Each Walkforward segment contains two separate periods:
- in-sample
- out-of-sample
They serve different purposes.
In-sample
The in-sample period is the historical data used to evaluate candidate parameter combinations.
Best Score or Stable then selects one configuration from those in-sample Results.
The in-sample period answers:
Out-of-sample
The out-of-sample period comes after the in-sample period.
The configuration selected in-sample is applied to this later data.
The out-of-sample period answers:
Why the periods are separated
The data used to select a configuration is kept separate from the later data used to evaluate it.
This separation is repeated for every eligible Walkforward segment.
What forms the final Walkforward performance
In-sample Results are used to select the configuration for each segment.
They are not added to the final Walkforward performance.
The final performance is calculated from the combined out-of-sample trades.
Period units
Walkforward period units follow the dataset timeframe.
Intraday and daily datasets use Trading Days.
Weekly and monthly datasets use the corresponding weekly or monthly periods.
Selection happens again in every segment
Each segment performs a new in-sample evaluation.
The selected configuration can therefore remain the same or change from one segment to another.
Where to go next
Continue with Anchored and Rolling Walkforward to see how the in-sample period moves between segments.