🕸️ Network Topology Generator: Random Graphs & Small Worlds

"Six degrees of separation" is the popular name for the empirical observation — going back to Stanley Milgram's 1960s small-world experiments and later confirmed at internet scale (e.g. Microsoft Messenger and Facebook studies finding average distances around 5–6) — that most pairs of people in a huge social network are connected by a surprisingly short chain of acquaintances. It's exactly what the Watts–Strogatz model reproduces mathematically: a network can be enormous and still have a tiny average shortest-path length once a few long-range shortcuts exist.

A power-law degree distribution means the fraction of nodes with degree k falls off as roughly k⁻ᵞ rather than clustering around a mean — so while most nodes have very few connections, a small but non-negligible number have very many, with no natural "typical" scale (hence "scale-free"). In this simulation's histogram it shows up as a long, thin tail stretching out toward the max-degree stat for Barabási–Albert graphs, in sharp visual contrast to the narrow, bell-shaped histograms produced by Erdős–Rényi or Watts–Strogatz.

Measured internet AS-level topology, the web's hyperlink graph, and most social networks all show both high clustering and short average path lengths together with heavy-tailed, often power-law-like degree distributions — properties an Erdős–Rényi graph with the same node and edge count essentially cannot produce, because independence washes out both clustering and hub structure. Small-world and scale-free generative models were specifically developed because they reproduce these measured statistical signatures, not just the raw node and edge counts.

This is a well-documented, published network-science result (Albert, Jeong & Barabási, 2000): because most nodes in a scale-free network have low degree, removing nodes at random overwhelmingly hits low-degree nodes and barely dents overall connectivity. But removing the handful of high-degree hubs — the same nodes this simulation visibly enlarges for Barabási–Albert graphs — fragments the network far faster than removing random nodes would, which is exactly why hub nodes (major internet routers, key social-network accounts) are both the internet's greatest strength against random outages and its clearest structural vulnerability against deliberate, targeted attack.

A node's local clustering coefficient is the fraction of pairs among its neighbors that are themselves directly connected — literally, how many triangles it participates in relative to how many are possible. This simulation computes it exactly for every node (counting real edges among each node's neighbor set) and averages across all nodes to get the network-wide clustering coefficient shown in the stats panel, which is the standard definition used throughout graph-theory literature.