Thanks, I got that reversed between and Chinese and Japanese scoring.

this is actually not a seki, because "strange things happen at the 1-2 point":

Click Here To Show Diagram Code

[go]$$B
$$ -------------
$$ | . 2 X 4 O O |
$$ | 3 6 X O O O |
$$ | X X X O O O |
$$ | X X O O O O |
$$ | X X O 5 . O |
$$ | X X O 1 . O |
$$ -------------[/go]

(black has to play 3, otherwise white plays it himself)

_________________
Spilling gasoline feels good.

I might be wrong, but probably not.

I would gladly provide you with a ranking system, BUT.. never programmed ruby not used GIT. Long time since I did any open source stuff. Ill at very least take a look at GIT however. In the meantime maybe I could draw your attention to this post..

viewtopic.php?f=10&t=4743

Based on the analysis a ranking system could be designed, where players can choose whatever komi and handi they want to, and the system can accurately estimate the probability of each player winning. Based on this probability you can then adjust their ranking. Also based on these probabilities you can deduce how accurate a players rank is by calculating mean square error for his recent games. This number could then be used as a multiplier when changing his rank, and conversely its reverse should be used when adjusting opponents rank based on the game result. And the same in pseudo code.


rank_weight=( 1 - mean_square_error( last_few_games ) ) ** cool_constant

if( won_game ) {
rank+= ( 1 - probability_of_winning ) * ( 1 - rank_weight ) * opponents_rank_weight
} else {
rank-= probability_of_winning * ( 1 - rank_weight ) * opponents_rank_weight
}


That should produce a ranking system where rank changes fast when its uncertain and becomes accurate when players result look consistent.

That's right, but how should we score it in case both players pass? It's certainly possible, I've just overlooked it.

Black's group would still be considered dead, because White doesn't need to make a move to ensure Black's death.

Black only dies if he plays... so if he passes its seki.... Play isnt mandatory here right?

_________________
I'm thinking...

It's a semeai; if Black doesn't play first, then he'll only be more liberties behind if he passes. Either way, Black will die.

You guys are missing the point. The correct answer to the question is 0 points, in all cases.

It doesnt matter what could've been, it matters what the players agree on. If they agree its seki, its 0 points under japanese rules, and the territory is counted as points in chinese rules (+all the stones).




I scanned over the post, and it would be great to be able to compare what you did to our current win-by-walk-over champion, Yoyoma's Glicko.

Specially, if you have your dataset in a comfortable format, we would really appreciate it. I got my hands on some chess results, but Go ones should be better as we can calculate with handi/komi.

It is not so difficult to use git for this purpose: you have to Fork the project( you can do that online), work on it, and then make a pull request. I will receive that and see your changes: after i review them, i merge them to the official branch (master).

All that you should be able to do online. Then, from the command line you have to do :

git clone git@github.com:conanbatt/OpenKaya.git
#work on your stuff
git add files_you_changed_or_created
git commit -m "i just made something really cool"
git push

There are several GUI's if you are a windows guy, but i cant recommend you one by experience.


Regarding Ruby, for such mathematically intensive code, you shouldn't have that many issues adapting yourself to it. You can see under rating/strategies/glicko to see how yoyoma implemented his ,its fairly straight forward.

If you need pointers or help, don't hesitate to ask. I'm available to help code-contributors. You can join #kaya.gs channel on FreeNode.

_________________
Founder of Kaya.gs

Here is a brief description from wms on how KGS does seki detection:
http://groups.google.com/group/rec.game ... 8f79?pli=1

Another related link:
http://senseis.xmp.net/?WmssDameSekiChallenge

From the rgg link:


If you put the above position into KGS it declares it to be a seki since the square-four shape cannot make 2 eyes. I believe the only eye shapes it will declare to be seki are 1-space, 2-space, and the square-four. All others such as 3-space can make an eye with sente, therefore the assumption is that since the player did not choose to do so it indicates that even if the opponent plays in, there must be a way to win a semeai.

excellent contribution.I will add the links on the repo for score.

_________________
Founder of Kaya.gs

OpenKaya is growing quite a bit as we select more libraries and functionalities to be Open-sourced.

We have added a very permissive License, CreativeCommons allowing Commercial use.

Today we uploaded the Timer code. After fighting a lot with the required testing, Polly was able to run an excellent test suite to check the clock is performing as it should.

The code is designed to new time systems can be easily added, and there is model code to see how to automatically test for it.

So many new systems are now ready to be developed, either by us or by eager contributors. Check it out at https://github.com/conanbatt/OpenKaya/

Code contributors will be added to the board of fame and recognized for their efforts!

_________________
Founder of Kaya.gs

Hi

Did you drop the idea of using the WHR algorithm ?
I'd like to know if you tried it, and if you compared it to Glicko. How does it perform ?

Good luck for your project