[Computer-go] pachi2

[fastgo]

:) Yeah, I was thinking of the bottom left corner. I did not read the whole
game. So naturally thinking of it in the bottom left corner.

I had trouble with the same situation as well. And I agree with both of you
that it is difficult to get it right without sacrificing speed.


On Thu, Jan 13, 2011 at 10:32 AM, <valkyria at phmp.se> wrote:

> I know the source of the confusion now. The first example was "Bottom left
> corner of 19x19 board" as Petr wrote it so the corner in the second example
> must be the bottom right. If it is located in the lower left corner then all
> you wrote made sense, but I am talking about the case where there is a
> "1-approach liberty" whcih is only for the case where this pattern is
> located in the lower right corner and is much hardwer to solve correctly.
>
> :)
>
> Magnus
>
>
> Quoting fastgo <fastgo at gmail.com>:
>
>  I guess we are not talking about the same case or I may have misread the
>> issue or the example. I was commenting on the second example. My take is
>> that it is corner as Petr stated "Example is the other bottom corner of
>> the
>> same game:"
>>
>>    O O . . . .
>>    O O X X X X
>>    . X O O X .
>>    . X O . O X
>>    . X O O . X
>>
>> I meant to say that if it is corner as the diagram shown, then the playout
>> should almost always get it right. As the middle bottom white group has
>> only
>> 2 liberties left and the left black group has 3 liberties. Any attempt by
>> the white group to take one liberty of the left black group will trigger
>> black to immediately check its surroundings and kill the white group.
>>
>> Even in cases where the liberties of the 5 liberty group are played first
>> by
>> the white group, and when it goes down to 2 liberties, the black will
>> check
>> its surroundings and kill the bottom white group first.
>>
>>
>>
>> On Thu, Jan 13, 2011 at 9:42 AM, <valkyria at phmp.se> wrote:
>>
>>  Quoting Go Fast <fastgo at gmail.com>:
>>>
>>>  I do not see any difficult in handling the second example.
>>>
>>>>
>>>> Whenever the white tries to take out one liberty of the corner X group,
>>>> the
>>>> 2-lib check should kick in and immediately put the bottom O group into
>>>> atarai and it has no escape at all.
>>>>
>>>>
>>>
>>> Look. If white takes one liberty away from X that group has 4 liberties.
>>> I
>>> do not undestand what you mean with a two liberty rule. But I fear you
>>> give
>>> it magic properties it does not have and it sound you made a tactical
>>> oversight from what you wrote.
>>>
>>> The atari is a self atari  (X must connect the two X stones to the 5 Xes)
>>> which means the two liberty white group has three liberties effectively.
>>> And
>>> since X needs to play an approach move and fill in a liberty X has
>>> effectively less liberties. This is a lot of computations to do and if
>>> you
>>> do it everytime any group loses a liberty you have a huge impact on the
>>> speed of the playouts.
>>>
>>> Capturing 2-liberty groups everytime a 5-liberty neighbor loses a liberty
>>> is also not a solution to the problem it will just make a lot of
>>> unneccessary moves and bias the playout in new weird ways, because most
>>> such
>>> situation are not a semeai with an eye and approach move.
>>>
>>> If this was trivial Petr and I would not independently struggle with the
>>> same problem.
>>>
>>> Magnus
>>>
>>>
>>>
>>>  On Thu, Jan 13, 2011 at 8:46 AM, <valkyria at phmp.se> wrote:
>>>>
>>>>  Quoting Petr Baudis <pasky at ucw.cz>:
>>>>
>>>>>
>>>>>
>>>>>    O O . . . .
>>>>>
>>>>>>   O O X X X X
>>>>>>   . X O O X .
>>>>>>   . X O . O X
>>>>>>   . X O O . X
>>>>>>
>>>>>>
>>>>>>  Funny! This is exactly the kind of shape that I had to struggle with
>>>>> in
>>>>> Valkyria and which I cover in my other reply.
>>>>>
>>>>> Magnus
>>>>>
>>>>>
>>>>
>>>
>>> _______________________________________________
>>> Computer-go mailing list
>>> Computer-go at dvandva.org
>>> http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
>>>
>>>
>>
>
> _______________________________________________
> Computer-go mailing list
> Computer-go at dvandva.org
> http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://computer-go.org/pipermail/computer-go/attachments/20110113/90a07388/attachment.html>