Machine vs Machine

The current thread re Chess Quotes contained this, which I had not previously encountered: "I think it's almost definite that the game is a draw theoretically." -- FISCHER

That brought to mind the computer in the movie War Games - nuke war averted via the machine coming to the analysis that tic-tac-toe is, when played right, inevitably, a draw. So what happens when Fritz plays Fritz, or Deep Blue, Deep Blue? Any literature or web links re such?

Was Fischer right about that?

Originally posted by snowblind2
The current thread re Chess Quotes contained this, which I had not previously encountered: "I think it's almost definite that the game is a draw theoretically." -- FISCHER

That brought to mind the computer in the movie War Games - nuke war averted via the machine coming to the analysis that tic-tac-toe is, when played right, inevitably, a draw. So wh ...[text shortened]... Deep Blue, Deep Blue? Any literature or web links re such?

Was Fischer right about that?

Playchess has an engine room where engine is pitted against engine, often with results favouring one or the other, having said that the time settings are normally blitz. There are computer vs computer long matchs, in which some engines do well and others not so well.

As far as the same engine vs itself goes its the engine has to make a mistake in not looking deep enough, which happens quite often. I havent played any recent engine vs engine games with the likes of Rybka or H10 etc but did play a few with F8, about only 60% of the games were drawn.

Maybe you should look up which engines the banned players on this site were using and see if any two had the same and if they played ;-)

Originally posted by snowblind2
The current thread re Chess Quotes contained this, which I had not previously encountered: "I think it's almost definite that the game is a draw theoretically." -- FISCHER

That brought to mind the computer in the movie War Games - nuke war averted via the machine coming to the analysis that tic-tac-toe is, when played right, inevitably, a draw. So wh ...[text shortened]... Deep Blue, Deep Blue? Any literature or web links re such?

Was Fischer right about that?

Whether or not chess is a theoretical draw doesn't matter when we are talking about engine-engine matches as chess cannot be analysed in the same way that tic-tac-toe (and connect 4 and to some extent checkers) can.

Here are the results for the 2005 Computer Chess World Championship. Lots of wins and losses happening there.
http://www.ru.is/wccc05/default.asp?Page=Notepad&ID=3

EDIT: You'll note that the winning engine Zappa scored 10.5/11 in the standard competition.

Bedlam & XanthosNZ, thx for the replies & info. Interesting, to me anyway, that 'engine vs engine' = the occasional loss, from which, I conclude that even if Fischer was right about that, we're a very long way from 'proving' it.

Originally posted by snowblind2
Bedlam & XanthosNZ, thx for the replies & info. Interesting, to me anyway, that 'engine vs engine' = the occasional loss, from which, I conclude that even if Fischer was right about that, we're a very long way from 'proving' it.

Depends on the style of the engine, if you made an engine that would just play 1.e4 e5 2. Ke2 Ke7 3. Ke1 Ke8, well draw every game. If your engine of choice enjoys a more daring approach to chess then you will get a lot less draws 🙂

Engines like gambit tiger or Hiarcs 10 set to agressive play or Rybka set to ultra tactical and speculative would get a lot few draws.

Game theory

http://en.wikipedia.org/wiki/Combinatorial_game_theory


Basically any 2 players game tht does not include luck when played perfectly will always be a draw....

on saying that however, if Chess is Ever mastered like naughts and crosses (tic tac toe) then we may find the advantage of being white, and moving first, is winning - or, draws every single game....

when you stand to grind down, we may find that the best opening is "x" -- being 0.000000000076 better than anything else......

anyhow -- were a long way off a complete solution to the game -- engines are yet to play the game perfectly...(with the execption of some endgames)

if you would pit two exactly identical engines against each other, wouldn't one always see a move farther than the other, before the horizon kicks in? so, neither of them would ever see if there was a blunder at horizon+1 when making their move. then, it would be pure luck wich side encounters the first blunder, losing the game. unless there is subsequent blunders similarly. sounds very chaotic.

105197 games played by 275 computers
http://web.telia.com/~u85924109/ssdf/list.htm

this site gives some info on computer play ...

it also seems that programs are rapidly getting better and better (on the same hardware).

on saying that however, if Chess is Ever mastered like naughts and crosses (tic tac toe) then we may find the advantage of being white, and moving first, is winning - or, draws every single game....

when you stand to grind down, we may find that the best opening is "x" -- being 0.000000000076 better than anything else......

anyhow -- were a long way off a complete solution to the game -- engines are yet to play the game perfectly...(with the execption of some endgames)

I've been thinking wayyyy too much about this.

1) In any chess game, there are a finite amount of possible moves
2) In any chess game, there are a finite amount of possible chess positions
3) In any chess position, therea are a finite amount of possible move choices
This implies:
There is a finite amount of possible chess games. (Even if the number is astronomically huge).


Since there are a finite amount of chess games
this implies: optimal games must exist for both players,
which implies: optimal positions must exist for both players
which implies: optimal choices must exist for both players

Since white always plays first, if both players play optimally from the start (ie. play a 'perfect game'😉, the outcome will always be the same (the 'perfect outcome'😉

There are three possible outcomes in a chess game: a tie, a win for white, and a win for black. The 'perfect outcome' is one of these outcomes, but it is only one of them. Until the day computers are fast enough to computer every possible chess game, we can only speculate which one it is.

By definition, every 'perfect game' will have the 'perfect outcome', but that does not mean that every game with the 'perfect outcome' is a 'perfect game', far from it.

Since a win is a win and a tie is a tie, there may be many different 'perfect games'. Perhaps hundreds, thousands, millions, or even more, games with different, but still optimal choices made by both sides, that still lead to the 'perfect outcome'. Perhaps, but unlikely, some of them may have been played already.

Anyways, I can go on or clarify this if anyone actually cares....

To get to the point:

calculating how many possible chess games there are:

white has 20 starting chess moves
black has 20 starting chess moves

After 2 moves, there are 20 * 20 = 400 possible chess positions

White then has about 400 * 20 = 8000 possible moves

Black then has about 8000* 20 =160,000 possible moves.

If every chess game was only 3 moves long, there would be 160,000 possible chess games!

Expanding this out, there are approximately 10^120 (10 to the power of 120)chess games.

Even if computers do solve every optimal solution, just imagine how many of them there potentially are! Even if only 10,000 games (10^4) are optimal, could one memorize them all??? Could we resist the psychological effect of a near-perfect, but not optimal queen sacrifice??

Originally posted by bosintang

Interesting posts. Some comments…

>> “Until the day computers are fast enough to computer every possible chess game, we can only speculate which one it is.”

It wouldn’t need to be every possible game. For any position, once a winning move is identified, there’s no need to try any remaining untried moves.

>> “Even if only 10,000 games (10^4) are optimal, could one memorize them all???”

Hypothetically, supposing you could memorise all the optimal games, what would that gain you? If your opponent played a sub-optimal move, you could only tell that it was sub-optimal, but not how to take advantage of it.

..double post..

Originally posted by Varenka
Interesting posts. Some comments…

>> “Until the day computers are fast enough to computer every possible chess game, we can only speculate which one it is.”

It wouldn’t need to be every possible game. For any position, once a winning move is identified, there’s no need to try any remaining untried moves.


This is correct. I should've written 'Until a game is mathematically proven to be perfect.....'

In all likelihood, if a perfect game is ever discovered, it will come from the 'bottom-up' rather than 'top-down', meaning that the proof that the game is perfect would start from possible endgames, and work backwards from the last move to the first move, rather than vice-versa.

As well, it would likely not be a brute-force solution (trying every possible combination), but from one that would systemically eliminate move possibilities, like current chess programs, and certainly humans, do already. Although to be a mathematically sound proof, it would still somehow have to be proven that those choices are not optimal choices, with 100% certainty.

>> “Even if only 10,000 games (10^4) are optimal, could one memorize them all???”


Hypothetically, supposing you could memorise all the optimal games, what would that gain you? If your opponent played a sub-optimal move, you could only tell that it was sub-optimal, but not how to take advantage of it.


And thats my point..I think guessing that there are 10^4 amount of perfect chess games is an extremely conservative guess. And for each of those games there may be a billion, or even more near-perfect games with drastically different outcomes. These numbers are pure guesswork, of course, but even if machines start printing off lists of perfect games, until the day that chess theory reaches a point where a player can almost certainly determine an optimal position just by looking at the board and calculating only a few moves deep, we're along way off from chess turning into checkers.