Re: [computer-go] Hardware rules in Competitions

[Paul Pogonyshev <pogonyshev@xxxxxxx>]

Frank de Groot wrote:

> GnuGo is one of the strongest Go programs around, perhaps even the
> strongest (soon).

Hardly _the_ strongest anywhen soon.  Development have been really slow
in the past year or even more.

> There is a big problem with that, and that problem appeared with SlugGo.
>
> The problem with the strongest Go program being Open Source is that winning
> a Computer Go contest simply becomes a matter of who is the richest person,
> not who has the best program.

This suddenly discards everything said here about entering only your own
programs in tournaments (even if the exact criteria of ``originality'' is
not anywhere clear.)

Also, what prevents anyone with a strong Go program to find a rich sponsor
and get superior hardware?  Yes, with programs such as GNU Go the sponsor
can find it instead, but it seems like a minor issue to me.  If anyone
with money gets serious about winning a computer Go tournament (which sounds
unlikely), he can always buy any non-free engine as well.

> If I rip of GnuGo's engine and buy a Tyan quad Opteron blade and put
> dual-core CPU's in it and modify GnuGo so that it uses multithreading well,
> I have a "winning" 8-CPU cluster that can be conveniently carried on
> handles by two people, it's not even heavy. I will most probably win the
> contest without doing much real programming.

Well, to make you feel a little more comfortable, I can assure that GNU Go
is _absolutely_ not scalable at the moment (you can scale up with levels,
but that is very weak and sometimes leads to more problems than gains.)
There has been some effort put into this, but it has lead to nothing.

> Not that I propose we force people to use standardized hardware, but it
> sure is a dilemma.
> The danger is that some guy with a lot of cash will forever win all
> competitions because a clever team (the GnuGo team) delivers the strongest
> program, and the only ingredient neccessary to remove all excitement about
> who will win the tournament is *money* for the fastest PC.

I believe if this happens, the ``forever'' will not last for more than some
two tournaments and rules limiting hardware will appear quickly.  But I
don't see any point in trying to set them up in advance, given that there
has been no significant issues about superb hardware so far.

It seems to me that Go difficulty is non-linear.  And the reported success
of SlugGo is because of additional layer (global lookahead), not because
of additional time.  Unlike others here, I don't think that the reported
increase in strength of 3--4 stones is not worth the 100x (or what?)
slowdown.  (When we are reluctant to accept patches that slow GNU Go down
10%, the reason is that we want to keep it fast enough for human opponents
to not get bored.  So such slowdown would be absolutely unacceptable in GNU
Go, but not because it is not worth it, but because it makes GNU Go
uninteresting for users.  SlugGo, on the other hand, is not limited by
this consideration, and I think that their slowness is well worth the added
playing strength.)

> So perhaps we can just continue like always, but *also* publish an
> interesting metric: The ranking normalized per SPEC MIPS / second of
> average thinking time.
>
> Just for interest, because that shows us programmers which program *really*
> is the strongest, normalized for CPU power and time used.
>
> (there are many "issues" with this metric as well, but I would be
> interested in knowing it..)

There are indeed many issues...

Paul

_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/