The Santa Fe Artificial Stock Market

Arthur, Holland, LeBaron, Palmer & Tayler (1997) — Complexity Economics in Action

This is one of the most famous models in complexity economics. Agents with evolving strategies compete to buy and sell a single stock. Each trader uses a genetic algorithm to breed better prediction rules over time — learning from the market they collectively create. Watch what happens when agents stop being perfectly rational and start adapting: bubbles inflate, crashes cascade, and fat-tailed returns emerge — just like real markets.

Beinhocker's Key Insight

When all agents use the same rational strategy, the market is boring and efficient. Price hugs fundamental value, volatility is low, and nothing interesting happens. The moment agents start learning and adapting, the market comes alive with bubbles, crashes, and power laws — just like the real thing. The volatility isn't noise. It's the emergent result of evolving strategies competing for dominance.

Simulation LEARNING Tick: 0
🔈 Sound

Mode

Preset Experiments

Market

Dividend

Evolution (Learning Mode)

Speed

Live Statistics

Price--
Fund. Value--
Dividend--
Volume--
Volatility--
Kurtosis--
Gini--

Market Mood

😐
Neutral
Extreme Fear Extreme Greed
What's Happening
📊
Hit Start to begin the simulation. Agents will begin trading and evolving their strategies.

Agent Leaderboard

Top 5
Bottom 5
BORING: Perfect Rationality
All agents use the same equilibrium formula
ALIVE: Learning Agents
Genetic algorithms evolve competing strategies

Price & Fundamental Value

Watch how price deviates from fundamental value (dashed line). These aren't random — they're bubbles and crashes created by agents' competing strategies.
?
SFI Fun Fact: The original 1997 model ran on a Connection Machine supercomputer at the Santa Fe Institute. Brian Arthur convinced the team that markets aren't just information processors — they're ecologies of competing beliefs.

Trading Volume

Spikes in volume often signal disagreement among agents — a hallmark of real market behavior.
?
SFI Fun Fact: In the rational-expectations regime, volume essentially disappears. Everyone agrees on the "right" price, so there's no reason to trade. But real markets have enormous volume. Why? Learning agents disagree.

Returns Distribution

The tall peaks and fat tails show that extreme events happen far more often than a normal distribution would predict — just like real markets.
?
SFI Fun Fact: Mandelbrot observed fat tails in cotton prices in 1963. The SFI model was one of the first to show how agent-based learning generates this pattern endogenously — no external shocks needed.

Wealth Distribution

Inequality emerges naturally as some strategies outperform others. Lucky early movers compound their advantage.
?
SFI Fun Fact: The SFI team found that wealth inequality in the artificial market follows power-law distributions similar to those observed by Pareto in real economies over 100 years ago.

Gini Coefficient Over Time

Rising Gini means growing inequality. In learning mode, strategy evolution drives wealth concentration then disruption.
?
SFI Fun Fact: The "punctuated equilibria" in Gini mirror patterns seen in biological evolution. Periods of stability are interrupted by bursts of change as new dominant strategies emerge.