Exact Scale Invariance in the Mixing of Binary Candidates in a Voting System
Preprint
Abstract
We study a voting model for binary candidates and discuss the scale invariance in the mixing of them. Candidates are classified into two categories, good one with score $s$ and bad one with score $w$. The probability that a candidate gets a vote depends on the sum of the score and the cumulative number of votes up to the time of the vote. After infinite times of voting, the probability function of the scaled number of votes $u$ obeys gamma distributions. The cumulative function of good candidates $x_{s}$ in terms of that of the bad candidates $x_{w}$ obeys power law $1-x_{s}\sim (1-x_{w})^{\alpha}$ for small $u$ region with a critical exponent $\alpha=s/w$. In the limit $(s,w)\to (0,0)$ with $s/w=\alpha$ fixed, the relation $1-x_{s}=(1-x_{w})^{\alpha}$ holds exactly over the entire range. We study the data of the horse racing of JRA from 1986 to 2006 and confirm the scale invariance.