[Computer-go] Combinatorics of Go

[Erik van der Werf]

On Mon, Jan 3, 2011 at 1:00 PM, Robert Jasiek <jasiek at snafu.de> wrote:
> On 03.01.2011 12:33, Erik van der Werf wrote:
>>>>
>>>> "Under Japanese style ko rules, the long-term history is never needed
>>>> to infer the best move".
>>>
>>> Do you mean long>2 or long>cycle_length?
>>
>> Roughly speaking 'long>2'; to be more precise these are the extra
>> state properties I typically use in my solver:
>>
>> - player to move
>> - basic ko
>> - number of consecutive passes
>> - last play was basic ko capture (before passing started)
>> - last two plays were basic ko captures '
>> - number of Black/White captures
>> - number of Black/White passes (not necessarily consecutive)
>
>
> Now why may you say "never" in your conjecture?!

Perhaps I'm being a bit provocative ;-)

> To start with, please demonstrate how you handle
> - sending-2-returning-1 cycle

This is handled trivially by observing that one sided passes/captures
more in each cycle.

> - eternal life cycle

This will generally lead to the search not returning a result. If the
eternal life shape is crucial for the win then I expect the search to
propose to keep on playing in it at least until perhaps some higher
level criterion at the root decides to claim jigo or no-result (just
analyzing the PV should be enough to determine when the game is in a
never-ending cycle).

> - 1-stone suicide cycle
> - 2-stone suicide cycle

Suicide is illegal under Japanese rules.

Robert, please forgive me if I don't reply to follow-up questions.
I've already talked too much off topic the last days and I really have
to get back to work now. I'm quite sure, if there really is an
interest, that you and/or others can work out the details.

Best,
Erik