In game 72 (7 komi, NZ rules) the entire sequence from move 46 - 82 felt like a giant “joseki on the side” (most of it playing out just the way Kata read it out at the beginning) with black getting a 36-point corner, whereas white makes a 12-point territory with thickness on the outside.
Non-19-x-19 square boards are finally accessible on https://www.baduk.com/next/ .I’ve played a 9 x 9 game in one go (about three hours) and am happy enough for this to be my only 9 x 9 high playout game (100,000 - 1,400,000 per move). From move 0 - 38, there wasn’t a single move at which KataGo’s second-favourite choice was close enough to its favourite move to qualify for my branching criteria. At move 39, we’re already in the endgame with everything settled.
Variation from move 0 (around 200,000 playouts on just F6)
I’ve played three 13 x 13 games (all in the same file). The second game turned out a bit more complicated and exciting than the first one (30 moves longer, too). For a while it looked like a white win was in the cards. Move 53 was the highlight for this 10 kyu 
In the end, all three games ended as draws. Shoutout to game 3, move 66 as well.
In 19 x 19 game 73 (NZ rules, 7 komi), 10p AI shows that the best way to play go is to play tengen and build three B2 bombers.
@terrific, OGS 5d has done a three-part YouTube review of this game!
I’ve included all of his (commented) variations, plus more KataGo variations, in the OGS review as well.
2 Likes
In our 11 x 11 game, KataGo plays four 4-4 points:
1 Like
One interesting thing is that on 11x11, KataGo very clearly prefers 4-4 over 4-3, but on 10x10, very clearly prefers 4-3 over 4-4.
4 Likes
Very interesting indeed! I’ll have to play a 10 x 10 game eventually 
I asked high-playout KataGo to play two games from the Tibetan starting position on a 17 x 17 board. I still used NZ rules and 7 komi.
White won the first game, but the second (bottom variation in the same file) was a breathtaking draw. Here is my 10 kyu interpretation:
Moves 13 - 48 are an exciting early middle game.
Moves 49 - 69 black builds a ko factory.
Moves 70 - 90 white builds a ko factory of its own.
Moves 145 - 247 is the ko fight.
And Clydevil (~2d)'s interpretation:
49-69 for black is just a way to invade the white framework. You just have to see the whole maneuver as invading the whole top left. it’s common when invading framework to play at several places and key points at the same time and deciding at some point what to sacrifice and what to make live. Here black discards the left and stabilizes the top part.
70-73 can be seen as probing exchanges, white makes black decide how he will kill it, thus deciding forcing moves in the vicinity.
74-75 can be seen as a first use of this white aji built in sente, white plays at the top 74, removing pbs in its shape and preparing an attack on the top group, black needs to answer to avoid things around K6 or H6 that could be large white reductions. So I guess here that white is just happy with 70-75 to play 74 in some kind of sente.
76-87 tries to recycle the white death into something useful, and it seems that what is mainly accomplished here is white huge yose at M2 followed by N2 (which is hard to see because in the game white changes it’s mind and do not play N2, but locally white N2 instead of the game move 92 would be logical).
2 Likes
I also asked high-playout KataGo to play two games from the sunjang starting position (still with NZ rules and 7 komi). Black won the first game, the second (bottom variation in the same file) was a draw. Kata really likes the 6-2 point in these games. She played it once in the first game, twice in the second.
1 Like
In game 75 (NZ rules, 7 komi), white doesn’t play northwest of the tenth line until move 100:
Terrific sent me here to this forum, says you’re interested in dan players doing video reviews. Got anything up for grabs right now? After looking through the games definitely a study topic I’m interested in.
3 Likes
Totally! :3 Feel free to browse the fuseki index and I’ll link you whichever game you think looks interesting: KataGo self-play fuseki
1 Like
In game 76 (NZ rules, 7 komi), by move 180, black is ahead by 4 points. Our current tally is 17 black wins, 27 draws and 32 white wins.
Game 77 (7 komi, NZ rules) sees some of the most bizarre endgame I’ve ever seen. White seems to play meaningless moves in black’s territory from 154 - 175, then from 195 - 222 black seems to be returning the favour. After move 213, black is at least 2 points ahead. At move 248, white starts giving away points and ends up losing by 12.
2 Likes
You piqued my curiosity enough with the description that I’ve uploaded a version with the usual Full OGS analysis of moves to look at:
I hope someone smarter than me will answer this:
1 Like
In game 78 (7 komi, NZ rules), black tries to build a center moyo. After move 43, white seems firmly ahead by 1 point and maintains a small lead throughout the game. There is a cool endgame trade in which white sacrifices its right-side group for a large bottom side.
@TangJieHao, CHN 5d did a one-hour video review of this game! https://drive.google.com/file/d/1oAA4hh … sp=sharing
I’ve included all of his (commented) variations, plus more KataGo variations, in the OGS review as well.
2 Likes
Black is forced to play tengen in Sunjang and its customary. Its essentially a white plays first board position.
Im not sure if i replied properly but im talking about Sunjang Baduk positions.
Yes, I am aware that that is customary. However, Kata regards it as a mistake. I simply borrowed the opening position (which gives an even winrate prior to tengen) and let Kata continue with NZ rules and 7 komi.
1 Like