I found this game https://sumplete.com/ and it was made by AI. Let's try to say mathematically what this game is about. We have $n\times n$ matrix $A$. Let $B= \{n\times n \text{ matrices, having } b_1, b_2, ..., b_n \text{ on its diagonal}\}$, $C= \{n\times n \text{ matrices, having } c_1, c_2, ..., c_n \text{ on its diagonal}\}$. In the game we need to find $n\times n$ matrix $X$ with elements from $\{0,1\}$ so that $AX \in B, A^TX^T \in C$. I think that there are a lot of questions we can ask.
For example:
- Does this matrix $X$ exist for any $A, B, C$? (obvious one)
- Is there a criteria for existence of $X$?
- Let all elements in all matrices have elements from $\mathbb{N}$ and an appropriate $X$ exists. Is there an algorythm for finding $X$?
- Let all elements in all matrices have elements from $\mathbb{Z}$ and an appropriate $X$ exists. Is there an algorythm for finding $X$? And so on... This list can be easily expanded.
I find this problem very exciting, but I don't have any ideas about significant steps that can be done here :)
This game is in connection with the domain called discrete tomography.
In fact if the issue is
"being given the margins, reconstruct a possible distribution of values in the 9 boxes"
(starting or not from an initial distribution ; I am not certain I have well understood the rules).
A major issue is the non unicity due to the "linear algebra" fact that
An underdetermined system of 9 unknowns and 6 equations can be "solved" by fixing $9-6 = 3$ parameters, which can be considered as "degrees of freedom"...
... this value $3$, not surprizingly, can be interpreted as the dimension of a certain kernel, with for example this basis with 3 elements :
$$\begin{pmatrix}1&-1&0\\-1&1&0\\0&0&0\end{pmatrix}, \begin{pmatrix}1&0&-1\\0&0&0\\-1&0&1\end{pmatrix}, \begin{pmatrix}0&0&0\\0&1&-1\\0&-1&1\end{pmatrix}$$
i.e., a basis of the subspace of matrices for which all the lines and columns sum up to $0$.
Please note that the elements belong to $\mathbb{Z}$. Limitations to elements of $\mathbb{N}$ would imply another branch of mathematics, i.e. linear programming (simplex method, etc.)
In this kind of game, strategies are of a very mathematical nature. Therefore, a (suitably programmed) computer should always beat human players...
Remark : a hidden constraint is that the right margin and bottom margin must have the same sum.