PageRank, explained.
PageRank represents a node's importance through a random-surfer stationary distribution combining link following with teleportation.
Why it happens
Four nodes start with links A→B,C; B→C; C→A; D→C. Toggle directed edges and link-following probability d. Iterate one hundred times from a uniform distribution, distributing dangling-node mass uniformly. Ranks sum to one.
PageRank represents a node's importance through a random-surfer stationary distribution combining link following with teleportation.
Read the result
Rank is conserved mass flowing through links and uniform jumps. Track the outgoing links as well as incoming ones before explaining a changed percentage.
A worked example
A link from a prominent page
Inspect the initial A–B–C loop and the weakly supported D node at link-following probability 0.85.
Toggle A→D and compare D's new rank with its starting value while the total rank stays one.
Set link-following probability to zero and observe uniform twenty-five-percent ranks, independent of the link graph.
OPTIONAL DEEPER DETAILGo deeper: inside the model
Inside this model
Four nodes start with links A→B,C; B→C; C→A; D→C. Toggle directed edges and link-following probability d. Iterate one hundred times from a uniform distribution, distributing dangling-node mass uniformly. Ranks sum to one.
Where this idea is useful
A practical use
A link from an already prominent node can transfer different rank mass from a link on a weakly connected page.
A common misconception
“PageRank is simply an incoming-link count.”
The source node's rank, outgoing links and teleportation all matter. A modern search engine also uses many signals absent from this tiny model.
What this explanation leaves out
- The tiny link-only model does not reproduce a modern search engine or imply a formula for organic traffic. Teleportation is uniform and the network is fixed during each calculation.
What happens to a node with no outgoing links?
Its link-following mass is distributed uniformly in this implementation, preventing probability from disappearing at a dangling node.
Are the signals supporting a ranking independent evidence of quality or partly a consequence of existing visibility?
Associated thinkers
Associations marked provisional are awaiting source review.
Further reading
Explore the original research or the teaching reference behind this experiment.