Life In 19x19

msgreg wrote:Haven't reviewed the accuracy of the scoring, but I'm getting and displaying results.

Nice! I didn't check the final number it gives, either-- it should be implementing AGA (area) rules.

Of course, you can always just use it to mark the dead stones and do the addition manually.

Since I'm between jobs for the week, I added some features to this, including the ability to take in SGF and set the number of playouts it uses (see the "iterations" parameter). It remains backwards compatible. Enjoy.

http://daily-joseki.appspot.com/score

Interesting. Do you have a recommended range for the iterations? (perhaps based on the board size and or number of unfilled stones, or other aspect)? And what are the minimum required SGF tags?
(i.e. are both B and AB recognized?) How about KM, and are results returned in JSON or SGF?

And especially, thank you for the backward compatibility!

It will pay attention to SZ, KM, B, W, AB, AW, and AE. SZ is required, KM is required if you want it to take komi into consideration. It plays through the first branch at every node and ignores everything else.

Results are exactly the same as before (json), no change there!

If the game is already finished, I'd leave the iterations parameter low, and you should still get consistent answers. Otherwise, set it high and don't expect quick or consistent results! I recommend playing around and seeing what the best balance of speed/accuracy is for your application.

Also, I just made the demo page much prettier (maybe you already saw the pretty version, depending on when you looked).

daniel_the_smith wrote:
Give it a shot:

http://daily-joseki.appspot.com/score

Hi Daniel.

I was sort of disappointed that your tool seems to give random scores for the same position. Are you still developing it?

cyclops wrote:

daniel_the_smith wrote:
Give it a shot:

http://daily-joseki.appspot.com/score

Hi Daniel.

I was sort of disappointed that your tool seems to give random scores for the same position. Are you still developing it?

It is a bit of a work in progress.

What position were you putting in?

Unfortunately, I don't have much time to work on this lately.

I just used the input of your example. But now I discovered that the grid input gives quite a stable result if I increase the repetition parameter to 250. The sgf input seems to be unstable even with 250 repititions though less unstable than with less repetitions.

Hm. I don't remember it being that bad. Maybe I broke something last time I changed it? Anyway, this is definitely a case where you're getting what you're paying for.

I really want to work more on this, but I just haven't had time and it doesn't really look like that's going to change much soon.