Svante Carl von Erichsen on Go WCH 4


Back to forum

Thibault de Vassal    (2010-09-17)
Svante Carl von Erichsen on Go WCH #4

As you probably read in the news, Svante Carl von Erichsen won the 4th FICGS Go WCH, beating his challenger Huayong Yang 3-2, Svante Carl wins the Go championship for the 4th time in a row!

Svante Carl kindly accepted to answer a few questions on his match & computer Go:

FICGS - Hello Svante Carl, congratulations once again for winning this match against a surprising challenger who started here a few months ago with a 10 kyu rank, Huayong Yang, now rated 2438 after scoring 2 points in your 5 games match (which is a great achievement for sure). What did you think about his play & yours in these games?

Svante Carl - I think that he greatly underestimated his rank initially. As far as I know, he had not played for a long time and believed that his ability had therefore deteriorated. I do not think that you can drop more than one or at most two stones, though -- it is like cycling or swimming, you never unlearn it. I had the impression that we were quite evenly matched in summa, but our strengths are in different aspects of the game; I cannot really put my finger on the difference, though.

FICGS - After a previous win, you said that you spend a quite long time to analyze, which probably helps you to reach a higher level than 2 dan (your EGF rating) compared to OTB play... It looks obvious to me that correspondence chess moves generally ask for much more time than Go moves at a high level but I may be wrong, how much time did you spend on your longest analysis during the match? Do you remember for which move?

Svante Carl - I usually spend at least a few minutes on each move, except when the continuation is obvious. I often use more, and if I do not find a satisfactory move then, I will even postpone the move to another day, so that I can sleep over it and let my subconcious work on it.

FICGS - Do you watch other games played by your future opponent before starting your match? Do you think that this is really important in preparation like it can be in Correspondence chess?

Svante Carl - I sometimes glance over the games in the championship qualification tournament, but I do not try to prepare this way. I do not think that such preparation has any value in Go, especially in correspondence Go, since you have time during the game to do deep analysis. I usually try to take each game out of standard fuseki patterns pretty quickly, anyway. Of course, I know that my opponents in these title matches are always very tough and demand my utmost respect.

FICGS - Do you still follow the recent developments in computer Go? What do you think about the latest Go engines? How much time do we have yet before the best Go players are caught by computers according to you?

Svante Carl - I have the impression that the currently most promising technology (Monte Carlo/UCT) has the potential to achieve a rank of about 2 or 3 dan (EGF/KGS). I think that the next fundamentally new idea or breakthrough might add 2 stones, to get to 4 or 5 dan. I do not have any idea where it might go from that, but I think that it gets always harder.

What I would find interesting is having more intermediate board sizes. The best bots are almost on par with the best professionals on 9x9 now. I would propose to try to achieve a similar level on 11x11, then 13x13, then 15x15 etc.. Regarding 9x9, I think that the currently predominant komi of 7.5 points is too big, and that this has a negative impact on the experiments because the bots do not play in a balanced environment. It might be worthwhile to introduce the Taiwan rule (last move compensation) to get more fine-grained scores.

FICGS - What programs did you use this year to analyze? (just trying, of course it may be part of your secrets ;))

Svante Carl - It is not a secret. I just use an editor, usually EidoGo or CGoban3, to visualize the variations I imagine.

FICGS - Finally, what thoughts would you like to share on your 5 games, that could help us not to miss the best times or to help us to understand the most complex moves...

Svante Carl - I cannot give a detailed commentary, but I can try to summarize my impressions.

I think that Game 5 was quite balanced until move 21, but I think that the white invasion was a bit ambitious then. Of course, White did not need to die there, but after moves 32-33 I think that Black had a good result anyway (move 32 should go out faster in my opinion; note how E14 helps Black in enclosing White).

In Game 3, I think things got quite difficult for White in the lower left, but I let him take the initiative by backing off at move 35 (I should have simply closed off F10 then). White gained control of the centre as a result, and in the large endgame, I lost too many points there.

In Game 4, I fell behind in the opening through some slow moves (there was some discussion on the Life-in-19x19 forum about this, see the link in the comments of that game). In the endgame, Black then lost some points in the centre, so that I was a bit ahead when the game timed out.

In Game 1, I made some bad decisions on the left side, and never managed to turn things around. I think I was behind by about 5 points in the end.

In Game 2, I think that Black should not have ignored move 24. After I got quite some territory from my moyo and also reduced his top side, I could play it safe.

I look forward to the games with Olivier Drouot that recently started, but I also hope that Yang Huayong will re-enter the championship cycle.