Bias and Evaluation

Status Draft briefPart Simulation and Backtesting

Look-ahead, survivorship, selection, overfitting, optimistic fills, missing costs, and repeated testing can manufacture attractive historical performance.

Central questions

• Was every input available at the simulated decision time?
• Which choices were made after inspecting evaluation results?
• Does reported uncertainty include parameter and model selection?

Planned model

Introduce common biases one at a time into a strategy with no true edge and show how each changes reported returns, risk, and significance.