Skip to main content

Weights (/weights)

Controls how much influence each model has on the ensemble score.

Current Weights

Shows Claude, GPT-4o, and Gemini weights (sum to 1.0) plus the disagreement penalty factor (k).

Simulate

Test different weight configurations against historical data:
  • Adjust sliders for each model
  • See simulated vs current avg score, trade rate, and accuracy
  • Compare before committing

Weight History

Audit trail of all weight changes with timestamp, source (manual/recalibration/simulation), sample size, and the weights at each point.

Auto-tune

tune_risk_params uses half-Kelly sizing from the last 100 outcomes to automatically optimize weights.