connect6, is it a draw with perfect play? Gomoku, Connect6

14 replies. Last post: 2020-12-05

Reply to this topic Return to forum

connect6, is it a draw with perfect play?
  • lguser at 2020-07-09

    connect6 has a special rule which is meant to balance playing first and second.

    has connect6 gotten to the point where players are reasonably sure that a certain outcome would

    occur given perfect play? if so what is it?

    optimally, that outcome would be a draw, because it would mean that connect6 achieves its goal of balancing the two colors.

    but i could very well see it being an advantage for either color.

  • lazyplayer at 2020-08-18

  • metzgerism at 2020-08-18

    Very likely. A draw can be achieved with <40% of the board being covered in nets (this is how you win at Gomoku Pro 9x9), and functional/tactical draws can occur sooner than that. The borders start to come into play after 20-25 turns.

    Building a net is NOT easy, though, and reading the board is tricky enough that draws are exceptionally rare…I can usually get really good players into a long dogfight, but have brainfarted and lost when the board gets really covered, around move 75.

  • add3993 at 2020-08-19

    Does anyone foresee a Renju-like set of carefully defined constraints to balance the game?  I imagine the designer's goal was partly to avoid the need for this, but is it in the game's best interest for expert play?

    More generally, I am interested in understanding the relationships between the GoMoku (and GoMoku Pro), Renju, and Connect6 player communities.  How do you feel about Renju, and has there ever been a move to include (some version of) it here?

  • lazyplayer at 2020-08-19

    add3993, if they settled on a single variant, with some kind of swap rule and without double stones like connect6 (I really dislike the idea), then it could be such a successful game. The simplicity of the rules is impressive and it's also easy to read ahead because there are no captures…

    In fact hex can be considered as a gomoku variant, except that instead of winning by placing 5 in a row, you win by occuping a string of cells connecting your two sides. It's actually the same game and the only difference is the shape of the winning sets of stones.

  • metzgerism at 2020-08-19

    I'd be surprised if there was another Connect6 player community, besides any possible in-person play in Taiwan. The BGA implementation in beta might change that.

    My hunch is there's a first-player advantage in Connect6 - there are guaranteed losing moves for white on turn 1, and black reaches the minimum forced victory and forced draw triggers first. That said, if there is an advantage, it's much, much smaller than Gomoku, where perfect play results in a guaranteed win. It's very likely still smaller than the advantage black gets in Renju as well…material advantage matters, and Black gains it back quickly enough that they can snatch it back with good early defense.

    Connect6 has a comparatively large game-tree complexity from the start, while Gomoku…doesn't. It's a lot easier to internalize a guaranteed winning tactic when there's only 25 stones on the board than when there's 50 or 100…and we still don't know if first-turn advantage even exists, like I said it's just a hunch.

    Personally, I wouldn't want to see Renju-style rules implemented in Connect6 if we do discover the game isn't quite fair. That said, I think highly competitive play might eventually benefit from the pie rule (pushing black to start towards the edges) and the GomokuPro anti-draw rule on a 20x20 board (where white reaches that trigger first).

  • lazyplayer at 2020-08-19

    Here is what ondik (the highest rated human player here?) said a few months ago in this forum:

    It’s definitely not balanced on the highest level. Currently there are around 3-4 openings which are playable for white. Other than that, everything is a significant black advantage surewin. And even with those 3-4 openings, I’m quite sure most of the top players prefer to play black. Last 2 championship: 46 black wins 25 white wins, 1 draw. So much for the “balance”. Either way, it’s boring to have just a few variants to play.

  • metzgerism at 2020-08-19

    You had mentioned “it's probably best to go back to Gomoku with swap rule,” and I disagree with that…I very much prefer playing two stones at a time - it's a lot more fun and feels a lot more dynamic. I just have a better time playing Connect6 than Gomoku.

    Yeah, maybe the three changes I suggested above for Championships. I'm finally at the point where I'd also like to see larger boards.

    Alternately, you could allow black to place 2 stones every turn, and white plays 3 stones on their first turn (only), or white may play 3 stones on a single turn of their choosing. It could just spin the advantage completely to white, but it's not obvious to me that this is the case.

  • lazyplayer at 2020-08-20

    You can play 3 stones per turn, no problem for me, if I ever play a gomoku-style game, it'll be without double stones. I'm so opposed to them because I think they emphasize the worst aspect of the game, the fact that a single wrong choice can have so drastic consequences…

  • metzgerism at 2020-08-20

    Hey, different strokes :)

    That said, it's going to be hard to find an N-in-a-row that doesn't ruthlessly penalize errors at any stage of the game…maybe YINSH or Pente?

    Make a really bad move in Connect6, Connect4, Gomoku, Renju, GomokuPro, Yavalath…you're either just about to lose or circling the drain. I find it a little more interesting in Connect6 because a winning position usually has to be carefully planned out, properly wheeled around center, and not run out of steam. You also have to make sure that any of your opponent's blocking moves don't make an threat that pauses your attack structure, and there's often 3 options for countering double-threats (instead of 1-2 in Gomoku).

  • add3993 at 2020-08-20

    How about, (k,m)-Gomoku is the game where the goal is to be the first player to form k *mutually disjoint* lines each of length k (not sure whether exactly k or at least k is preferable).

    (3, 5)-Gomoku could be a reasonable thing to try.  The hope is exactly that (1) the first-turn advantage is lessened, (2) mistakes are less fatal.

    By the way, there is a huge monograph by József Beck called “Combinatorial Games: Tic-Tac-Toe Theory” which studies games of this kind in considerable generality (encompassing Hex as well).  It's quite possible what I'm proposing is in there or the literature already.

  • add3993 at 2020-08-20

    Sorry, above “each of length m” not length k.  Ordinary Gomoku = (1, 5)-Gomoku.

  • metzgerism at 2020-08-20

    One of the highest rated n-in-a-row games (YINSH) has this goal, but the gameplay is very different. I haven't gotten a chance to play it yet, unfortunately.

    When you achieve a 5-in-a-row, you remove one of your (initially 5) rings from the board. The rings give you flexibility of movement on your turn, so as you approach the goal (remove 3 of your rings), you get more hamstrung. Also IIRC, making an n-in-a-row usually doesn't allow your opponent to make one immediately afterwards except in rare circumstances.

    The bigger problem is that this would be awfully drawish for basic Gomoku, and exceptionally drawish for Connect6. The grid on the board doesn't guarantee a winner for anything bigger than 4-in-a-row.

  • crazyvvi at 2020-12-05


Return to forum

Reply to this topic