Forums

If a supercomputer with depth ℵ₀ exists, shouldn't a chess move just return 6 type of evaluation?

Sort:
tygxc

#120
Any position is within reach of present computers. You get a fairly accurate evaluation rapidly, but with more and more calculation time you get more and more confidence.
However, there is no point in expending time, money, and effort to evaluate a nonsensical position.

R3tdog

In 10 moves (5 for each player) there could be over 69 trillion combinations. To calculate an entire chess game (starting from about this point) would be around 10^100+ combinations. A computer system with one PetaFlop could calculate 10^15 positions per second (less but just for estimation). At this rate it would take 1*10^84 seconds to finish, or more time than has passed in the entire universe so far, and more time than probably will ever pass, as it is 3*10^76 years. So that is why this can never happen, every possible calculation cannot be performed.

R3tdog

And the problem with quantum computers is that there is no way that we know of to eliminate the "wrong" positions using destructive convergence or whatever it's called.

jonnin

If you have an unrestricted, theoretical computer, then every possible position would be cataloged and everything would be known.  At that point, one of two things is going to happen in perfect computer vs other perfect computer:

1) its going to prove that white can always force a win. 

2) its going to prove that chess is always a draw.

-- the 3rd possible, that black can always win, seems too remote to consider, but it 'could be'. 

in that context, every position will have, as you said, one of 6 outcomes (win,lose, draw for 2 sides = 6) if say the computer took over midgame for both sides.  

the problem may be solved in the next 100 years.  Chess does not have that many actual board positions, given modern computer power and considering transpositions (same position, different rotations, eg k vs K+R endgames are all the same, just rotate the board)

haiaku
R3tdog wrote:

In 10 moves (5 for each player) there could be over 69 trillion combinations. To calculate an entire chess game (starting from about this point) would be around 10^100+ combinations. A computer system with one PetaFlop could calculate 10^15 positions per second (less but just for estimation). At this rate it would take 1*10^84 seconds to finish, or more time than has passed in the entire universe so far, and more time than probably will ever pass, as it is 3*10^76 years. So that is why this can never happen, every possible calculation cannot be performed.

That was my objection in post #81, the Shannon number, but tygxc says that only 10^38 positions are sensible and only 10^19 "relevant", and claims that 5 years could be enough to solve chess, quoting a claim by Sveshnikov. AFAIK, no computer scientist or software developer supports that claim with a proof of concept, but he insists it is feasible. You have to read all tygxc's posts in this thread to decide if his POC does convince you. At least, it seems that we all agree that chess will not be solved in the next 5 years, whatever the reason. Anyway, as Ilampozhil25 noted in #116, the discussion about how to solve chess is quite OT.

p1mpinpauly

did not real all of this, but did super glance at it all, these guys are going at it, but in a respectful way. thats whatsup!

CheesePrix2314
p1mpinpauly wrote:

did not real all of this, but did super glance at it all, these guys are going at it, but in a respectful way. thats whatsup!

Yeah, agreed. Appreciated. Although it's partly because trolls do not understand chess correctly.