[Computer-go] God Go Rank

Nick Wedd nick at maproom.co.uk
Tue Oct 12 13:31:04 PDT 2010

In message <alpine.DEB.1.10.1010121924520.17507 at phare.normalesup.org>, 
Kahn Jonas <jonas.kahn at math.u-psud.fr> writes
>> Brian Sheppard wrote:
>>> I am not sure what this means for dan ranking. But if 1 stone = 6 points,
>>> and 1 error = 2 points, then perfect players are about 4 stones better
>>> than top human pros.
>> The rest of your analysis seems good, but one stone is surely much more
>> than six points (twice komi is the obvious approximation), and (at least
>> for Japanese and Korean pros) it seems more natural to say that one
>> error is one point.
>Under Chinese rules, the minimum error is more or less systematically 2 points.

This is only true if you use a very pedantic measure of the size of an 
error, assuming perfect play for all subsequent moves.  More 
realistically, a move can give away say 0.37 of a point, because it 
makes it that much less likely that the player will play subsequent 
moves correctly.


>In fact, you could argue that one-point error under Japanese rules are 
>either two-point or zero-point error under Chinese, but knowing which 
>is almost
>impossible: you have to play out the entire game to see who gains one
>dame in the end.
>My gripe would be the other way around: there is no reason to think that
>every error should be minimal. Of course I know the previous analysis
>worked by assuming that all errors had the same value, but then mean
>error value looks like a better estimate of the value of an error. Maybe
>this compensates the other underestimation.
>Computer-go mailing list
>Computer-go at dvandva.org

Nick Wedd    nick at maproom.co.uk

More information about the Computer-go mailing list