. The Go master took a job elsewhere. Initially there is a lot in the way of data structures and initial code for stone groupings...

It was fun and it was able to beat Godummy easily without any search. I believe the real research in Go needs to be in the eval. I said the same thing about Chess in the late 1980s and 90s and 00s...

Micro and Macro patterns are a big deal in Go and depth of search is too limited thus eval and strategy rule.

[pocopito]

An interesting match between CrazyStone and Franz-Josef Dickhut (6d) is going on these days.
Here's the info of the match and two short interviews with Franz-Josef and Remi Coulom:
https://blog.codecentric.de/en/2014/10/ ... mi-coulom/