Slashdot Mirror


Pentago Is a First-Player Win

First time accepted submitter jwpeterson writes "Like chess and go, pentago is a two player, deterministic, perfect knowledge, zero sum game: there is no random or hidden state, and the goal of the two players is to make the other player lose (or at least tie). Unlike chess and go, pentago is small enough for a computer to play perfectly: with symmetries removed, there are a mere 3,009,081,623,421,558 (3e15) possible positions. Thus, with the help of several hours on 98304 threads of Edison, a Cray supercomputer at NERSC, pentago is now strongly solved. 'Strongly' means that perfect play is efficiently computable for any position. For example, the first player wins."

28 of 136 comments (clear)

  1. Comparison to Chess? by jratcliffe · · Score: 4, Interesting

    Out of curiousity, does anybody know what the number for chess that compares to the 3e15 number for pentago is? In other words, how much "bigger" is chess?

    1. Re:Comparison to Chess? by vux984 · · Score: 5, Informative

      chess
      state space complexity 10^47

      go
      9x9 - 10^38
      13x13 - 10^79
      19x19 - 10^171

    2. Re:Comparison to Chess? by iggymanz · · Score: 2, Funny

      woman
      state space complexity: 10^191
      23 out of 28 days

      three orders of magnitude added 5 out of 28 days

                                                                                     

    3. Re:Comparison to Chess? by jellomizer · · Score: 2

      Well lets find out.
      1... 2... 3... 4... 5...

      --
      If something is so important that you feel the need to post it on the internet... It probably isn't that important.
    4. Re:Comparison to Chess? by Anonymous Coward · · Score: 4, Informative

      According to the summary, 3e15 is the number of possible positions. The number of possible positions in chess is around 4e46.

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

    5. Re:Comparison to Chess? by noh8rz10 · · Score: 3, Funny

      keep it classy man. ha ha, women. double ha ha, menses. and yet you're still single, how can that be?

    6. Re:Comparison to Chess? by ledow · · Score: 2

      I was lucky enough to have a lecturer at university who was actively working on solving Go. Professor Wilfred Hodges at QMW, University of London.

      It was his talk about the complexity of the game of Go on my induction day that convinced me to go to that university, and I was able to have him as a course lecturer for certain related courses later on (graph theory).

      He is a typical, mad-haired, Einsteinian-looking sandals-in-winter professor, and he gave a marvellous intro to Go, complexity and the work he was doing.

      All everyone else took away was that he showed photos of himself at a Go convention (on 35mm film). But I thought it was brilliant, and it made me teach myself Go.

      I hope he's still working on it. Judging by the website I've found for him, he's busy with a LOT of other areas of mathematics than Game Theory, though.

      But I'll never forget the mental "Woah" I got when he explained how much more complicated 19x19 Go was than Chess, even though the rules are vastly simpler.

      Apparently there is no decent Go computer player in the world that can beat more than an average Go human player.

    7. Re:Comparison to Chess? by Kjella · · Score: 2, Insightful

      three orders of magnitude added 5 out of 28 days

      Somehow 10^194 doesn't seem significantly worse than 10^191.

      --
      Live today, because you never know what tomorrow brings
    8. Re:Comparison to Chess? by Laxori666 · · Score: 2

      Supplicating white knight is also being quite manipulative. What, you think if you're nice you deserve to get a chick?

    9. Re:Comparison to Chess? by bluefoxlucid · · Score: 2, Interesting

      Go on 9x9 is very similar to Chess neurologically: it's a tactics game. I'm bad at it. I hate chess for emotional-political reasons: it is technically inferior to Go, being that it is tactical, and thus I cannot stomach nor can I support wasting my time on such an inefficient pursuit. Someone has been reasoning, correctly, that I should practice Chess to gain additional, non-domain tactical exercise so as to improve my Go playing. I have been, incorrectly, resistant.

      Go in 13x13 is strategically different to Go in 19x19. On a 13x13 board, abstract strategy and tactics both apply simultaneously: control of a corner exerts strong influence over another corner, and you can develop very quickly. Strategic moves must also be tactical.

      Go in 19x19 is, early on, strategic: the stretch of influence is too big, and only vague tactical considerations are helpful because of too much variation. Early play accounts for strategy alone: it accounts for providing a position that has strong but non-specific tactical value for all variations. As the board develops, tactical situations arise: the strategy employed provides a variety of tactical responses to tactical approaches. This exercises, both separately and conjointly, the full utilization of abstract strategic thinking and logical tactical computation.

      Thus Go 9x9 and Chess are both tactical calculation; Go 13x13 is strategic-tactical or tactical depending on position, weakly exercising abstract reasoning and more strongly exercising tactical calculation; and Go 19x19 exercises the full breadth of abstract strategic thinking, blended thinking (including feedback looping tactics into strategic impact and strategy into tactical options), and direct tactical thinking (when the immediate position is only a win-lose proposition, not a resultant strategic position proposition).

      I am better at Go 13x13 because I can cover the weaknesses in my abilities while abusing the weaknesses in my opponent's abilities. It is an easy game for me because I start with the option of using whichever mode of thinking I'm better at in the current position, and retain that until it's too late to really swing the game. 9x9 relies on tactical calculation, while 19x19 relies on effective use of the greatest breadth of mental skills based on what's on the board more than what I feel like I can handle at the time; I like 19x19 because it forces me to grow and learn.

      Computational analysis is not relevant for wide-play of Go because it's impossible and there are other, more abstract ways to do this. There are even computer algorithms to analyze influence and use this to analyze strategic strengths and weaknesses, then analyze those areas of the board to analyze the strategic position. Humans are better at it, but direct computation isn't the only way to approach the problem.

    10. Re:Comparison to Chess? by MrBigInThePants · · Score: 3, Informative

      Why is this insightful? He doesn't even understand the word "complexity" in the context it was used!?

      GO IS complex. Yes, the brute force complexity is stupidly large.

      But also the (non-optimal) pattern strategies that humans use and researchers ARE STUDYING (and have done for some time) are very difficult to emulate.

      I love how you just dismiss a whole field of research as if they had not bother to try anything else in the decades this has been study?!

      Sheesh...talk about failing to give people the benefit of the doubt...

    11. Re:Comparison to Chess? by Spy+Handler · · Score: 2

      A simple neural net (I hate that fucking term) algorithm trained against average Go-playing humans will end up being average at playing Go.

      Seems to me like you took a cursory glance at Go and decided it'd be simple to create an algorithm that would be good at it. Without any actual experience.

      Have you actually played it? More importantly, have you ever seen a decent player play against a computer?

      My dad is an amateur and is not ranked, in chess terms he's not even an expert and basically he's nothing. He stomps the living crap out of the best Go software commercially available; he wins every time. And from reading articles about the state of the art in computer Go programming, it would probably be a winning bet to say that my dad (a barely decent player) could beat any computer program ever made.

    12. Re:Comparison to Chess? by krkhan · · Score: 2

      A machine operating at the rate of one variation per micro-second would require over 1090 years to calculate the first move

      I'm guessing that we have gotten a little faster since then with our current peta and soon to be exascale machines... micro-second? An eternity. Recalculate that spreadsheet, professor.

      Shannon -- the "professor" -- was simply taking into account the technology available at the time.

      Hans-Joachim Bremermann has also made an interesting argument:

      "Speed, memory, and processing capacity of any possible future computer equipment are limited by specific physical barriers: the light barrier, the quantum barrier, and the thermodynamical barrier. These limitations imply, for example, that no computer, however constructed, will ever be able to examine the entire tree of possible move sequences of the game of chess.

    13. Re:Comparison to Chess? by Vintermann · · Score: 2

      Go is a perimeter game and is really no more complex than Othello. A simple neural net (I hate that fucking term) algorithm trained against average Go-playing humans will end up being average at playing Go.

      How impressive that you can know that without having tried. How do I know you haven't tried it? Because it's totally wrong...

      --
      xkcd is not in the sudoers file. This incident will be reported.
    14. Re:Comparison to Chess? by nojayuk · · Score: 2

      As far as I know most worthwhile computer-Go programs don't use alpha-beta for the reasons you stipulate. A lot of the in-game scoring involves trying to calculate territorial control and areas of influence, moyo in Japanese. Like chess, at the beginning of a game computer programs tend to use standard openings (joseki) and as in chess an experienced human player will break out of the book early on in the game to force the computer to start evaluating moves instead. The end-game, like chess simplifies down as there are many fewer possibile moves on each turn and ownership of large areas of the board have already been decided or conceded at which point calculating a full solution to the end is possible. Its the tricky bit in the middle that's the fun part (especially ko fights).

      Small-board Go of 7x7 size has a small enough number of unique positions that its gamespace could be exhaustively evaluated today given the ubiquity of rentable cloud computing and cheap storage to hold the results. All it would cost is time and money. Creating a good Go-playing computer program which can evaluate positions and choose moves wisely is more tricky.

  2. Chess by Anonymous Coward · · Score: 5, Funny

    After playing in chess tournaments for 20 years, I have strongly solved that chess is a forced win for any player facing me.

  3. A rough approx. is 10^60 moves by Anonymous Coward · · Score: 2, Interesting

    Assuming the game goes for at least 30 moves, and that each player has roughly 10 options per move you get 10^(2*30). 10 options times 30 moves * 2 (there are two players, so two moves per "move").

    1. Re:A rough approx. is 10^60 moves by Impy+the+Impiuos+Imp · · Score: 2

      IIRC, if the entire universe were crammed to neutron star density with computer processing, it still couldn't touch a perfect game in the entire life of the universe.

      --
      (-1: Post disagrees with my already-settled worldview) is not a valid mod option.
    2. Re:A rough approx. is 10^60 moves by Kyont · · Score: 3, Funny

      Plus, where would you put the chess board?

      --
      You shall see a cow on the roof of a cotton house.
  4. Re:different than tic tac toe or connect 4? by mrchaotica · · Score: 4, Informative

    No, tic-tac-toe is always a tie.

    --

    "[Regarding the 'cloud,'] ownership was what made America different than Russia." -- Woz

  5. Re:different than tic tac toe or connect 4? by i+kan+reed · · Score: 2

    All these posts have the caveat of "with perfect play"

    Human play is about creating mental shortcuts that create useful heuristics for "winningness", and developing those shortcuts is most of the fun to be had.

  6. Re:different than tic tac toe or connect 4? by Anonymous Coward · · Score: 2, Insightful

    If you are always losing as the second person in tic-tac-toe, you might want to lay off the drugs or stop posting on Slashdot as your IQ is too low.

  7. Re:Grammar? by zerosomething · · Score: 5, Funny

    Can't the editors write a headline that meets the basic rules of grammar? How about "In the game of Pentago the first player can always win", or "Pentago is strongly solved".

    No cause with out those grammar mistakes their would be 30 pricent fuer com-mints on /.

    --
    It all starts at 0
  8. Wargames... by Rhaban · · Score: 5, Funny

    If Matthew Broderick had played pentago, the computer would have concluded the first country launching a nuclear missile always wins the war.
    Il came close

  9. Re:Grammar? by Anonymous Coward · · Score: 5, Funny

    No, now that Pentago is solved, we're reduced to online games of Pedant.

    HINT: Last player always wins.

  10. Re:Grammar? by noh8rz10 · · Score: 2

    I thought it meant, the game of Pentago is a really awesome first person shooter. or other game from the first person perspective. You know, the opposite of "Pentago Is a First-Player FAIL"

  11. Re:This is why I don't play deterministic games by bluefoxlucid · · Score: 2

    Again Go proves superior: Memorizing Go openings is a start. Learning why they work the way they do is required.

    In Go, your opening is strong for a reason. Somebody plops a stone in the middle of it, or does an approach out of turn, you have an advantage or at least are no worse off than if you played a standard opening. Fast players just play standard openings and get into midgame--Koreans like to do this. The Japanese like to analyze their opponent, the board, and form a long-term strategy based on standard openings, often not even following the standard more than 2 or 3 moves. Standard openings like Ni Ren Sei and San Ren Sei are only a few moves (Ni Ren Sei is black on two adjacent corners, white on two adjacent corners, black's move--which is usually approach, or strengthen, sometimes split).

    In Go, your opponent can do more than just move a piece toward you in a bumbling and foolish manner. He can jump right into the middle of your opening and screw it up. Mid-level players do this a lot: a slightly stronger opponent will disrupt a weaker opponent because he can make a position off the disruption, and the weaker opponent does not know why he is playing the Low Chinese other than because it is a correct opening--and thus makes a really screwy position responding poorly.

    Go is about concepts. Memorizing Joseki is often advertised as a good way to start; it is a poor way to continue. Joseki study is done to learn why: to learn about variations, about how a joseki is strong, about why a korean Jung-Sek is formed differently from a similar looking Japanese Joseki--what was the goal?

    Life and Death is often by rote... for the simple shapes. Vulnerable points. Recognizing the shapes before they come.

  12. Connect Four by multimediavt · · Score: 2

    Connect Four was the same way. Whoever went first wins. Didn't take a supercomputer to figure that out, either. Once you did figure that out, though, it pretty much made playing that game pointless. Up in the back of the closet it went. Something tells me Pentago will be joining it, soon.