D-Wave Systems breaks the 1000 qubit quantum computing barri

[Ozymandias]

Winning with Black is much more difficult in
computer chess btw, but having a drawing strategy for Black
isn't so difficult, and fundamentally certainly not impossible.

Let me be positive, and instead of using a double negative, say that it's "certainly possible". Only when black tries to win, can a withe win take place.

[bob]

The problem is, you can't hand-wave a "bad position" away when you are trying to actually prove something. You can't confirm sacrificing the queen on move 5 is bad until you force checkmate which might not be possible with optimal play (i.e. a fortress position, etc).

Playing a game is one thing, proving white can win or only draw is something else entirely.

[bob]

If the game is a draw, then you will almost certainly have to search to depth 11,000 plies to make sure there is no way out.

Well in top correspondence chess they don't play out drawish games,
the ICCF rules are already allowing 6men egtb's (with 50 move rule)
to decide whether and endgame is a draw. And we know (see eg other posting) that in top correspondence chess the drawing percentage is approaching 100 pct. With the current rules.

Despite the immense high nr of possible positions, most positions i.e. mistakes can be discarded because they lead to a win (for Black or White), for example, unless a sound sacrifice, in situations where one side lost a piece in the opening or (early) middle game it's fair to say that this side will lose for sure against solid (correspondence) play.

And if we would analyze 'only' games which are played reasonably well
it's possible to build up a tree, and see if an (opening) strategy can lead to a win. In *theory* there *could* be an still unfound strategy (eg with 1.d4 with steadily but slowly increasing positional advantage), still leading to a win but the chances for that are slim; and if there is *no* winning strategy(*) there is a theorem by Zermelo that the game then is a draw. This also holds true for chess although ' proving' there is no winning strategy according to proper mathematical methodology i
ndeed is a difficult task.

The empirical approach not using GM evals like eg with the NIC yearbooks but a simple Stockfish approach in a huge positionbase with now apparently some 8 billion positions has yielded nothing at:
http://www.arshah.com/
Installation doesn't work anymore, but a year ago i looked at it
and all initial values were zero (backsolved via lines after every game)
and they stayed zero. Not surprising for me, although the quality of the database of this Arshah system didn't look great; i played some games on ICC with my own book, still often managed to win with standard
time control, due to my faster comp and a later stockfish and better worked out opening strategies. Apparently arshah(C) still is running sometimes on Fics so you could have a look.

So despite not believing anymore in such a ' holy grail' i admit i still search sometimes for better/promising lines starting with 1.d4, fundamentally i can' t find more ' backsolved' advantage than eg 0.16 or so with Komodo, and that's not enough to guarantee a win. So for me it's obvious that chess with it's current rules is a draw, and in the end my most promising lines will also be equalized, although this would significantly and maybe unnecessarily increase the size of my already large opening book leading to nothing and being able to demonstrate that provided i would live that long

jef

PS (*) other example (not related to a two person non-chance game with perfect information, but a game of chance): Q) is the game of roulette a draw or a loss ? Answer: this depends on the commission for the Casino.
Do you need to simulate it with quantum computers into trill-bill-gazillions of millennia to find the answer ? No sir. With such games you need to have an *edge*, otherwise forget it. In Blackjack apparently a player with card counting could have a theoretical edge that's proven.
But in chess, even if White would have a slight edge, the drawing margin of the game prevents a fundamental win for White. And even the 'edge' is doubtful some experts argue that while White has the first move, by releasing the info which opening he wants to play, Black can use such info to his advantage thus cancelling out the 'edge'. Ergo chess = an =.

It is a loss unless you remove the 0 (and sometimes 00) slots from the wheel.

As far as claiming chess is a draw, it is only a claim. Nothing even approximating a proof is available now. And never will be. You'd have to show that EVERY possible path from starting position to final position is a forced draw. Not possible.

[Ozymandias]

The problem is, you can't hand-wave a "bad position" away when you are trying to actually prove something. You can't confirm sacrificing the queen on move 5 is bad until you force checkmate which might not be possible with optimal play (i.e. a fortress position, etc).

Playing a game is one thing, proving white can win or only draw is something else entirely.

True, you can't confirm if the Queen was sacrificed, or blundered, unless you perform some extensive analysis. But with the aid of engines, such a position becomes correctly evaluated, without the need to search all the lines. Similarly, positions that aren't extremely complicated, can be defended by an engine, even if there's a small edge, for the opposing side.

Playing a game is nothing, a million is something, a billion will tell you interesting things.

[Ozymandias]

As far as claiming chess is a draw, it is only a claim. Nothing even approximating a proof is available now. And never will be. You'd have to show that EVERY possible path from starting position to final position is a forced draw. Not possible.

I don't consider theoretical truth to be the only one of relevance. Even if we can never obtain the proof you mention, from the practical point of view there's already proof enough, that chess is a draw. By the time we get to a 99% draw rate, almost no one will be left in computer chess, to bother about the remaining 1%. It may be the most relevant, for the absolute truth to be achieved, but even though chess isn't just a game, which is why some will persist in the search for that ultimate truth, it's basically a game, which is why it'll be too boring, for the majority to stick around and find out.

[mvk]

You only need to store the number of moves, with optimal play, until a zeroing move. If the 50 move counter in a position plus that number of moves is 50 or more, it's a draw. Otherwise it's whatever result is stored for the position. This is at most seven bits per position, and on average much less.

Why are you writing about pathways? They don't matter.

You are correct. That is how DTC (Thompson), DTZ (?) and DTZ50 (Syzygy) can work at all. There are no path complications needed in the addressing, because the information you need is already available in the value.

[bob]

As far as claiming chess is a draw, it is only a claim. Nothing even approximating a proof is available now. And never will be. You'd have to show that EVERY possible path from starting position to final position is a forced draw. Not possible.

I don't consider theoretical truth to be the only one of relevance. Even if we can never obtain the proof you mention, from the practical point of view there's already proof enough, that chess is a draw. By the time we get to a 99% draw rate, almost no one will be left in computer chess, to bother about the remaining 1%. It may be the most relevant, for the absolute truth to be achieved, but even though chess isn't just a game, which is why some will persist in the search for that ultimate truth, it's basically a game, which is why it'll be too boring, for the majority to stick around and find out.

It would seem not everybody believes it is a draw, else nobody would waste the time playing the game.

"proof" has one meaning only. It establishes a fact, not a guess or a conjecture.. If you leave the word proof/prove out, you can say most anything about the probable outcome of the game of chess. But when you use the word proof, that means an exhaustive search from beginning to end. IE Schaeffer's work with chess (Chinook).

[bob]

The problem is, you can't hand-wave a "bad position" away when you are trying to actually prove something. You can't confirm sacrificing the queen on move 5 is bad until you force checkmate which might not be possible with optimal play (i.e. a fortress position, etc).

Playing a game is one thing, proving white can win or only draw is something else entirely.

True, you can't confirm if the Queen was sacrificed, or blundered, unless you perform some extensive analysis. But with the aid of engines, such a position becomes correctly evaluated, without the need to search all the lines. Similarly, positions that aren't extremely complicated, can be defended by an engine, even if there's a small edge, for the opposing side.

Playing a game is nothing, a million is something, a billion will tell you interesting things.

"correctly evaluated" implies we don't need to go any deeper and we can stop worrying about parallel search and better search methods. We are a LONG way from that point. You can always go deeper, and you can sometimes expose something that was not visible to the last iteration of the search...

[Ozymandias]

It would seem not everybody believes it is a draw, else nobody would waste the time playing the game.

Or they just believe their opponent, to be further from perfect play, than they are.

[Ozymandias]

You can always go deeper, and you can sometimes expose something that was not visible to the last iteration of the search…

My point is that, chess being a draw, those things you find, when searching deeper, won't have any effect on the outcome. Not with the current state of engines, and no unless you stumble upon a complex enough opening position, which your opponent doesn't have stored.