Universality of Top Rank Statistics for Brownian Reshuffling

TL;DR

This paper introduces an observable called the overlap ratio to analyze the top rank dynamics of particles undergoing Brownian motion, revealing universal behavior across various stochastic systems.

Contribution

It derives an analytical formula for the average overlap ratio in a system of Brownian particles with boundary conditions, demonstrating universality in top rank statistics.

Findings

Overlap ratio follows a simple erfc function in the large N limit.

Universal behavior observed across multiple stochastic models.

Approximate validity of the formula for moderate top-n sizes.

Abstract

We study the dynamical aspects of the top rank statistics of particles, performing Brownian motions on a half-line, which are ranked by their distance from the origin. For this purpose, we introduce an observable that we call the overlap ratio $\Omega(t)$, whose average is the probability that a particle that is on the top-$n$ list at some time will also be on the top-$n$ list after time $t$. The overlap ratio is a local observable which is concentrated at the top of the ranking and does not require the full ranking of all particles. It is simple to measure in practice. We derive an analytical formula for the average overlap ratio for a system of $N$ particles in the stationary state that undergo independent Brownian motion on the positive real half-axis with a reflecting wall at the origin and a drift towards the wall. In particular, we show that for $N\rightarrow \infty$, the overlap ratio takes a rather simple form $\langle \Omega(t)\rangle = {\rm erfc}(a \sqrt{t})$ for $n\gg 1$ with some scaling parameter $a>0$. This result is a very good approximation even for moderate sizes of the top-$n$ list such as $n=10$. Moreover, as we show, the overlap ratio exhibits universal behavior observed in many dynamical systems including geometric Brownian motion, Brownian motion with a position-dependent drift and a soft barrier on one side, the Bouchaud-M\'ezard wealth distribution model, and Kesten processes.