Two-mode network depicts triathlon competitors attending consecutive yearly races in Hamburg from 2009 to 2015.
Data sample scraped from Timing & Result Service mika:timing for Triathlon Hamburg Everyman Olympic Men Races 2009–2015.
Participants not included in the sample:
- DNS (Did Not Start)
- DNF (Did Not Finish)
- DQ (Disqualified)
Participants with missing split times have been removed from the sample.
Data has been concatenated, and recurrent participants identified by name. This is only a rough approximation for several reasons:
- data is self-reported and spelling varies (e.g. “André” and “Andre”, titles and middle names have been included and/or omitted, typographic errors)
- homonymous participants are treated as one person
Graph has been created with Gephi 0.8.2 using Yifan Hu layout (default settings), and exported with SigmaExporter. Nodes have been coloured by modularity class, size of nodes has been ranked by In-Degree. Edges are directed.
- No. of nodes: 10,729
- No. of edges: 18,655
- Density: 0.000062
- No. of modularity classes: 7
All races show large clouds of (one-time) participants in the outer ring of the network diagram that could not be re-identified by the method applied. These sum up to 61&thsp;% of the total. Middle-sized clusters show some overlap between two races (20&thsp;% of the total). Small clusters of athletes that participated in more than two races hover in the center of the network diagram.
Since the number of potential connections grows exponentially with the size of the network (n * (n-1)), large networks usually display a low density (no. of connections divided by no. of potential connections). The density of this network is especially low, since connections between the races’ participants have neither been surveyed nor considered.
A different and arguably better visualisation is a one-mode network, where the number of recurring competitors is used as edge weight between races.