Agent-based prediction-market simulation of the BTC spot market

What it is

A discrete-time agent-based model of the BTC spot market. Traders are heterogeneous: attention-based participation decides whether they trade each step, and a herding term decides which way. Belief signals come from Polymarket-style prediction-market prices.

I owned the system infrastructure — the simulation core, the parameter sweeps, and the Solara dashboard used to explore runs interactively.

Findings

• Stronger herding amplifies signal transmission: as the herding parameter rises, participation increases and order imbalance shifts to the extremes.
• Price dynamics move from fluctuating around fundamentals into persistent one-sided regimes — the model reproduces momentum-like bursts from a purely behavioural mechanism.

Why it matters

It’s a small, inspectable laboratory for how sentiment feeds become price. The same machinery is the natural home for the news/Twitter sentiment work I keep coming back to.