Life In 19x19

This has become known as the Black Hole fuseki - White stones in various places, of course.

It was introduced in pro go by Ohashi Hirofumi in 2012, and has appeared quite a few times since then - even Cho U has used it. But it has apparently been an almost exclusively Japanese phenomenon. When it first came out, it was just referred to as New Fuseki, but over time the name Black Hole was attached. I'm not sure how. It may have come from amateurs and/or the West, but it is now being called Black Hole fuseki in Japan.

One of the main adherents has been So Yokoku. He played it last week (22 August 2020) in the Yugen no Ma championship which has been set up in the Nihon Ki-in to provide games for pros now that real games are scarce and study groups have been banned during the CV rampage.

Except that he didn't really play it. He played the first two moves and then had to diverge, but it was still called a Black Hole fuseki in the reporting. But can you have half a hole?

Actually, that's not my main question. The reporting also said this opening has become established in the prime of the AI era. I am aware that there's a program somewhere set up to run with this opening and it's apparently having some success, but the Japanese report seems to imply it's got rather wider popularity than that across several AIs - at least the ones they are able to see. Is that true? My Leela is certainly very sniffy about it, but maybe Leela is now antediluvian? (So lost BTW)

While on the topic of AI, I was a little startled to see a comment by Shibano on Game 1 of the 45th Meijin. He got a very poor opening, and the reason, he said, was that he had not studied the joseki in the upper left properly. This is one of the bog standard Direct 3-3 invasions (where he was the invader). Looks like his method of studying AI (i.e. not using a bot himself, but just observing what other people play on the internet) let him down. He almost won that Game 1, nevertheless, helped by an Iyama mistake. A calamity ko erupted, which led to a big trade. Iyama was in a "cold sweat" and at one point spent 65 minutes on his move. But he survived by the slimmest of margins. Shibano gets his favourite White stones next time round.

John Fairbairn wrote: Actually, that's not my main question. The reporting also said this opening has become established in the prime of the AI era. I am aware that there's a program somewhere set up to run with this opening and it's apparently having some success, but the Japanese report seems to imply it's got rather wider popularity than that across several AIs - at least the ones they are able to see. Is that true? My Leela is certainly very sniffy about it, but maybe Leela is now antediluvian? (So lost BTW)

There's a bot at or near the top of the computer go server rankings that plays black hole fuseki. But I have no reason to believe it has naturally discovered this opening and thought it is similar to or better than normal corner openings. Instead I strongly suspect it is simply hardcoded to play these initial moves for the spectacle, and then the bot just carries on playing with the disadvantageous position as best it can. But because it likely has greater resources than the other bots on the server it manages to win despite this handicap. We've seen here that KataGo with 200k playouts can beat KataGo with 5k playouts and 3 handicaps, and I doubt the black hole loses as much as a stone (don't have access to KataGo atm to check). So all that's needed to troll the go world with a black hole bot is the money to rent some beefy GPUs. A lot of CGS bots use less than 1000 playouts.

Fatkata