29

Imagine a six-sided die, D6, the right size to exactly occupy a square on a chessboard.

The die can move to any adjacent square, but does so by rolling rather than sliding, so the topmost side of the die will show a different value.

Now suppose the chessboard is infinite in every direction: north, south, east & west. And there is a constraint: the die must at no point show a 6 on top. 1, 2, 3, 4 & 5 are all ok.

Subject to this constraint, can you define a sequence of moves for the rolling die across the infinite chessboard, so that each square of the board is occupied exactly once?

EDIT: have added my own suggested solution below.

EDIT 2: have awarded the legendary green tick to what I consider to be the best-explained answer. A clear picture that didn’t confuse people is a big part of this. I excluded my own (very different) answer from contention, although I do like it! Thanks all!

Laska
  • 1,919
  • 10
  • 22

4 Answers4

17

Here is I believe a simple way of demonstrating that

it is possible.

First note that it is easy to make a straight 3 squares wide line (let's call it a street): For example, moving N start with the 6 facing W. Move WWNEENWWNEE.. The six will face WdEEdWWdEEdW.. between these moves (d for "down").

At some point we will need to turn 90°: Picking up where we dropped the previous example at the E edge of the street this can be achieved: NNNWWSESWW . The 6 will be WWWWdEEdNNN. This turns our N-bound street to the W and does so traversing a clean 3x3 square. Conveniently, we can use the same sequence backwards in time and mirrored in space if we want to achieve the same turn while happening to be on the other edge of the street at the moment we want to turn. Indeed, N-bound on the W edge: NNESENNWWW turns W with the 6 facing EEEdNNdSSSS.

With these building blocks thinking in terms of

3-by-3 blocks

we have transformed the original problem into the same without the six constraint. Indeed, we can move a

3-by-3 square freely forward, left or right.

Without this constraint the problem is trivial, for example, anything

"spirally"

will do.

Picture:

enter image description here

loopy walt
  • 21,184
  • 33
  • 95
  • Any chance of a pic? I promise when I post my own solution, I will show it graphically – Laska Mar 05 '23 at 10:47
  • This seems to be essentially the same as AxiomaticSystem's answer. Their image shows such a spiral. – Daniel Mathias Mar 05 '23 at 12:27
  • 1
    @DanielMathias I can see your point. Though I'd argue the spiral bit is the most obvious part of the construction. I've edited the answer to reflect this. (Also, the spirals are not the same. Mine has long straights and the occasional left (or right, but always the same) turn, theirs is arguably more convoluted to the point of being non obvious.) – loopy walt Mar 05 '23 at 15:36
  • 1
    @Laska Since you are asking so nicely. – loopy walt Mar 05 '23 at 18:23
  • @loopy wait Thanks very much for the clear diagram. However I don’t think it can work because of parity. Must enter and leave a 3x3 square by a corner, because there are 5 light squares and 4 dark squares – Laska Mar 06 '23 at 00:14
  • 2
    @Laska If you look at the small arrows attached to the small squares you'll find that that is what they do. For example, the bottom right 3x3 is entered at the bottom left corner and left at the top left corner. – loopy walt Mar 06 '23 at 00:34
  • @loopy wait. Ok I see that’s sound - sorry I had not expanded the image on my phone. Your turning 3x3 is the same as AxiomaticSystem’s but you also have a straight-ahead 3x3. It looks as if you can define a recursive 9x9 made of 3x3. This could be a different non-spiral way to cover the plane – Laska Mar 06 '23 at 01:56
  • 1
    @Laska I thought of that but in order to make sure it expands in all directions you'd have to do something like cycle trough orientations which is kind of spiralling. – loopy walt Mar 06 '23 at 02:27
8

If we start with 1 facing up, our single movement option is to tilt, roll some distance with 2, 3, 4, and 5, then tilt in the opposite direction to go back to 1. This lets us think of a path as consisting of U-shaped segments with length-1 legs.

These are sufficient, as evidenced by this jagged spiral consisting of a starting section and a repeating 3×3 block: enter image description here
The white section is the initial block and red cells indicate places where 1 is up.

AxiomaticSystem
  • 9,864
  • 16
  • 38
  • 1
    Definitely agree with the characterization as U segments. Finding it harder to see the algorithm for orientation of the 3x3 blocks, and why this must create a spiral. If this is a solution as it seems to be, it’s very different from mine! – Laska Mar 04 '23 at 14:12
  • Each 3x3 block turns the path left or right. This reminds me of fractal dragon curves: https://jurassic-pedia.com/dragon-curve-cn/ but this one fills the plane with no attempt to make a spiral – Laska Mar 04 '23 at 14:34
  • 15
    I'm struggling to make sense of the graphic. What do the green/black/grey square denote? – fljx Mar 04 '23 at 16:10
  • 8
    @fljx This image should make it clear. – Daniel Mathias Mar 04 '23 at 20:29
  • 4
    To show the path clearly, each square is represented as 2x2 shape. It would be good if the answer can explain this – Laska Mar 04 '23 at 21:33
  • 13
    The graphic can definitely be improved—it's very confusing right now. – Greg Martin Mar 04 '23 at 21:50
  • The 3x3 block here (represented as 6x6 for clarity of seeing the path) can be freely placed in one of two orientations: one turns the path left and the other turns it right. I guess the algorithm chosen to make the spiral is always turn left if you can. “Obvious but not obvious” that this fills the plane – Laska Mar 04 '23 at 23:57
  • 1
    Why can't the arrows cross the boundary of the two squares that the die is rolling too and from? Limitation of the software used? – Lamar Latrell Mar 06 '23 at 05:52
6

Is it possible?

Yes.

Why?

The die could roll on three possible "axes": 3146, 2156 and 5432. You can roll on 5432 in one direction, then turn so that 1 and not 6 would be on top without going the same direction twice in a row right after rolling 1, and then these are possible:

enter image description here

The bottom right shape shows that it's possible (the colors on the cells are the 5-4-3-2 blocks starting from the red one). Just extend it indefinitely.

Nautilus
  • 6,534
  • 10
  • 32
  • 2
    Edited my answer because there are multiple "net directions" to arrive at the same number on top. – Nautilus Mar 04 '23 at 15:31
  • 1
    It’s not entirely obvious to me how this pattern would extend indefinitely. Can you show some more cells please – Laska Mar 05 '23 at 00:09
1

Here's my own solution:

double spiral

Just keep spiraling outwards, adding two "meccano pieces" of length 2k (together with the connectors) at each stage.

If one tries the same simple-minded approach with a single spiral, there seems to be a parity issue, leaving at least one square unvisited by the path.

Enjoy!

Laska
  • 1,919
  • 10
  • 22
  • could you explain how you make both spirals ? Maybe I misunderstood the solution, but I don't see how you pass from one spiral to the other, and so only half the squares are occupied. – Alois Christen Mar 08 '23 at 11:42
  • From the green square showing 1, one can roll north to show 2,3,4 or 5, and that begins one of the spirals. Or one can roll south to begin the other spiral. It's still a single path. The distance between any two specific squares will always be finite – Laska Mar 08 '23 at 12:37
  • Your question asks for a sequence of moves, which can only traverse a path in one direction, but your path is unbounded in two directions. Your comment says it best: you can visit all of the yellow squares or all of the white squares. You cannot visit all of both while following the path you propose. – Daniel Mathias Mar 15 '23 at 20:02
  • I asked specifically for a sequence of moves, but I didn’t ask that this start at one point. I wanted to allow the possibility that the sequence be infinite at both ends – Laska Mar 16 '23 at 10:39