[computer-go] GTP and handicap

[Mark Boon]

On 7-sep-06, at 16:47, Gunnar Farnebäck wrote:

> Except for the conceptual difference between making moves and placing
> handicap stones there is a practical difference between knowing how
> many handicap stones to place and not knowing. For example an engine
> might prefer placing three handicap stones all on one side (e.g.
> san-ren-sei) but four handicap stones in all four corners.

That's a good argument.

The protocol seems a bit convoluted when it comes to handicap, so I  
was just wondering if there's a way to simplify.

Next the discussion moved to rule-sets. I agree the protocol needs to  
cover that. But from the discussion it's also clear that it will  
complicate the protocol even further to cover all the possibilities.

It's amazing we have such a plethora of complications for what is  
basically the same game. All the rules intend to define exctly the  
same game, they just have a different philosphy how to go about doing  
so. How do Go players manage at all?

The reason we manage is that the number of rules applied in practice  
is very small. We have Japanese rules, Chinese Rules and Ing rules.  
Computers maybe use Tromp/Taylor rules at times. That's about it. If  
we'd restrict ourselves to these four, what do we lose? Nothing IMO.  
If you want a mathematically sound set of rules you need to go with  
something like Tromp/Taylor rules anyway, the others are just  
practical interpretations for human players that have a historical root.

So if you ask me we just use a simple enumeration of these four rule- 
sets. All the rest of the discussion is boring and nitpicking to the  
point it doesn't add any real value. Whether you use simple ko or  
super-ko, suicide or not, it has an effect on the game in practice  
that approaches zero to the point I really couldn't care less. All it  
does is preventing from having a good standard protocol.

Mark

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://computer-go.org/pipermail/computer-go/attachments/20060908/2c28af85/attachment.htm