I think this page is actually specific to some limited cases. The best known is probably the Chinese fuseki with a star point extension. More generally we do not invade based on which is the wider side but rather where we have support. Consider the "extreme example" shown on the page. That position does not occur in GoGoD at all. However, if we tighten up the sides we find as shown below that White is far more likely to invade the narrow side than the wider side. Why is that? Examining the individual games, it is because more often this framework occurs with a White stone just below the narrow side than just to the right of the broader side. In other words White decides based on the White stones on the board rather than the Black ones. The more pertinent advice seems to be, "Approach from support". What do you think?
Search results
Dave, you're absolutely right. And we should back up statements more by database search than suspect conventional wisdom.
(later)
A gobase search yields, for (c,d,a,b):
This means that there is an overwhelming pro game support for approaching at the smaller side, regardless of support, for this particular configuration. BUT, there is a catch, as usual with database search ...
The figures also suggest that, in case of a White stone at the top, Black rarely leaves this situation alone. Indeed, already 7 games are found where Black plays a one space jump to defend his moyo. So, maybe Black should not omit a move there. On the other hand, a White stone at the left is often unanswered and then White will obviously move at b.
So, the most relevant result is the one without support. That one clearly shows that pros en masse choose to approach at the small side, or rather invade it.
Dieter,
We seem to be seeing very different search results. Using the current summer 2011 GoGoD CD, I get 289 hits...
Search #1
If I search with a stone below on C10, I get 177 hits. Note that this is what I expected and the reason that I checked in the first place. The formation with the 3-space extension from the star point is mainly used to attack a stone on C10. (A supplemental search showed that of those 177, 19 also had a stone at M17 on the right side.)
Dieter: gobase puts a maximum of 200 on the results, so this one is 200+
Search #2
Dieter: 0.
So that leaves a further 41 with some miscellaneous stones on the outer edges. However, if I read your post correctly, your results from GoBase are very different. In particular, you have most of your hits on my search #4 while I get most of mine from search #2. Could you recheck that your results are as you described? Thanks!
Dieter: I parsed it into your message. Gobase seems to have quit updating since 2009. GoGod is definitely a fresher db, but I don't have it (yet).
I parsed my comments into yours. We should probably have made this a discussion subpage i/o a forum.
I now understand my way of searching is flawed. The big results I get for an "unsupported" invasion in the small extension obviously come from this one:
I'm a bit embarassed. This is a good lesson.
(later and less embarassed) It seems that the "sente" feature of the gobase search is defunct. My acquisition of GoGod + Kombilo is imminent.
That's fabulous. It would be nice if there were a search feature along the lines of "this was the last stone to be played" in addition to just who's sente it is.