Upgrade to Chess.com Premium!

Centaurs Vs Humans... A forced competition, why?


  • 21 months ago · Quote · #41

    FirebrandX

    miriskra wrote:
    ElKitch wrote:

    Perhaps they find something very innovative and not very expensive? Ah well, not a solution at this time anyway.

    Magnetic-resonance imaging (MRI) could be a solution.

    Still very expensive, and it was later shown to be even less reliable than a polygraph.

  • 21 months ago · Quote · #42

    miriskra

    Very soon we will come to a conclusion that will produce just one single word "brute-force". Are you ready to congratulate an oil oligarch or a drug baron (kingpin) on his world correspondence chess champion title?

  • 21 months ago · Quote · #43

    browni3141

    sapientdust wrote:

    Making a game that rewards humans' strategic strengths and  is much less amenable to the brute-force calculation that computers are so good at was one of the motivations for the game Arimaa.

    It's a really fun game to play, and it can be played with a standard chess set.

    Here's what the creator of the game said about the origin and motivation of the game (from The Creation of Arimaa):

    It was a cold winter day on January 15th 1999. I was watching the kids while my wife went to get the groceries. Aamir was only four and a half years old at the time, but I was trying to teach him how to play Chess. I soon realized that playing a game with the full set of Chess pieces was not yet a possibility, so I started with just the Pawns and King and planned to introduce the other pieces one at a time later. While playing these games with just Pawns and King I realized that even with pieces that had such simple movements there can still be some very interesting games. This brought back some thoughts and memories from 1997 when Garry Kasparov, the world Chess champion was defeated by Deep Blue, a Chess playing computer built by IBM. For some reason after watching that match I really felt sorry for Garry Kasparov because I thought that in a way he was not able to show the depth of his real intelligence over a game of Chess. I felt that Garry had just been out calculated, not really out smarted. I had this feeling that using just the board and pieces of a standard Chess set, it should be possible to make a game which would require the kind of real intelligence that humans possess and computers have not even begun to acquire.

    As one of the top Arimaa players I think I am qualified to estimate that the top human is roughly 400 points stronger than the top bot, if anyone's interested in that. The game is still quite young so it could go either way. You can decide whether or not 400 ELO qualifies as highly computer resistant. Personally I'd like to see a larger gap, and hope to personally increase it Wink

     

    Also, Humans can give handicaps of a horse to bots, which is quite a large advantage. It is very difficult to compare to chess, but I'd say that is like a rook. When I gave this handicap I did take advantage of a specific weakness to win though, so it might not be the best indication of human superiority.

  • 21 months ago · Quote · #44

    LeggomyEigo

    [COMMENT DELETED]
  • 21 months ago · Quote · #45

    FirebrandX

    miriskra wrote:

    Very soon we will come to a conclusion that will produce just one single word "brute-force". Are you ready to congratulate an oil oligarch or a drug baron (kingpin) on his world correspondence chess champion title?

    Actually the opposite effect is happening. The more games that are logged on ICCF, the more prepared the average player can be by using that archive. For example, the Najdorf is so played out to death on ICCF, that I can take that archive and ensure at least a draw even against a super-computer.

  • 21 months ago · Quote · #46

    miriskra

    FirebrandX wrote:

     I can take that archive and ensure at least a draw even against a super-computer.

    You draw against a super-computer with your six years old desktop? Well done FirebrandX. How do you describe super-computer? And how do you know, you play against a super-computer?

  • 17 months ago · Quote · #47

    Alec89

    [COMMENT DELETED]
  • 17 months ago · Quote · #48

    ponz111

    In the  anti ICCF group my question is are books not allowed and if so why?

  • 17 months ago · Quote · #49

    Polar_Bear

    Ponz, your question has been already answered.

    Paper books are indeed allowed & safe as well as online articles written by humans.

  • 17 months ago · Quote · #50

    FirebrandX

    miriskra wrote:
    FirebrandX wrote:

     I can take that archive and ensure at least a draw even against a super-computer.

    You draw against a super-computer with your six years old desktop? Well done FirebrandX. How do you describe super-computer? And how do you know, you play against a super-computer?

    The opening database from the ICCF archive you can create includes every game ever logged on ICCF in modern times. Some of the top CC players have been known to make use of computer clusters or "supercomputers" in order to get the deepest calculation they can. You combine that with the Najdorf being the most popular (by far) opening played on ICCF, and you end up with an opening database you can set to 80 moves per side (120 ply) and it covers every reasonable try by both sides. Thus, not even a supercomputer is going to outperform someone armed with it.

    To this day whenever I see a new win logged in the Najdorf on ICCF, I can simply plug the game into my openings database and spot exactly where one side or the other went wrong. Last time I did it, I found white had lost because he had played 32.Rg1 instead of 32.Rh3, which cost him the endgame dozens of moves later. That's how deep and technical it goes now without making use of that archive, and is also why the Najdorf is a dead opening on ICCF if both players do know to use it.

  • 13 months ago · Quote · #51

    MrSkull

    Here is a suggestion to make it difficult for the engines. Change the rules so that a player may elect not to move on, say, two occasions at any time during the game.  Or maybe have the choice on every move throughout the game.  I am not at all sure whether the computer could list the best moves in order of strength if it does not know whether the opponent is going to choose not to move, and therefore, itself, making consecutive moves. The calculations for each of its candidate moves are based on opponents' responses. Even if the new position is set up afresh each time, it will still need to base its suggested move on the opponent's responses. It would ruin a whole line of calculating, would it not?  Players could still make moves and never "pass", but the computer cannot be sure, and would probably always suggest the best lines as those in which the opponent is standing still.

    Perhaps I am completely wrong.

  • 13 months ago · Quote · #52

    ponz111

    Mr.Skull   Every time a player would pass the other player would increase his advantage. In fact even one pass could lose the game.

  • 13 months ago · Quote · #53

    ponz111

    Here is one good thing about Centaur Chess--it is as close to perfect chess as you can get.  Centaur theory has discovered such things as it is no use trying to play the Ruy Lopez [or the Ponziani] for a win in Centaur Chess because Black can play 1. e4  e5  2. Nf3  Nf6!  [the Petroff Defense] and almost 100% sure of a draw.

    So no good even to play 1. e4 unless you want to try something risky such as the Kings Gambit

    By the way this situation and also the situation with the Nardoff Defense vs The Sicilian point more and more that chess is a draw with correct play.

  • 13 months ago · Quote · #54

    MrSkull

    ponz111

    Mr.Skull   Every time a player would pass the other player would increase his advantage. In fact even one pass could lose the game.

    ponz111     Yes, but a player may never actually choose to pass, while the computer cannot be sure which of the next five or six moves the player will pass, and sometimes make inappropriate "best" choices of moves based on the wrong assumption !  The aim is to make engines untrustworthy because some of their moves are unreliable.

  • 13 months ago · Quote · #55

    ponz111

    MrSkull  your idea will not work.

  • 13 months ago · Quote · #56

    Polar_Bear

    MrSkull, in other words - you suggest some chess variant instead of classic chess in correspondence to avoid engine cheating. It doesn't work. It is too easy to modify existing modern engines if needed.

  • 13 months ago · Quote · #57

    FirebrandX

    ponz111 wrote:

    Here is one good thing about Centaur Chess--it is as close to perfect chess as you can get.  Centaur theory has discovered such things as it is no use trying to play the Ruy Lopez [or the Ponziani] for a win in Centaur Chess because Black can play 1. e4  e5  2. Nf3  Nf6!  [the Petroff Defense] and almost 100% sure of a draw.

    So no good even to play 1. e4 unless you want to try something risky such as the Kings Gambit

    By the way this situation and also the situation with the Nardoff Defense vs The Sicilian point more and more that chess is a draw with correct play.

    To clarify:

    The Petroff by itself doesn't ensure the draw, but combined a well-booked database and at least a quadcore desktop, the chances black can be outplayed in centaur cc are virtually non existent. Thankfully the concept only comes up sparingly in centaur cc games because a majority of the players still like to try to win from both sides of the board. But say if everyone on ICCF started playing the Petroff, you'd see the move 1.e4 get less and less played with each passing year until it becomes an "offbeat" try rather than the most popular move.

    With the Sicilian Najdorf, the situation is a little different. So many games have been played on ICCF in the Najdorf, that one can put together an opening book extracted from the ICCF archives and find that, as complex as the Najdorf is, it's very nearly solved. When I review modern wins in the Najdorf from either side on ICCF, I can follow along in my opening book of ICCF games and always spot the point where one side or the other went wrong. Stuff like:  This person played 32.Rh3, which the book shows a 17% performance, whereas the book also shows 32.Rg1 has a 53% performance. Thus, this person went wrong on move 32. Now that's a very very basic understanding and usage of the book percentages, but you get the idea of how well the Najdorf is covered on ICCF.

    Edit: I see I posted the same point 4 months ago! Oops!

  • 13 months ago · Quote · #58

    ponz111

    Firebrand  you put what I meant much better than I could


Back to Top

Post your reply: