A safe process for changing strategy parameters

Treat every input change as a new experiment so a live system does not drift away from the evidence that justified it.

Change one meaningful thing at a time

Changing risk, session, stop, and entry filters together makes the result impossible to attribute. Write the reason for a proposed change and identify whether it addresses strategy evidence, execution evidence, or an operational bug.

If the change is intended to fix a bug, reproduce the bug and test the fix separately from performance optimisation.

Run the same validation ladder

A new parameter set should pass a small logic test, a historical comparison, an out-of-sample check, and a demo or controlled execution stage before live use. Do not promote it because it improved one recent window.

Record old and new values.
Keep the prior configuration available for rollback.
Run a risk and execution diff, not only a profit diff.
Set a review date after deployment.

Make rollback easy

Save configuration files with a version label and know how to disable or revert the EA. A change process is incomplete if an operator cannot restore the last known-safe state during a live incident.

Good change management keeps learning possible without turning the live account into an uncontrolled experiment.

Discuss this article

Ask a setup question, share a backtest, or compare notes with other algorithmic traders. Use a display name; your email remains private.

Create an account or sign in above to join the conversation.

Loading discussion…