As an absolute beginner, a big thank you to both of you for your comments. Yes, I understand that the visualization will NOT be perfect, as there are many complications that can arise. Highlighting the correct areas of influence can ONLY be done manually.

Still, it is at least interesting to see this video to have some 'rudimentary' grasp of the concepts of territory, inter-relationship between each stone and how it strengthens one another to form an area of influence. Also, to see how the opponents' stone diminishes it.

Perhaps, one day, we could have just a EXHIBITION game where these areas of influence are shaded properly with comments, so as to get a beginner interested.

Robert, I understand your fear that it will give PRECONCEIVED and MISTAKEN notions to anyone watching, but as a beginner, at least it makes me more interested in the game, as per my above comments. Anyone interested would probably be able to rectify these misconceptions after they start playing the game more seriously, methinks?

Hi agewisdom, this varies enormously from person to person.
My anecdotal evidence shows the range from a short time (a few years) to never.

Haha... now I understand why you're so concerned about trying to prevent any misconceptions from the start.

Hopefully, I don't fall into the NEVER category.

As a kyu player and maybe even as a low dan player, very much effort (30% ~ 70%) has to be spent on unlearning, correcting or completing wrong, partially wrong, inconsistent, confusing or very incomplete knowledge. The amount of interesting useful, good, correct knowledge is already so great that there is simply no point in wasting a similar amount of effort on bad knowledge. It is better spent on learning good knowledge immediately.

I do understand what you mean. Thank you for the sound advice. Appreciate it

I think the video looks neat (and maybe that's useful in itself for generating more interest in the game), however, it seems to have significant shortcomings in actually visualizing influence.

I wonder how the visualization would look if you first set down four stones in a ponnuki shape, and then filled in its eye with a fifth stone. Any reasonable analysis would have to say that the influence of that group has gone down (since it has only become weaker in losing its eye), but I think the simple linear technique used in this video might actually suggest that the influence has increased.

The description on the youtube page states that it is based on "a simple linear driven metric exponential distribution influence function".

I'm not too sure exactly what that means, but if I had to guess, I think the keywords "simple", "linear", and "exponential distribution" suggest that the status of each point is simply computed as the sum of the exponentially scaled distances to each black stone minus the sum of the exponentially scaled distances to each white stone (so essentially a proximity-based method). Then, these values are normalized and colored, with the larger values being more orange and smaller values being more blue.

The way the colors change in this video seems to be in line with the above hypothesis. Note how each stone always seems to have at least a small effect every other point (even those on the other side of the board with many intervening groups in between).

I suspect that the video ends short since such a simple linear method would become absurd later in the game when closed borders and life/death come more into play.

Does anyone want to collaborate to do one professional game using the visualization principles? I'm an absolute beginner, but I prefer to have some visualization along these lines to get an idea.

The game should be reasonably simple but a professional one with a variety of techniques/principles/problems(???) involved, so that it can be a good example to be used to assist. It's quite a bit of work involved, so let's make it a good one.

I don't know what I'm getting myself into, but I will do my best to assist to finish one game. Here's a rough sample of what I'm talking about. Hope it makes sense. This is just MEANT to help beginners visualize how the stones affect the whole board and how the territories are affected by the placement of each stones. It's not meant to do anything else. Just to ease beginners into the game like the earlier Youtube video.

Everyone can chip it and instruct me, and I'll make the images accordingly, if they feel it's useful. If it's a bad idea, do forgive a newbie, ok!


upload pics

It is an idea of 1970s computer programs. Needless to say, they were very weak.

In game 5 of the alpha-Lee match, Redmond at one point mentioned a "black shadow" extending into white's moyo. Redmond pointed out that this black shadow only existed because the few black stones creating it were unequivocally alive (in that case, because they already had one eye and could make another).

Pardon me, if I'm mistaken but during the AlphaGo vs Lee Sedol matches, I heard remarks about AlphaGo having a heat map over the entire Go board, to visualise the entire situation.

Anyone have any idea whether this was correct? It would be interesting to see what this visual representation looked like.

After watching most of Redmond's commentary, I think automating a heat map would be an impossible task. Even a manual heat map would be only based on the highest probability assumption on what would occur, which may or may not be correct. I wonder whether stronger Go players actually do visualize the territorial influences of just intuitively 'know' it based on the patterns of the stones...

Take a look at Bruno Bouzy's work. He has some paper's which will give good starting points.
The initial step of projecting influence is relatively easy, but then you have to factor in life and death status. As far as I know that it is a harder problem.

Sorry to give a useful answer.

_________________
North Lecale

Thanks. Would appreciate if you post the link here, if possible.

If you can't find it from google let me know

_________________
North Lecale

The "heat map" was not of influence but of the probability of the next move being played at that location. So in a given position it might think that Black was 43% likely to play move #1, 19% likely to play move #2, etc. This was its "policy network" that was a neural network trained to imitate human games.

Oh, I see. Thanks for clearing that up.

Did they make available the heat map for any of the games played? Either against Lee Sedol or Fan Hui?

There are some examples of various heat maps in a specific position in one of the informal games against Fan Hui in Figure 5 of their paper (PDF link).

Many thanks!