Walk-forward optimization without curve fitting

A strategy-optimization report can rank candidate parameter sets by column headings such as net profit, placing the highest historical-profit result first. The top-ranked result on a net-profit-sorted optimization report is not automatically a usable trading procedure.

How optimization fits the sample

System optimization varies rule inputs and ranks candidate parameter sets on historical results so entry, exit, and abstention rules can be tested as one procedure.

Optimization, as defined in the source glossary, tailors system rules so they fit the historical data under study as closely as possible. Overfitting occurs when parameters are chosen specifically to return the highest profit on the historical sample. That selection describes past data rather than a repeatable decision procedure.

Indicators do not themselves issue trading signals. Strategies issue the signals that produce profits and losses. A trading signal is the buy, sell, or stay-out instruction produced by a complete strategy, not by an indicator reading alone.

Checks a ranked report cannot supply

Backtesting means fitting or testing a strategy on historical data and then applying the same strategy to later data to see whether the results remain consistent.

Walk-forward analysis is a rolling in-sample optimization followed by a later out-of-sample test that checks whether the selected rules still produce usable signals under new market data.

Robustness testing asks whether a selected parameter set remains usable when nearby values, later periods, or execution constraints change, instead of relying on a single peak result.

What this workflow does not decide

Knowing optimization techniques does not by itself identify which market techniques will work. That question is treated as a later research step.