I’ve seen this beautiful video years ago, and I think I shared it then, too, but that was long before the new OGS It definitely gave my understanding of influence a boost. Wish the video were longer … but then I guess I will just have to develop my own “X-Ray Vision” for my play
Quote from the YT page:
This video is a demonstration of a piece of code that I wrote to visualize the concept of influence in the game of GO. It’s a simple linear driven metric exponential distribution influence function. This game was played on 1933-10-10 in the Oteai, between Kosugi Tei (Black) and Go Seigen. White won by resignation.
[quote=“Mourner, post:5, topic:3224”]
[…] leaves a lot to be desired
[/quote]LOL, what in the world would NOT leave a lot to be desired? As long as beings we are, as long will we desire even more more more.
Influence and score estimation is such a complex thing, it would be like having the strongest AI watching our game, I can’t even imagine the server “brain” power that would be necessary for that. And, of course, to have that available for every running game, not just for yours
While the effect looks somewhat similar, I think that the method in this video might be quite different than the OGS score estimation algorithm.
The OGS score estimator has been described as performing some sort of Monte Carlo method, which involves simulating many random play outs from the current position while using some heuristics to ensure reasonable simulations and to aggregate them into a judgement about which groups survive and which empty points are taken as territory by each player. I think the visualization of the estimate is based on the confidence of the judgment for each point.
As for the method used in the video, I’m not too sure exactly what is meant “simple linear driven metric exponential distribution influence function”, but based on the description and how the video looks, I would guess that something along the lines of the following is done:
Each intersection starts at a value of zero at the beginning of the game.
When a black stone is placed, the value of each intersection is increased by an amount proportional to an exponentially decaying function of the distance between the stone and the intersection.
When a white stone is placed, the value of each intersection is similarly adjusted, except in the negative direction.
Values equal to zero are visualized as black, while larger positive values become more orange and larger negative values become more blue. These values would need to be normalized (or, equivalently, the coloration scale dynamically adjusted) to avoid ruining the visualization.
Animation frames for smoothly changing colors between moves could be generated by interpolating between the before and after for each move.
Such a method could be considered “linear” since the visualization of each intersection is just the sum of the effects from each move. The animation interpolation would also likely linear.
The video looks neat, but I don’t think the method really illustrates influence nor territory properly. Even a group cut-off and forced to live small would diminish the influence of a surrounding wall. Also, this simple coloring method by itself is not able to factor in life/death considerations.
I don’t think the intention of the OGS score estimator is to visualize influence. Instead, I think it is showing the confidence of its territorial and life/death judgments. As a territorial confidence visualization, perhaps one could say that it is the philosophical opposite of an influence visualization.
Nice to see influence visualised, though the colours and the (IMHO too fast) movements make it painful for me to watch.
This is a java program, which uses moving waves of colour to explore the concept of “influence” in the game of Go. To the right is a screenshot of it – static, unfortunately. With the program, the colurs are constantly moving.
If you have java installed on your computer, download it and watch it running. If you have any feedback or suggestions, please report them to its creator, Nick Wilkinson.[/quote]