AlphaGo Teach — whole-board joseki

I’m not sure about that.

Loading that page didn’t make my memory consumption blip at all in the activity monitor.

There was a burst of CPU going on with “Chrome renderer”, but that’s all.

2 Likes

It’s not the memory. Firefox and Chromium both are using about 10% RAM when I open the demo board, but only Chromium has the rendering issues (on Desktop).

4 Likes

It is not an insufficient memory issue on my machines.

I see similar graphical glitches on two different machines. Both of them are running Linux and Firefox.

These two machines have 16 GB and 32 GB of RAM, only a small fraction (less than a quarter) of which was utilized (across all processes) during which these glitches occurred. The problem seems to be persistent, and hence easily reproducible.

Let me know if you want me to do additional troubleshooting, if it might be of interest to the developers.

1 Like

I added some optimizations that should help with loading and viewing this.

There is a separate issue with graphical glitches, I think Chrome is breaking things with their canvas again.

4 Likes

does it feel like black especially has to try some incredibly adventurous stuff in order to compensate for komi?

The current setting of komi is probably too big at 6.5; it should probably be 6.0. With integer komi, some small percentage of games would end as draws, which some Go players don’t like.
A draw is a perfectly sensible result if both players play equally well. Other games (chess, for example) have a much higher percentage of drawn games than Go would ever have, and still managed to run decent tournaments for the whole of the 20th century. For example in the famous Linares 1998 chess tournament, over half of the 84 games played were drawn. In Go, with integer komi, probably less than 5% of pro games would be drawn. Between amateurs the percentage would be even less. How many of your games were decided by half a point?

3 Likes

I plugged AlphaGo Teach’s favourite starting position into KataGo (courtesy of ZBaduk) and continued the game to the end. Might do the same for other AGT positions in the future.

4 Likes

Another AGT position continued to the end with KataGo from ZBaduk. It was even throughout the game (as is par for the course) until black crashed and burned in the endgame:

2 Likes

I completely re-played both of the above games with better playouts (same links).

And here’s an (IMO) spectacular third game with huge territories on both sides. At one point, one player sacrifices 20 stones, apparently with the purpose of living with 6 points on the other side. The game stays balanced after the “trade”. https://online-go.com/review/434009

Another game with lots of sacrificed stones and huge territories on both sides (but it’s a half-pointer, of course): https://online-go.com/review/434346

More games (let me know if I should stop spamming these… or alternatively post them in a “Kata Go self-play games” thread?):

Game 5
Game 6 (continued from a High Chinese position)
Game 7 (in which white not only makes a wall, but successfully uses it to build a moyo)
Game 8 (the first game I’ve continued from a position with an early 3-3 at move 5 - amazingly, the projected winrate remains almost perfectly even until move 330)

Game 9 started from a super complicated fuseki and stayed complicated throughout.

Game 10, in which black actually creates a sizeable center territory rather than it all turning into dame.

Life in 19x19 user xela (Australian 2d) added LZ-258 variations and comments for game 1. Super fascinating IMO!

xela has added a few variations and comments for game 2, as well: https://online-go.com/review/432118

KataGo continuing from an AlphaGo Teach position, game 11: Black not only goes for vast amounts of influence, but actually uses it to build a giant center territory. White takes three and a half corners and wins by 2.5 points or so.

1 Like

KataGo continuing from an AlphaGo Teach position, game 12: In which white almost makes a B2 bomber shape in the fuseki. Black refuses to be shown up and builds a proper B2 bomber later. The game is decided in a ko that is fought over 76 moves.

Cool winrate graph, too:

2 Likes

My next project is to explore AlphaGo teach positions in which white makes a mistake on move 2 in order to get the winrate closer to 50 % (continuing with KataGo at the end of AGT’s sequence). Game 13:

1 Like

My current project, which started with game 13, is to explore AlphaGo teach positions in which white makes a mistake on move 2 in order to get the winrate closer to 50 % (continuing with KataGo at the end of AGT’s sequence).

White won game 14 by 0.5 points:

In game 15, black slays white’s dragon and wins big. Noteworthily, Leela Zero gives white 64 % winrate as late as move 275:

In game 16, it is black’s dragon that dies:

1 Like

White wins game 17 by 26 points or so:

Likewise, white is ahead by about 26 points again by the end of game 18:

Black wins game 19 by 1.5 points or so. Leela Zero gives white 62 % winrate as late as move 308:

1 Like

White wins game 20 by 2.5 points or so:

1 Like