Workbench
Lab Mode
Interactive research workbench driven by live data. All 6 domains fetch real-world data to parameterize BTUT simulations.
Dual API — Yahoo Finance + Frankfurter (ECB)
Yahoo FinanceFrankfurter (ECB)
Domain
Parameters
Agents (N)1000
Kernel σ0.1
Drift Coeff0.1
Diffusion0.01
Max Iterations50
Game Type
Controls
Agent Distribution
Bullish
Bearish
Nash Gap Convergence
Run a simulation to see convergence data
Integration Logs0 entries
No logs yet. Start a simulation to see API integration.
Experiments
0 runsNo experiments yet
Run a simulation to create one
Theory Reference
∂μ/∂t = −∇·(μ v[μ]) + σ²Δμ−∇·(μ v[μ])
Drift: Agents move toward better strategies (exploitation)
+σ²Δμ
Diffusion: Random exploration prevents local minima
Variable Reference
μPopulation density (where agents are)
tTime (simulation iterations)
v[μ]Velocity field (optimal direction)
σExploration noise (diffusion strength)
∇·Divergence (flow out of region)
ΔLaplacian (curvature of density)
Convergence:ε = E[||uᵢ − u*||²] → 0
Nash gap measures deviation from equilibrium
Higher σ
More exploration, slower but more robust convergence
Lower σ
Less exploration, faster but may miss global optimum
Higher Drift
Agents follow gradient more aggressively
More Agents
Better mean-field approximation, O(N) scaling