How this works
Seven specialist agents (macro, fundamental, quant, news, risk, skeptic, PM) run a full investment process on a simulated market generated in your browser: regime-switching price paths, drifting fair values, and scheduled news events across 10 assets.
- Research cycle — each researcher pitches their strongest signal with a confidence level; the skeptic attacks the two highest-conviction theses; risk checks sizing; the PM builds the book, weighting agents by live reputation.
- Advance time — the market moves, predictions resolve, and every agent is scored with the Brier score (p − outcome)²: 0 is perfect, 0.25 is a coin flip.
- Reputation is earned — PM position weights scale with each agent's trailing Brier score. Calibration curves show whether "70% confident" really means 70%.
- The ledger is append-only — every prediction is permanently recorded and later graded. Nothing is cherry-picked. State persists in your browser; ↺ starts a fresh market.
⚠️ Simulation, not advice. All prices, agents, research, and P&L are fictional and generated locally in JavaScript. No real money, no real securities, no real predictions.