The name “Go” can be a bit frustrating for us English. It happens to be the same word we use for a very common english verb. However, you have to admit that the simplicity of the word “Go” really goes (QED) with the game. This is no “Terra Mystica”, or “Mahjongg”. This is just “go”, and like humble old man, it conceals an ocean of ideas to marvel at.
So, this afternoon, I found myself playing this game with a friend. We played on a 13×13 board with a 2 stone handicap (which I felt was just about right). He’s been improving. I slipped up and lost a 10 point group right at the end (of course I was being greedy, as usual). I was a little frustrated of course, but he had beat me fair and square.
Anyway, on the way home, the idea hit me. I wonder if I could remember (or deduce from partial memory) the exact sequence of moves made in the game, start to finish. I had made no attempt to memorise the game while playing it, so it should pose a good challenge. I had done a similar thing before on a 9×9 board, but that was only about 13 moves (a half finished game), but I was pretty sure that this time, the number of moves it was more like a hundred…
So, I got home, opened a demo board on the computer, and started reviewing the game. At first it was easy. I remembered about 70 moves, and then stopped. Something was wrong. I went back to the seventh move, trying out variation after variation. I slowly made progress, remembering sequences and then stopping on pivotal moves…
The game tree became very complex. I had remembered remarks we had made about the game while playing it “that move is boring, I’m going to jump” or “that move looks like a slap”. I knew the end score was 12 points apart, 16 to 28 maybe (I wasn’t paying much attention at the time, I was rather frustrated!), yet I felt it was an odd number, since I remember making a slight mental leap trying to count the difference in scores. Memory is a funny thing…
Eventually I had got quite far in the game tree. There were definite problems here or there, but I had about 60 moves (in about 3 sequences) that I was pretty sure of. I remembered shapes, and these shapes helped me realise problems. Therefore, I decided to try and figure out the final game state, and work back from there:
So I got out my 13×13 go board, and plunked some stones down, correcting shapes as I went. This proved to be a lot easier, and I got quite far in it. I found inaccuracies by scoring the game, and comparing the final score. I remembered white passing 2 stones and black 1 at the end, so I knew that black played an additional stone. I went back to the tree with this information, and perfected that. Both approaches played off each other quite well.
So what did I get, in the end? Well, according to my latest board state, black has played 62 moves, and white had played 61 (123 moves). Black’s final score (space minus prisoners) is 29 and white’s is 20. As you can see, white is at least three short. Here is the final game state:
As you can see, I didn’t quite manage it. However, I am pretty sure my final board state is at least 119/123 = ~97% accurate, which isn’t quite there. I did a lot of transposition in the tree, though it mostly makes sense (though I seem to have sente more than I should!). Despite the failure, I enjoyed doing this. It was fun, and I think I’ll get better at it the more I play.
Really, I think the fact that this is possible is due to the nature of the game:
– Go is a very visual game (there are lots of patterns).
– Before every move, we see the game state, and reflect on it.
– Pieces placed earlier generally stay on the board, so this memory is reinforced.
I think these three factors (along with experience) make Go games rather memorable.
To be honest, I’ve given up evangelising this game. in the end, it is just a game. Nevertheless, it continues to surprise me. As I continue to explore it, more and more concepts begin to emerge. This little experiment is just one example of why I’m fascinated by the “ultimate game” of Go.