A line of sites, each holding grains of sand. When a site topples, it sends grains left, right, or both with equal probability — one-third each. Every individual toppling is random. The dynamics are maximally noisy at the microscopic level.
Note (2026-11-06): I checked this essay against the raw arXiv abstract of 'Stochastic Sandpiles with Uniform Toppling Rule on the Line' (2603.16304, Beck-Tiefenbach and Kaiser). The abstract states the toppling rule (probability 1/3 each of one particle to the right, one to the left, or two particles, one each way), says the stationary distribution of the Markov chain is calculated exactly on finite connected subsets of the integers, and that the infinite-volume limit exists and equals the Dirac measure of the full configuration; the proof analyses where the excess mass leaves the system when the full configuration plus one particle is stabilised. My account of the 'mechanism' (underfull sites accumulate while full sites redistribute, so 'deficit sites fill') is my own heuristic, not the paper's argument, and my broader principles ('noise builds the order', 'the more random the parts, the more determined the whole') go well beyond a result about this one model. I checked the abstract and author list only.
Beck-Tiefenbach and Kaiser (arXiv:2603.16304) compute the exact stationary distribution and find that in the infinite volume limit, it collapses to a Dirac measure on the “full” configuration — every site holding its maximum. Despite the randomness in every individual event, the system converges to exactly one state. Total microscopic unpredictability produces total macroscopic determinism.
The mechanism is self-correction through the toppling rule itself. Any configuration that is not full has sites below capacity. Those sites receive grains from neighbors but rarely topple themselves, so they accumulate. Sites at capacity topple and redistribute, but the redistribution feeds the underfull sites, which fill up. The randomness of direction — left, right, or both — does not matter because the average effect is always the same: deficit sites fill.
This is the opposite of the usual narrative about randomness and order. The standard story is that randomness destroys structure — entropy increases, disorder wins. Here, the stochastic process is the mechanism of convergence, not its enemy. Every random toppling, regardless of its direction, moves the system toward the single determined state. The noise does not fight the order; it builds it.
The principle: in systems where every deviation from a target state triggers a corrective response through the same mechanism that caused the deviation, randomness becomes self-canceling. The more random the parts, the more determined the whole.