Re: computer-go: prevent cheating

[Mika Kojo <mkojo@xxxxxx>]

birk@xxxxxxxxxxxxxxxxx writes:
> > swang@xxxxxxxxxxxxxxxxx wrote:
> > What about mapping the search-depth in 256 level, and the Go-playing
> > program send the resulting play with the search-depth value, the one
> > where the program make the playing decision.
> 
> But 'search-depth' is not the only variable in a go-program!
> E.g. strategic decisions like 
> - if I  have a lot of time but the opponent does not, I could try 
>   a more complicated variation
> - I am short of time so I omit a complete life-and-death analysis
>   but use some heuristics 
> etc.

The 'search-depth' should be seen in a more abstract way. Basically
the point seems to be that the 'search-depth' can be at most 256
levels in size, but it is not defined what it actually means. (As it
would be very hard to prove.)

Basically, cheaters would encode enough information in this
'search-depth' block to guide their programs in the replay. It is
quite easy to see that even a pseudo-random player program could play
some ok games with this additional information. 


> Besides that I think your proposal has even bigger problems:
> 2) An official standard GUI
> 3) An official standard for initial parameters
> 5) An official standard for running platform

These could to be very nice for the organizers! 


-- Mika