I don't think ouzo removes atoms from the brain. It just changes their arrangement.
True or False Chess is a Draw with Best Play from Both Sides
Sort:
True. Now, please stop talking about this. This is getting tiresome. Move on.
Aug 18, 2020
0
#3504
Thee_Ghostess_Lola wrote:
is Leela a female engine ?...cuz its a girl name uknow.
My robot has a female voice. My wife says she cleans better than a man.
Aug 18, 2020
0
#3505
Thee_Ghostess_Lola wrote:
is Leela a female engine ?...cuz its a girl name uknow.
If you take away its graphics card it plays like a big girl's blouse.
Prometheus_Fuschs wrote:
btickler escribió:
Prometheus_Fuschs wrote:
No, you do not need to traverse all of the game tree to make a proof. It's the only way we know of to prove it but it doesn't mean it's the only possible way we could prove it. Wikipedia explicitly says so...
Define what you think you mean by"game tree" .
You don't have to cover the Shannon number (10^120) in a proof, but you do have to traverse 10^46 positions. Maybe you should read the actual thread on it, which is 6 years old and 343 pages and covers everything WIkipedia does and a lot more.
https://www.chess.com/forum/view/general/will-computers-ever-solve-chess?page=1
Ok, show me the proof that you need to "traverse" all legal positions to solve chess...
It's in the thread I linked. Educate yourself. The current situation is that all technologies in existence and even technologies predicted to be reasonably possible in the far flung future cannot handle solving chess. They are not even able to scratch the surface. If your position is otherwise, the burden of proof is clearly on you and your ilk 
.
GMproposedsolutions wrote:
btickler wrote:
GMproposedsolutions wrote:
10^46 is excessive in my opinion
Luckily, opinion is largely meaningless in this particular area. But hey, thank goodness you came along and at least tried to let the world know that they were looking at this all wrong, and that you could solve the problem, alone, with a donated PC.
That was a feeble attempt of you making a case about anything I wrote. Feel proud of being you as I wouldn't feel comfortable performing your actions. Making logical fallacies and making unwarranted personal attacks is not in me, but good thing we have people like you else what would our existence be like had we stopped wars, stopped all crimes, and treated everyone with respect? It would be a rather dull place.
Translation: "You got me there".
Moving on (from this exchange).
MARattigan wrote:
You're assuming that the only valid method of proof is brute force and ignorance. If you want to prove that the arithmetic mean of two positive real numbers is greater than the geometric mean, a really bad way of attempting it would be to try it out for every pair of positive real numbers.
I didn't say that a chess beginner could solve the game in 100 bytes. I said that he could write a routine to win from any winning king and rook versus king position in less than a couple of hundred bytes. I'll write you a javascript for it if you like.
For this I would have to consider maybe a half dozen generic positions, not the the circa 400,000 positions in the endgame.
You omitted to note that I would estimate less than 1K for the 1MB+ EGTB bishop and knight endgame. I could also write you a javascript for that. Again this would be based on about fifteen properties of the position rather than considering the 11 million individual positions. The compression ratio has jumped from 14 to 1000+ and I expect would continue to increase exponentially for an optimal routine as the number of positions in an endgame increases.
This is not a proof that a conceivably practicable solution of chess on these lines exists, just a conjecture, but @pfren has already pointed out, a solution by EGTBs appears to be completely impracticable simply from a storage point of view. If such a practicable solution does exist, the hard bit is finding it.
Your premise is flawed. It's right there in your first claim: "You have completely solved (Basic Rules) chess if you can produce a routine that guarantees a win from any winning position."
Sure. This would be pretty darn easy to solve if human beings just had a definition of a "winning position" that carries all the way to the starting position in chess ...we could also predict the future of the entire universe, if only we knew the relative velocity and vector of every particle in it and then had a computer that could just crank out the answer instantly, and then store it, and then communicate it to us in some way we could actually absorb. Simple and easy to communicate method for predicting the future. Same basic level of impossible.
We're already doing that (trying to prove what a winning position is), via working backwards from what humans beings and their understanding of mathematics have already determined to be the best "winning position" understanding that we have. You are extrapolating that you can define in a direct rules-related manner what a "winning position" from some opening or middlegame position. But that understanding does not exist, nor can you prove it without going through the same traversing we are talking about. The current method is not entirely brute force. It's already taking into account all the basic rules of chess. Brute force would be checking every position regardless of legality. Can there be some ways of cutting down the traversing farther? Perhaps. But even if you do this so well that you eliminate 999,999 out of every million positions, you are still solving for 10^40. It's not even remotely in the range of possible for the reasonably foreseeable future of humanity.
Go ahead and write your Javascript . You might want to at least switch to assembly language, though. Javascript is hugely inefficient as a high level language, even more so than most compiled languages.
Aug 18, 2020
0
#3509
With like maybe a half dozen rules I think you could define a winning strategy for a rook and king vs a king. No need to have analysed every possible position for the best move or generate a "game tree". I know this is a simple example, but couldn't similar rules be made for different sets of pieces and/or types of positions? I know this is a wild speculation, but might a future computer be able to from the opening position find a set of rules of strategy that forces either a win or draw without examining every position?
Aug 18, 2020
0
#3510
btickler escribió:
Prometheus_Fuschs wrote:
btickler escribió:
Prometheus_Fuschs wrote:
No, you do not need to traverse all of the game tree to make a proof. It's the only way we know of to prove it but it doesn't mean it's the only possible way we could prove it. Wikipedia explicitly says so...
Define what you think you mean by"game tree" .
You don't have to cover the Shannon number (10^120) in a proof, but you do have to traverse 10^46 positions. Maybe you should read the actual thread on it, which is 6 years old and 343 pages and covers everything WIkipedia does and a lot more.
https://www.chess.com/forum/view/general/will-computers-ever-solve-chess?page=1
Ok, show me the proof that you need to "traverse" all legal positions to solve chess...
It's in the thread I linked. Educate yourself. The current situation is that all technologies in existence and even technologies predicted to be reasonably possible in the far flung future cannot handle solving chess. They are not even able to scratch the surface. If your position is otherwise, the burden of proof is clearly on you and your ilk .
Lol, yeah sure I'll sweep through dozens upon dozens of thread pages just to find out there is no proof, then again, you don't even seem to understand what you are discussing.
exactly Ziggy. one doesnt hafta do a exhaustive compilation. u can stop once the beautiful, remarkable, & boneheaded human brain has correctly confirmed a winner. like u say...simplify by combining silicon+organic. take 7-piece TB's e.g. there are abt 4.23x10^14 unique and legal positions (in a abt 18 terabytes).
now. eliminate the dubious & blunderful openings that rely on opp error to convert a point. of which there are many...many. it ends up boiling to a light mix of openings and a not-so-light bevy of middlegames. so. applying some common sense w/ a brain box may produce a #.
philosophically, does this # exist ?...yes. trust me its out there. s/w. so. if we find it can we comprehend it ?...not if we use IM Pfren's brain atoms count azza reference we wont. but if we use a few a Avacados #'s its possible...i think (blink-blink).
Arthur a child could write a strategy how to win with R and K vs K.
But as you learn more about chess writing such programs is not so easy with more pieces on the board.
A child could also write a strategy on how to DRAW with a lone K vs K and B and P with the bishop and pawn protected. [by the way this type of endgame where Black might force a draw in such an endgame is one more piece of evidence that chess is a draw.]
MARattigan wrote:
pfren wrote:
...
Chess will never be mathematically solved, as the universe does not have enough atoms to store the data of the solution.
I think you're assuming an EGTB type solution is the only possible mathematical solution.
I think I could write a routine in 1 KB for example that would do the same job as the 450 KB KBBK Nalimov EGTB (faster).
The Nalimov EGTBs in any case are overkill. You have completely solved (Basic Rules) chess if you can produce a routine that guarantees a win from any winning position. It doesn't have to be in the minimum number of moves and it doesn't need to produce more than one move for a given position.
A chess beginner with an aptitude for programming could probably produce a routine to replace the 14 KB KRK Nalimov EGTB in less than a couple of hundred bytes if you drop that overkill. I could also replace the 1 MB + KBNK EGTB in around 1 KB with the same proviso.
Such EGTB replacement routines are no doubt possible for all generic endgames (including all 32 pieces). Moreover I think the ratio of the size of such routines as a fraction of the corresponding EGTB size would probably exponentially decrease with EGTB size.
The problem is, of course, that the larger the EGTB size, the harder is the analysis for such replacement routines, but this is not the same as saying that such routines do not exist.
If storage capacities continue to grow at their historical rate, I wouldn't be surprised if in a decade or two, a complete solution to chess would fit on the average person's mobile. I would be surprised if it actually did.
He's right .... back around 1990-91 I was experimenting with writing code in the smallest space possible that would work as fast as possible. I designed many different varieties of prime number generators, doing it completely from scratch and without deliberately using previously designed techniques so I didn't fall into traps. Then I went to university and went through the first year of a computing degree course before switching to philosophy, which was more fun.
The programs I produced were probably about 40 times more compact than the standard top-down programming techniques we were taught. I think they also ran around five times faster. I used to time them with a stopwatch.
Not true
Aug 18, 2020
0
#3515
btickler wrote:
MARattigan wrote:
You're assuming that the only valid method of proof is brute force and ignorance. If you want to prove that the arithmetic mean of two positive real numbers is greater than the geometric mean, a really bad way of attempting it would be to try it out for every pair of positive real numbers.
I didn't say that a chess beginner could solve the game in 100 bytes. I said that he could write a routine to win from any winning king and rook versus king position in less than a couple of hundred bytes. I'll write you a javascript for it if you like.
For this I would have to consider maybe a half dozen generic positions, not the the circa 400,000 positions in the endgame.
You omitted to note that I would estimate less than 1K for the 1MB+ EGTB bishop and knight endgame. I could also write you a javascript for that. Again this would be based on about fifteen properties of the position rather than considering the 11 million individual positions. The compression ratio has jumped from 14 to 1000+ and I expect would continue to increase exponentially for an optimal routine as the number of positions in an endgame increases.
This is not a proof that a conceivably practicable solution of chess on these lines exists, just a conjecture, but @pfren has already pointed out, a solution by EGTBs appears to be completely impracticable simply from a storage point of view. If such a practicable solution does exist, the hard bit is finding it.
Your premise is flawed. It's right there in your first claim: "You have completely solved (Basic Rules) chess if you can produce a routine that guarantees a win from any winning position."
Sure. This would be pretty darn easy to solve if human beings just had a definition of a "winning position" that carries all the way to the starting position in chess ...
--- It's your response that is flawed. We have a definition of what is a winning position (actually two because the FIDE rules define two distinct games). In neither case is it pretty darn easy to solve.
We're already doing that (trying to prove what a winning position is), via working backwards from what humans beings and their understanding of mathematics have already determined to be the best "winning position" understanding that we have. You are extrapolating that you can define in a direct rules-related manner what a "winning position" from some opening or middlegame position.
--- I'm not extrapolating that I could do that. I'm conjecturing only that such routines exist in a form that might fit on computer storage not too remote from what is currently feasible. That such rules exist is obvious, the question is only, "what is the minimum size?". As I said, finding small routines is non-trivial. (Openings and middle games are just types of endgame.)
But that understanding does not exist, nor can you prove it without going through the same traversing we are talking about. The current method is not entirely brute force. It's already taking into account all the basic rules of chess. Brute force would be checking every position regardless of legality.
--- The understanding doesn't currently exist for all endgames. That is not to say that relatively compact rules do not exist.
Where the (human) understanding does exist, it doesn't consist of traversing large numbers of positions.
I'd still regard EGTB generation as a BFI routine. The fact that many (not by any means all) illegal positions are excluded doesn't really change that.
Can there be some ways of cutting down the traversing farther? Perhaps. But even if you do this so well that you eliminate 999,999 out of every million positions, you are still solving for 10^40. It's not even remotely in the range of possible for the reasonably foreseeable future of humanity.
--- As I pointed out, writing a routine for KRK would be very straightforward. It would solve the 400,000 or so positions in that endgame. No specific positions need be included in the routine (not even the draws - the only requirement is that it wins from winning positions).
Go ahead and write your Javascript . You might want to at least switch to assembly language, though. Javascript is hugely inefficient as a high level language, even more so than most compiled languages.
--- Most of the programs I've written have been in IBM 360 Assembler and its descendants and my storage estimates were based on that. I offered Javascript as easy to translate into whatever takes your fancy. I doubt if a user would actually notice any difference in performance.
"Solving chess" doesn't necessitate finding the solution of every possible chess position. Logically, it necessitates an accurate and efficient categorising of potential positions within a tree branch as more complex examples of a generic that is already "solved". This should reduce the number of positions that need to be solved by a factor of maybe a thousand, rising to probably several millions if all trivial examples are included.
I don't think there is sufficient understanding here of the banality of brute force in this context, together with the fact that there will be many generics and sub-generics that can be dismissed.
If chess isn't to be a draw, there would have to be positions that came about through normal, fairly aggressive and pointed lines of play that could be extended to produce a zugzwang in all possible variations. The idea of that is ridiculous, since the procedure, working one way, can obviously be reversed. Thinking chess could be a forced win with optimum play isn't thinking clearly.
Aug 18, 2020
0
#3517
Optimissed wrote:
...
If chess isn't to be a draw, there would have to be positions that came about through normal, fairly aggressive and pointed lines of play that could be extended to produce a zugzwang in all possible variations. The idea of that is ridiculous, since the procedure, working one way, can obviously be reversed. Thinking chess could be a forced win with optimum play isn't thinking clearly.
I don't understand what you mean by reversing the procedure here. Obviously moves like QxQ can't be reversed under normal playing rules. Can you clarify?
Why would your reasoning apply to the initial chess position but not the positions in posts #3523 and #3532?
In general, if one side can lose several moves in a complex and unclear but sharp position to produce a zugzwang, the other side can do that too and since that's the only way to win (hypothetically) it can't be done.
3523 etc are irrelevant since they can't be forced into.
Even using the false assumption that each of these 10^46 positions can be stored at just one atom, it is still way too much to get the available storage for solving chess. The number of atoms in the Universe according to the latest evaluations do not excheed 10^82, and it is more than apparent that we cannot take a short ride to gazillions of these atoms.
i have abt 7 quintillion-billion atoms in my brain. i have no idea how many u have left. tho no dbt ouzo has probably scorched plenty a many.