Candy-passing Games on General Graphs, I

We undertake the first study of the candy-passing game on arbitrary connected graphs. We obtain a general stabilization result which encompasses the first author’s results (arXiv:0709.2156) for candy-passing games on n-cycles with at least 3n candies.

💡 Research Summary

The paper introduces and studies the candy‑passing game on an arbitrary connected undirected graph G = (V,E). The game proceeds in discrete rounds triggered by a whistle. At the start, a positive integer c candies are distributed among the |V| vertices. In each round every vertex that holds at least as many candies as its degree deg(v) simultaneously sends one candy to each neighbor; vertices with fewer candies do nothing. The authors drop the “student” metaphor and treat the candies as piles attached to vertices.

A central notion is that of an “abundant” vertex: a vertex v is abundant if it holds at least 2·deg(v) candies. The paper proves Lemma 1, which states that after a finite number of rounds the set of abundant vertices becomes fixed and each abundant vertex stabilizes (its candy count no longer changes). The proof is straightforward: the total number of candies on abundant vertices never increases, and whenever an abundant vertex loses candy the total strictly decreases. Since the total cannot drop below zero, only finitely many such losses can occur, forcing the set and the amounts to become constant.

The main result, Theorem 2, asserts that if the total number of candies satisfies

  c ≥ 4|E| − |V|,

then every vertex of G eventually stabilizes, i.e., its candy count becomes constant from some round onward. The proof proceeds by case analysis. If no vertex is abundant after the finite “clean‑up” phase guaranteed by Lemma 1, the inequality forces c = 4|E| − |V| and each vertex v must hold exactly 2·deg(v) − 1 candies. In this configuration every vertex is already stable.

If at least one abundant vertex remains, that vertex must be stable (by Lemma 1) and therefore receives a candy from each neighbor in every round. Consequently each neighbor must have at least deg(v) candies each round, which forces those neighbors to also fire every round. By the same argument their neighbors must fire, and so on. Because G is connected, this propagation reaches all vertices, showing that all of them eventually stabilize.

The theorem subsumes earlier work on cycles. For an n‑cycle we have |E| = |V| = n, so the condition becomes c ≥ 3n, which matches the bound previously obtained for cycles (the earlier paper gave c ≥ 3n − 2). For a k‑regular connected graph the bound simplifies to c ≥ (2k − 1)·|V|, a clean expression linking the required total candy to the degree and size of the graph.

The authors note the close relationship between the candy‑passing game and the well‑studied chip‑firing model. Both are discrete dynamical systems on graphs where vertices “fire” when a threshold is met, redistributing a conserved quantity to neighbors. The present work extends the chip‑firing paradigm by using a threshold based on the vertex degree rather than a fixed integer, and by providing a universal stabilization criterion that depends only on elementary graph parameters.

In the concluding remarks the paper suggests several avenues for future research: determining tighter thresholds when c is below the 4|E| − |V| bound, analyzing non‑connected or directed graphs, and exploring stochastic variations where the whistle intervals are random. Overall, the paper delivers a concise yet powerful generalization of candy‑passing dynamics, establishing that a simple linear lower bound on the total number of candies guarantees eventual equilibrium on any connected graph.