Re: How long would it take a computer to completely "solve" chess?

From: fs (a_at_b.c)
Date: 09/15/04 

Date: Wed, 15 Sep 2004 05:37:40 GMT

"Sean O'Leathlobhair" <jwlawler@yahoo.com> wrote in message
news:d1835a57.0409140718.4a9f228b@posting.google.com...
> dwb1729@yahoo.com (David Bandel) wrote in message
news:<a88af92f.0409131300.23273807@posting.google.com>...
> > "fs" <a@b.c> wrote in message news:<Bb21d.543$g9.470@trnddc06>...
> > > "Luigi Caselli" <luigicaselli@anyspamrefusediol.it> wrote in message
> > > news:oD11d.90599$1V3.2233631@twister2.libero.it...
> > >
> > > > But solving the game, I mean every chess positions, is another
thing and
> > > > IMHO it's impossible to do for a computer in our universe lifetime.
> > > >
> > > > Luigi Caselli
> > >
> > > An alpha-beta infinite-ply search (with an obscene amount of time)
would
> > > "solve" chess without examining every position, so the terms "solving
the
> > > game" and "every chess position" are two different things.
> >
> > the exact point i've been making. PRUNING occurs
>
> But if we do any pruning, can we be sure that we have not missed some
> very obscure trick? We may have solved the game for all practical
> purposes but if we have pruned anything, could we really say that we
> have theoretically solved chess?
>
> Seán O'Leathlóbhair

Yes, we can be sure because alpha-beta does not *forward* prune. For
example, if 1.e4 is searched and is found to be a forced win for white, do
we need to search 1.d4 and all the rest? No, we can stop and say we have
theoretically solved chess. Alpha-beta pruning does the same thing on a
smaller scale along the branches and stems of the tree. Searching black's
moves after 1. g4 e5 2. f3 we find that white will lose to 2...Qh4#
Therefore, we can go back to f3 and prune all possibilites branching out
from there.

(the rest is kinda off-topic)

Forward pruning is speculative. An engine looks at the most promising
lines, and ignores the rest, possibly missing winning lines. From what I
understand, Fritz does this based on the "null move" concept. Usually, it
works great, but sometimes it overlooks zugzwang concepts. On infinite
analysis mode, Fritz8 swears that 49...Ra8 will draw this position:

3r4/Pk5p/R2P2p1/1P3p2/2K2P2/6P1/7P/8 b - - 0 49

So, Fritz annotated the actual move (49...Ka8) by saying that it "ruined a
great position!" After either move is made, however, Fritz realizes that
Black is hopeless.