TVMcp
Common Pine Script Backtesting Mistakes and How to Avoid Them
The mistakes that quietly ruin strategy confidence, plus practical ways to prevent each one.
Repainting and lookahead are silent killers
If your signal depends on future information, your backtest is not predictive. It is a spoiler.
Audit each indicator dependency and confirm behavior on closed bars where required.
Ignoring execution costs
No slippage and no fees make almost anything look smart. Reality then sends the invoice.
Use conservative friction assumptions as the default baseline.
Over-optimization by default
Huge parameter sweeps can manufacture “perfect” results that fail instantly outside sample.
Prefer robust zones over isolated peaks every single time.
No run documentation
If you cannot rerun yesterday’s “great result,” that result has no operational value.
Store code version, inputs, dates, and assumptions with each decision-grade run.