—σ above degree-null
Calculating the evidence trail.
DTU 02805 · go nuts · week 3
Who is Marvel’s most under-credited broker?
High betweenness may only mean high degree. We rewired Marvel 80 times while preserving every character’s number of links, then asked whose brokerage refused to disappear.
Open the evidence →- articles
- —
- undirected links
- —
- null universes
- —
01 / evidence field
Fame on one axis.
Brokerage on the other.
Each dot is a character. Change the lens from observed betweenness to a z-score against degree-preserving rewires. Positive z means “more brokerage than this degree normally buys.”
Hover or click a dot to open its dossier. Log scales reveal the low-degree brokers.
- connections
- — —
- betweenness
- — —
- null expectation
- — same degree, rewired
- stranded if removed
- — —
highest-degree neighbours
most surprising brokers
03 / stress test
Now pull people out.
We ranked the untouched network once, removed characters in that order, and tracked the largest surviving component. This tests whether a centrality ranking actually identifies structural weak points—not merely prestigious pages.
04
case finding
One character brokers far more than their degree predicts.
Degree asks who knows many people. Betweenness asks who controls routes. The degree-preserving null is what turns that distinction from a slogan into evidence.
- Representation
- A node is a Marvel superhero Wikipedia article. An undirected, unweighted edge exists when either article links to the other. Analysis uses the 277-node giant component.
- Null model
- 80 sequential degree-preserving double-edge shuffles, three swaps per edge between samples. The script asserts that every node keeps exactly the same degree.
- Important limit
- Wikipedia links measure editorial attention, not friendship or story-canon relationships. A high z-score identifies unusual linking structure, not a causal social role.
- Human check before publishing
- Pick one reported subject, count its links in the frozen edge list, and inspect two neighbours on Wikipedia. Record the checker and whether the “broker” interpretation survives.
- Reproduce
uv run analyse_week3.pyrebuilds this page’s JSON.