TVMcp
Pine Script Optimization Best Practices for Robust Results
A practical optimization process focused on stability and robustness, not lucky peaks.
Do not worship the single best number
If your “best” setup only works at one exact parameter value, that is not a strategy. That is a coincidence wearing a suit.
Optimization should find stable neighborhoods, not lottery tickets.
Constrain the search space first
Define parameter ranges from market logic, then run coarse sweeps. Narrow only where behavior stays stable.
Blind brute force across huge ranges creates beautiful overfitting artifacts and very expensive disappointment.
Measure robustness, not vanity metrics
Track drawdown behavior, trade count quality, and consistency across neighboring values.
A slightly lower CAGR with resilient behavior is usually the better production candidate.
Always include out-of-sample validation
Pick candidates in-sample, validate out-of-sample, then re-check across regime shifts.
If performance collapses outside one window, optimization found history, not edge.