Life In 19x19
http://www.lifein19x19.com/

Scoring estimate question
http://www.lifein19x19.com/viewtopic.php?f=24&t=604
Page 1 of 2

Author:  tj86430 [ Sun May 23, 2010 7:07 am ]
Post subject:  Scoring estimate question

Any idea why SE shows this situation as seki:

Click Here To Show Diagram Code
[go]$$
$$ . . . . . . . |
$$ . . . O . O . |
$$ . . . . . . O |
$$ . O . O O O X |
$$ . . O X X X X |
$$ . O X X . O . |
$$ . O O X X . . |
$$ --------------[[/go]


I believe black is unconditionally dead?

Author:  Magicwand [ Sun May 23, 2010 7:11 am ]
Post subject:  Re: Scoring estimate question

dead 100%

Author:  tj86430 [ Sun May 23, 2010 7:13 am ]
Post subject:  Re: Scoring estimate question

Magicwand wrote:
dead 100%

So the original question remains, why does SE not recognize it as such? It recognizes IMHO much more complicated situations correctly

Author:  Magicwand [ Sun May 23, 2010 7:15 am ]
Post subject:  Re: Scoring estimate question

what is SE?

Author:  tj86430 [ Sun May 23, 2010 7:20 am ]
Post subject:  Re: Scoring estimate question

Magicwand wrote:
what is SE?

Scoring Estimate (in CGoban) - although I'm not sure if it is called exactly that in English client, but I believe I've seen that abbreviation used somewhere.

Author:  nami [ Sun May 23, 2010 7:25 am ]
Post subject:  Re: Scoring estimate question

SE is awesome!

Features:
can kill 2 eyes group. 12d tesuji
can live with not enough space
can make territory with dead marked stones
can create territory with tengen stone

99,99% correct, 12d tesuji + L&D function.
Thousands of people already know: SE is trustworthy. When will you start to trust SE?
: )

Author:  amnal [ Sun May 23, 2010 7:28 am ]
Post subject:  Re: Scoring estimate question

tj86430 wrote:
Any idea why SE shows this situation as seki:

Click Here To Show Diagram Code
[go]$$
$$ . . . . . . . |
$$ . . . O . O . |
$$ . . . . . . O |
$$ . O . O O O X |
$$ . . O X X X X |
$$ . O X X . O . |
$$ . O O X X . . |
$$ --------------[[/go]


I believe black is unconditionally dead?


The answer is just 'SE isn't very accurate'. It isn't something of bot level complexity, that reads out situations to calculate their life and death status. I don't know what exactly it *does* do, but it will have some simple algorithm that is right a reasonable fraction of the time.

Author:  tj86430 [ Sun May 23, 2010 7:34 am ]
Post subject:  Re: Scoring estimate question

amnal wrote:
The answer is just 'SE isn't very accurate'. It isn't something of bot level complexity, that reads out situations to calculate their life and death status. I don't know what exactly it *does* do, but it will have some simple algorithm that is right a reasonable fraction of the time.

Ok. I know that it isn't accurate, but having seen it estimate more complex situations more or less correctly somehow led me to believe it would recognize the basic shapes. Apparently I was wrong.

Author:  Tryphon [ Sun May 23, 2010 10:57 am ]
Post subject:  Re: Scoring estimate question

I think a wider question that could answer all kinds of question about SE would be :

Does anyone know what algorithm SE is using ? Can we find it somewhere ?

IIRC, it wasn't coded by wms.

Author:  Joaz Banbeck [ Sun May 23, 2010 11:56 am ]
Post subject:  Re: Scoring estimate question

tj86430 wrote:
Any idea why SE shows this situation as seki:

Click Here To Show Diagram Code
[go]$$
$$ . . . . . . . |
$$ . . . O . O . |
$$ . . . . . . O |
$$ . O . O O O X |
$$ . . O X X X X |
$$ . O X X . O . |
$$ . O O X X . . |
$$ --------------[[/go]


I believe black is unconditionally dead?


It is idiotic, but there is an interesting point buried in here. It happens because SE has an incomplete definition of seki.

SE seems to be assuming that a seki exists because each side has 3 common liberties. There are indeed common liberties, and 'having a set of common liberties' is part of the definition of seki. What SE is missing is the part of the definition that continues "...and there are no unused liberties".

The 1-1 point is an unused liberty. In other words, it is a point that either side could use to gain a liberty, and thus possibly break the alleged seki.


Click Here To Show Diagram Code
[go]$$B Black attempts to use the previously unused liberty
$$ . . . . . . . |
$$ . . . O . O . |
$$ . . . . . . O |
$$ . O . O O O X |
$$ . . O X X X X |
$$ . O X X . O 1 |
$$ . O O X X 2 . |
$$ --------------[[/go]


What does SE think of this position?

Author:  Solomon [ Sun May 23, 2010 12:19 pm ]
Post subject:  Re: Scoring estimate question

Joaz Banbeck wrote:
Click Here To Show Diagram Code
[go]$$B Black attempts to use the previously unused liberty
$$ . . . . . . . |
$$ . . . O . O . |
$$ . . . . . . O |
$$ . O . O O O X |
$$ . . O X X X X |
$$ . O X X . O 1 |
$$ . O O X X 2 . |
$$ ---------------[/go]

What does SE think of this position?

Using the SE used for KGS, it considers that position seki. Along with the following conclusions:

Click Here To Show Diagram Code
[go]$$W Dead.
$$ . . . . . . . |
$$ . . . O . O . |
$$ . . . . . . O |
$$ . O . O O O X |
$$ . . O X X X X |
$$ . O X X . O 1 |
$$ . O O X X . . |
$$ ---------------[/go]


Click Here To Show Diagram Code
[go]$$W Seki.
$$ . . . . . . . |
$$ . . . O . O . |
$$ . . . . . . O |
$$ . O . O O O X |
$$ . . O X X X X |
$$ . O X X . O . |
$$ . O O X X . 1 |
$$ ---------------[/go]


Click Here To Show Diagram Code
[go]$$W Seki.
$$ . . . . . . . |
$$ . . . O . O . |
$$ . . . . . . O |
$$ . O . O O O X |
$$ . . O X X X X |
$$ . O X X X O . |
$$ . O O X X . . |
$$ ---------------[/go]


Click Here To Show Diagram Code
[go]$$W Seki.
$$ . . . . . . . |
$$ . . . O . O . |
$$ . . . . . . O |
$$ . O . O O O X |
$$ . . O X X X X |
$$ . O X X . O . |
$$ . O O X X . X |
$$ ---------------[/go]


Click Here To Show Diagram Code
[go]$$W Seki.
$$ . . . . . . . |
$$ . . . O . O . |
$$ . . . . . . O |
$$ . O . O O O X |
$$ . . O X X X X |
$$ . O X X . O . |
$$ . O O X . 1 . |
$$ ---------------[/go]


Click Here To Show Diagram Code
[go]$$W Dead.
$$ . . . . . . . |
$$ . . . O . O . |
$$ . . . . . . O |
$$ . O . O O O X |
$$ . . O X X X X |
$$ . O X X . O 1 |
$$ . O O X . . . |
$$ ---------------[/go]



Click Here To Show Diagram Code
[go]$$W Seki.
$$ . . . . . . . |
$$ . . . O . O . |
$$ . . . . . . O |
$$ . O . O O O X |
$$ . . O X X X X |
$$ . O X X . O 2 |
$$ . O O X . 1 . |
$$ ---------------[/go]


Click Here To Show Diagram Code
[go]$$W Seki.
$$ . . . . . . . |
$$ . . . O . O . |
$$ . . . . . . O |
$$ . O . O O O X |
$$ . . O X X X X |
$$ . O X X . O 2 |
$$ . O O X . 1 3 |
$$ ---------------[/go]

Author:  Phelan [ Sun May 23, 2010 1:34 pm ]
Post subject:  Re: Scoring estimate question

Tryphon wrote:
I think a wider question that could answer all kinds of question about SE would be :

Does anyone know what algorithm SE is using ? Can we find it somewhere ?

IIRC, it wasn't coded by wms.

I remember a mention of this somewhere. The algorithm itself wasn't posted, but the author was. It was a japanese name.

As for me, ever since I've heard it called that way, I refer to it as the "Score Guesstimator". ;-)

Author:  Mef [ Sun May 23, 2010 4:02 pm ]
Post subject:  Re: Scoring estimate question

Phelan wrote:
Tryphon wrote:
I think a wider question that could answer all kinds of question about SE would be :

Does anyone know what algorithm SE is using ? Can we find it somewhere ?

IIRC, it wasn't coded by wms.

I remember a mention of this somewhere. The algorithm itself wasn't posted, but the author was. It was a japanese name.

As for me, ever since I've heard it called that way, I refer to it as the "Score Guesstimator". ;-)


Hehe, some affectionately call it "Score Randomizer"

Author:  xed_over [ Mon May 24, 2010 9:07 am ]
Post subject:  Re: Scoring estimate question

Mef wrote:
Phelan wrote:
As for me, ever since I've heard it called that way, I refer to it as the "Score Guesstimator". ;-)


Hehe, some affectionately call it "Score Randomizer"

and its even more fun when non-Japanese rules are used -- drives the kibitzers crazy

Author:  wms [ Mon May 24, 2010 9:59 am ]
Post subject:  Re: Scoring estimate question

Yeah, the score estimator was written by a Mr. Yoshikawa. He was a friend of Richard Bozulich's. I met him in Japan and he offered his board evaluation code to KGS, I built the estimator out of it from that.

I do not know how the algorithm works. Several people have looked at the code (I give it away when Mr. Yoshikawa gives me permission to). They all wanted to make small, simple changes (work with non-19x19 boards, let the player say that a group is alive or dead, etc.)...they all ran away screaming when they saw the code. It is very, uhhh, peculiarly, written. :o

Author:  Li Kao [ Mon May 24, 2010 11:49 am ]
Post subject:  Re: Scoring estimate question

Why does SE only work with the offline client? Does it contain unverifiable or native code?

Author:  wms [ Mon May 24, 2010 12:35 pm ]
Post subject:  Re: Scoring estimate question

Originally, the SE was almost as big as the rest of the client put together, so I left it out of the applet to help people with slow internet connections.

But now the client is bigger and very few people are still on dialup, so it makes no sense to leave it out. When I get around to it I should add it to the applet.

Author:  Marcus [ Tue May 25, 2010 8:53 am ]
Post subject:  Re: Scoring estimate question

wms wrote:
Yeah, the score estimator was written by a Mr. Yoshikawa. He was a friend of Richard Bozulich's. I met him in Japan and he offered his board evaluation code to KGS, I built the estimator out of it from that.

I do not know how the algorithm works. Several people have looked at the code (I give it away when Mr. Yoshikawa gives me permission to). They all wanted to make small, simple changes (work with non-19x19 boards, let the player say that a group is alive or dead, etc.)...they all ran away screaming when they saw the code. It is very, uhhh, peculiarly, written. :o


This makes me VERY curious. Any chance you could get permission to pass along the code to me? I don't need it or anything, I just want to look at the code. Anything that resists refactoring like that MUST me interesting. :D

Author:  wms [ Tue May 25, 2010 9:25 am ]
Post subject:  Re: Scoring estimate question

Marcus wrote:
This makes me VERY curious. Any chance you could get permission to pass along the code to me? I don't need it or anything, I just want to look at the code. Anything that resists refactoring like that MUST me interesting. :D
You'll have to find and ask Yoshikawa yourself.

But here's a hint as to the troubles with it:

* No comments
* Functions thousands of lines long
* All variables (and most functions and even many of the classes) have single-letter names
* Chars used as enums (e.g., code like "if (y < 5) { z = 'j'; }" is all over the place.)

So you've been warned. :twisted:

Author:  Harleqin [ Tue May 25, 2010 9:55 am ]
Post subject:  Re: Scoring estimate question

Heh, me too :).

Page 1 of 2 All times are UTC - 8 hours [ DST ]
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group
http://www.phpbb.com/