3
$\begingroup$

When I was a child, my mother taught me a simple pencil-and-paper game. I would like to know whether this game, or an equivalent formulation of it, has already been studied.
Let the integer $n > 1$ be given. Then, consider the following two-player impartial game played on a finite connected region $R \subset \mathbb{R}^2$, consisting of a union of unit squares (cells). The outer boundary of $R$ is fixed.

At each move, a player adds a unit-length edge along a side shared by two adjacent cells inside $R$ (i.e., a single horizontal or vertical segment of the square lattice).
If this move disconnects the remaining free region into two connected components $R_a$ and $R_b$, then the player immediately captures the smaller component (i.e., the one containing $\min\{|R_a|,|R_b|\}$ cells; in case of equality, the player may choose either component), and that component is removed from the playable region.
The game ends when no further disconnection is possible. The score of each player is given by the total number of cells they have captured, and the winner is the player with the larger score.

I am particularly interested in the case where $R(n)$ is the diamond-shaped polyomino consisting of $2 \cdot n^2 - 2 \cdot n + 1$ unit squares arranged in rows of lengths $1, 3, 5, \ldots, 2 \cdot n - 3, 2 \cdot n - 1, 2 \cdot n - 3, \ldots, 5, 3, 1$ (see the figure below for the case $R(4)$, shown in an intermediate stage of a game). enter image description here

Question(s):

  • Has this game, or an equivalent formulation of it, already been discussed in the literature?
  • If not, what can be said about optimal play for each given $R(n)$? In particular, I would like to know which player wins, for each given $n$, under perfect play. For example, a draw clearly occurs for $n=2$, while I strongly suspect that Player II can force a win for $n=3$, possibly by a score of $8$-$4$.
$\endgroup$
11
  • $\begingroup$ If you have unit squares whose centers have integer coordinates, then the vertices of these squares have half-integral coordinates, so the edges of the squares are not line segments joining points of ${\bf Z}^2$, right? $\endgroup$ Commented 2 days ago
  • $\begingroup$ Thank you, Gerry. I have fixed the definition of the region $R$ and added a figure illustrating the case $R(4)$ at an intermediate (possibly sub-optimal) stage of a game; note that $R(1)$ would consist of a single non-playable unit square. $\endgroup$ Commented 2 days ago
  • $\begingroup$ With $R(3)$ having 36 (if my tally is right) moves available, with a bit of alpha-beta and some transposition tables it might be within reach of a tree search. $\endgroup$ Commented 2 days ago
  • $\begingroup$ So, the unit squares forming $R$ are meant to be axis-aligned (otherwise, "horizontal" and "vertical" don't make much sense). $\endgroup$ Commented 2 days ago
  • 1
    $\begingroup$ I calculate that $R(3)$ has a value of $-2$ for the first player. The full game tree contains $726$ graphs (using nauty to canonicalise). $\endgroup$ Commented 19 hours ago

0

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.