Model 2 · replication of Gode & Sunder (1993) · validation gate live
Buyers and sellers bid at random. No learning, no strategy, no rationality. Watch the price find equilibrium anyway — then remove the one constraint and watch it fail.
Each buyer holds a few units and a private value for each; each seller holds units with a private cost. Sorting values from high to low and costs from low to high gives the demand and supply schedules — the two staircases on the canvas. Where they cross is the competitive equilibrium: the price and quantity a textbook says a market of rational agents will find.
The traders here are not rational. On each step one is picked at random and submits a random bid or ask. A new bid must beat the standing bid, a new ask must undercut the standing ask, and when they cross a trade happens at the standing quote. That is the whole mechanism.
The order is in the market structure, not in the agents.
Gode & Sunder, "Allocative Efficiency of Markets with Zero-Intelligence Traders," Journal of Political Economy 101(1), 1993. With the budget constraint, random traders extract nearly all the available surplus and prices converge toward equilibrium. Without it — ZI-U — prices wander the whole range and efficiency drops. The single constraint does the work that "rationality" was supposed to do.
This is Chapter 2 of the book, runnable. Equilibrium never required everyone to be homo economicus. That was a device for doing the mathematics by hand.
Model 1 failed three times because every parameter was invented and nothing could be checked against a known answer. This model has one. Until the gate reads PASS — ZI-C efficiency above 95%, ZI-U clearly lower, ZI-U prices clearly less converged — nothing of ours goes on top. If it fails, the bug is in the implementation, not the world.