Each coloured line = one user’s net ⊙ position over time · same IOU amounts & receivers in both panels — only the sender-selection rule differs
⚠ Large user counts may slow or freeze your browser. The effect is equally clear at 1K users.
A · CONVENTIONAL — uniform random sender
Sender chosen with equal probability · no social pressure · all records kept forever
min —
max —
mean —
SD —
crossings —
live records —
IOUs —
IOUs/s —
B · IOUNet — rank-biased sender, same amount & receiver
Sender by net rank: 0× (most indebted) → 2× (most net-positive) · only rank matters, not absolute amount
min —
max —
mean —
SD —
crossings —
live records —
IOUs —
IOUs/s —
debt eliminated —
What you’re watching: Both panels receive IOUs with the same amounts and the same recipients, tick by tick. The only difference is who sends.
Panel A picks senders uniformly at random — no social feedback. Lines drift away from zero and crossings become increasingly rare as the simulation runs.
Panel B applies rank-biased emission: the most net-positive user (has given the most) sends at 2× the base rate; the most net-negative (most indebted) has zero probability of sending. Only rank matters — not absolute balance. No erasure in this panel — all records kept, so the spread difference is purely from sender selection.
The result: 23× tighter spread and 3× higher crossing rate. Panel B’s crossing rate stays high as the simulation runs, while Panel A’s collapses as lines drift from zero. These are two fundamentally different dynamical regimes: one that runs down, one that sustains itself indefinitely.