On 'Solving' Chess...

[swami]

If you had a CHOICE, which 'Scenario' would you choose?!


Scenario 1:

I support the idea that chess should never get solved, In short, I DON'T want to see chess getting solved in near future, I want it to last forever, I'd love to get a lot of chess engines as possible, either new or updated.

I would love to see computer chess alive, and new young programmers to contribute engines via their own unique programming techniques.

I want the ELO limit to go beyond 3500 elo. I want programmers to find hidden weaknesses in their strongest engine. If chess is solved, I see it as the drawback to our giant Computer Chess world.

Bottomline: don't want to see chess getting solved in forseeable future.


Scenario 2:

I love to see Quantums or thousand core comps(or whatever it takes) solving the game of chess and be done with it. I would love to wait and see who would be the first guy to solve the game. I don't care whether the game is a draw, loss or win, I think We have had enough chess already, there's nothing new in the game anymore, 3500 Elo is the limit and chess is draw!

Just solve it as soon as possible and put the game back in the league of Sudoku, Checkers and other solved board games. I'd move on to other complex board games like Go.

I'd very much hope that the 'GO' would then be popular as chess was, I'd hope to see more programmers take up 'GO' Programming, offer free Go engines, just like the how computer chess was. Promote Go Compeitions in real life.

Bottomline: Solve 'Chess'.

[tano-urayoan]

What do you mean by "solve"?

[swami]

What do you mean by "solve"?

Well, It's a bit complex question. If you had followed the discussion in Rybka forum and some threads in here, you'd know what it means!

[Kirill Kryukov]

What do you mean by "solve"?

Check this out.

[CRoberson]

Lets see what is needed to do this: (the math below is simplified to
get the point across)

In 1 minute on my Core 2 Duo, Rybka 2.3.2a 32 bit searched
17 ply from the root position. Solving requires us to start from
the root position.

Now, lets make things super easier: we will look at only the first
50 moves that is 100 ply. So, how much longer
than 1 min for 17 ply will it take to complete 100 ply from the
root position?

Assuming an EBF of 2 (this is for Rybka to search in its normal
way which is not a full search and is not enough to answer the
question at hand as that would need a average branch factor of 40),
then the number of nodes in the search is 2^100 - a simple search
starting at 100 ply (no iterative deepening).

Code: Select all

     Number of minutes to complete = nodes / (nodes to 17 ply in 1 min)
                          = 2^100/2^17 = 2^83.

Now, lets asume we have a computer about 1 billion times faster
than mine. The time to complete is 2^83 / 2^30 = 2^53.

2^53 minutes is approximately 8,000,000,000,000,000 minutes.
Actually, it is more than that since 2^10 = 1024.
In Hours that is about 2^53 / 2^6 = 2^47
In 5 day periods it is about 2^47/2^7 = 2^40
That means 5 * 2^40 days which is about 5,000,000,000,000 days
which is about 15 billion years.

So, Rybka on a computer 1 billion times faster than mine could search
from the root position to the 100th ply (to move 50 by human counting) in 15 billion years.
It would take more than that to solve chess.

[Kirill Kryukov]

Well.. Forward search is not the only way to solve chess. If this was all we could do, 6-men chess endgames would be still unsolved.

(Plus, there is some hope that quantum computing may help, which works fundamentally different from what we use now.)

[Bill Rogers]

With millions and millions of games that can be analized it is possible that the perfect evaluation subroutine may be created and thus regardless of what opening may be used what we might call a 'perfect' game may be played and thus considered 'solved'.
Bill

[Dann Corbit]

With millions and millions of games that can be analized it is possible that the perfect evaluation subroutine may be created and thus regardless of what opening may be used what we might call a 'perfect' game may be played and thus considered 'solved'.
Bill

It is possible that there is a forced solution 30 plies from the origin that nobody has seen. In a few years the game of chess would be solved by alpha-beta and exponential hardware strength increase. Of course, it is also possible that the solution is 10000 plies away, in which case computers won't solve it by conventional search.

Of course, a forced 15 move checkmate *ought* to have been discovered by now but you never know.

Rybka can currently search about 26 plies from an opening position on my machine in 3 hours. Given one ply added per year, 30 plies is only 4 years away and 60 plies is only 34 years away.

It could be that a computing revolution will allow machines trillions of times faster than the current machines and it could be that the solution to chess (if any) is not far from the origin. So whether or not chess is possible to solve will depend very much on whether there is a solution nearby or not. The full 12K ply monstrosity is "neigh onto impossible" to search.
IMO-YMMV.

[Steelman]

Chess will some day be "solved".

I think that someday a computer may be able to execute software using no time at all. Or perhaps very close to no time at all.
If this happens then solving the game of chess......well - no problem.

How many years could it take from now? 100 years? 1000? years? Longer?
Look where computer science has taken us in such a short time.
Well I figure it will happen.

What would be the answer?....DRAW in 150 moves? Now that would be disappointing!