2D Financial Bubble — Agent-Based Market Simulation

Trader Population

Number of agents 300

Herding intensity κ 600

Strategy Sensitivity

Momentum sensitivity β_c 2.0

Fundamentalist sensitivity β_f 1.5

Market liquidity impact λ 0.25

Idiosyncratic noise 0.010

Fundamental growth g 2.0%

Actions

Simulation speed 3×

No scripted price path: every tick each agent submits real buy/sell demand from its own strategy, price moves from the net of all of it, and agents periodically re-evaluate which strategy has been paying off and switch.

Live Readouts

Price

100.0

Fundamental value

100.0

P / V ratio

1.00

Max drawdown

0.0%

Momentum traders

50%

Fundamentalists

50%

Net order flow this tick

0.000

How it works

This is a genuine agent-based market: every one of the N trader objects in the simulation has its own strategy type and its own sensitivity, and submits its own demand d_i each tick.

Fundamentalists compute d_i = sens_i · β_f · clamp((V−P)/P), buying when the price is below fundamental value V and selling when it's above — a stabilising force.
Momentum traders compute d_i = sens_i · β_c · clamp(trend), buying because the price has recently been rising — a destabilising, trend-amplifying force.
Price formation: the market price moves from the real average net demand across every agent that tick, P ← P·(1 + λ·mean(d_i)). Nothing about the price path is pre-scripted.
Strategy switching: each tick a small random subset of agents re-evaluates which strategy has recently been more profitable (tracked as an exponential moving average of realised payoff) and switches with a herding-weighted probability. When momentum traders start winning, more agents copy them — this is what fuels the mania. When the price/fundamental gap grows large enough, fundamentalist payoffs turn attractive again and traders switch back, triggering the correction.

The boom, mania and crash you see are an emergent property of the population's changing composition — not a phase state-machine.

Fundamentalist agent Momentum agent