The Numbers
We crawl the Nostr follow graph continuously. As of February 2026, our scoring engine indexes:
- 51,551 nodes (unique pubkeys)
- 622,402 edges (follow relationships)
- Average degree: ~12 (follows per account)
We ran standard graph metrics on this data. One number stood out.
Gini Coefficient: 0.049
The Gini coefficient measures inequality. 0 means perfect equality (everyone has the same number of followers). 1 means total concentration (one account has all followers).
Nostr's follow graph has a Gini of 0.049. That's extraordinarily flat.
For comparison, Twitter's follower distribution has an estimated Gini of ~0.9. Most accounts have near-zero followers while a handful have millions. Nostr is fundamentally different: attention is distributed more evenly across the network.
Why This Matters for Trust
On Twitter, follower count is the primary trust signal. But when 0.01% of accounts hold most of the followers, that signal is easily gamed. Buy followers, get influence.
On Nostr, the flat distribution means PageRank-style algorithms work much better. When connections are distributed, graph topology reveals genuine trust patterns that raw follower counts miss.
Here's an example from our data: the account ranked #42 in trust score out of 51K nodes has only 3 followers. The account ranked #43 has 2 followers. They're trusted not because of popularity, but because of who follows them.
Power Law: Alpha ~2.0
Despite the low Gini, the degree distribution follows a power law with alpha ~2.0. This is the classic "scale-free" pattern found in all social networks — a few hubs, many peripheral nodes. But the hubs in Nostr are much less extreme than Twitter's.
A power-law alpha of 2.0 means the network has organic hub formation without runaway concentration. The top accounts have hundreds of followers, not millions. This is a healthier topology for trust scoring.
Single Connected Component
The entire indexed graph forms a single connected component. Every node can reach every other node through follow paths. There are no isolated clusters or disconnected islands.
This means trust can propagate across the entire network. PageRank converges cleanly, and trust path analysis (finding the shortest trust chain between any two pubkeys) always returns a result.
Try It Yourself
All of these metrics come from our /network-health endpoint:
curl https://wot.klabo.world/network-healthReturns graph size, Gini coefficient, power-law alpha, density, average degree, and connectivity status. Free tier: 50 requests/day per IP.
Or explore interactively at wot.klabo.world/demo — the Network Health card loads automatically with no pubkey needed.
What's Next
We're presenting these findings live on Thursday Feb 12 at 8am PST on Zap.Stream / nosfabrica as part of the WoT-a-thon. The demo includes trust scoring, sybil detection, influence simulation, and more across all 49 API endpoints.
Full API docs: wot.klabo.world/swagger