This is a problem from mathematical contest for kids.
Split integer numbers $1, 2,\ldots, 31$ into sets such that a maximum value in each set is the sum of the others, or show that it's impossible.
My analysis so far has been that if we suppose that it's possible, and we end up with the collection of sets $G$, then if we denote the maximum element of each set $g \in G$ as $m(g)$, $$\sum_{x \in g} x = \sum_{x \in g \setminus {m(g)}} x + m(g) = 2 \cdot m(g)$$ $$\sum_{i=1}^{31}i = \sum_{g \in G} \sum_{x \in g} x = \sum_{g \in G} 2 \cdot m(g) = 2 \cdot \sum_{g \in G} m(g) $$
However $$\sum_{i=1}^{31}i = \frac{1 + 31}{2} \cdot 31 = 16 \cdot 31$$ So $$2 \cdot \sum_{g \in G} m(g) = 16 \cdot 31$$ and $$ \sum_{g \in G} m(g) = 8 \cdot 31$$
Thus I need to find a set of maximum elements sum of which is 31, and then for each of the maximum elements, augment their sets by some other numbers so that their sum is equal to the respective maximum elements.
I wasn't able to analyse it further, and when I tried to work out the solution by hand, somehow I always was unable to find a required partition.
There seem to be many ideas to prune the search (like 31 will be a maximum element, then it's either that we take $31 = 30 + 1$ and then $29$ will be a maximum element, or we'll have $31$ and $30$ as maximum elements). However I still don't see how this could be done without the use of a computer. Implementing a search procedure in a programming language should also be an interesting problem, and I think it could be done, but if it won't find a solution, I think such a brute force rejection won't be considered as a satisfactory solution in a sense that there should be some less computationally hard argument.