A random walk is a sequence of points in space where each step is chosen randomly and independently. A central question is: what is the probability that the convex hull (the tightest enclosing shape) of the points visited so far never traps the origin, the starting point? This question matters in probability theory and has connections to classical results going back to the 1950s. Previous answers required special assumptions, like the steps being symmetric around the origin or the points being in "general position" (no special geometric coincidences). This paper removes all such assumptions and proves a clean formula that works for any random walk in any dimension.
The key formula expresses the probability that the origin avoids the convex hull of the first n steps in terms of similar probabilities for independent copies of the walk running for shorter durations. The structure mirrors how permutations decompose into cycles, which is a classical idea in combinatorics. Remarkably, the result is rooted in a purely deterministic geometric identity: for any fixed sequence of vectors, the number of orderings that keep the origin outside the convex hull of partial sums equals the number of permutations for which the origin stays outside the convex hull of the "cycle sums." This deterministic statement is the engine behind the probabilistic formula.
The paper unifies and extends a surprising range of earlier results. The classical Sparre Andersen and Spitzer identities from the 1950s, which describe how often random walks stay positive, fall out as special cases. So do more recent distribution-free absorption probability formulas due to Kabluchko, Vysotsky, and Zaporozhets, as well as a generating function result by Petrov, Randon-Furling, and Zaporozhets. The authors also derive a new formula for the probability that the origin is never absorbed at any time. The approach works without any symmetry or non-degeneracy conditions, suggesting that these absorption phenomena are fundamentally combinatorial rather than geometric in nature.