From intersection names to street rankings in Kyoto: Comparing ranking methods for incomplete pairwise data

Rankings constructed from incomplete pairwise comparisons may depend on the method used, yet applied studies often report a single ordering without quantifying disagreement among methods. Kyoto's intersection names provide both a methodological example and a subject of broad public interest. Central Kyoto has an approximately grid-like street network, and many intersections are named by combining the names of one east-west and one north-south street. The order of the two street names is not fixed. Inspired by a newspaper article asking which street is the "strongest," we treat the street mentioned first in an intersection name as the winner and the second as the loser. When an intersection bears a neutral name containing neither street name, the two streets are assigned a draw. The dataset consists of 59 observed comparisons, 51 decisive outcomes and eight draws, among the 64 possible pairs formed by eight east-west and eight north-south streets. The remaining five street pairs, for which no intersection name exists, are treated as missing comparisons. Rankings were calculated using Colley, Keener, and Markov chain-based methods. Nishioji-dori ranked second under the Colley method and first under both the Keener and Markov methods. Simple averaging of the normalized scores placed Nishioji-dori first, closely followed by Kujo-dori. Although the methods broadly agreed on the top-ranked streets, their score spacing and lower positions differed. HodgeRank further localized residual disagreement among the methods. This study demonstrates ranking and rank aggregation for an incomplete bipartite comparison network while showing how familiar civic data can connect quantitative research with public interest in Kyoto's history and everyday geography.