Walk-forward analysis

Walk-forward analysis tunes a rule on one window of history, tests it on the window that follows, then rolls both windows forward and repeats.

How Walk-forward analysis is calculated

Split time into consecutive pieces. Pick parameters using the first training window, then record the result on the next test window without changing anything. Slide the pair forward and repeat. Stitching the test windows together gives a result that never used its own data to choose its settings.

How it is read

It mimics how a rule would have been re-tuned in real time. The weakness is that it still allows many tries on the training windows. Read the stitched result together with how many parameter sets were tried in each window and how many trades fell in the test windows, since a few trades say little.

Common mistakes

• Re-running the whole procedure with new settings until the stitched result looks good.
• Making the test windows too short to contain enough trades.
• Letting information from the test window leak into the next training window.

What Tickfloor tested

Tickfloor has not published a backtest of a rule built only on Walk-forward analysis. It describes or manages something rather than giving a signal, so there is no strategy to score. It still shapes how any tested rule should be read.

Lessons that cover it

• Testing on prices the strategy never saw

Related concepts

• Out-of-sample testing
• Overfitting
• Look-ahead bias
• Backtest
• All concepts

General information only. It doesn't consider your objectives, finances or needs. Tickfloor holds no financial services licence and never places trades.

• See a full worked example (free, no account)
• Browse every course lesson by topic