Walk Forward analysis (WFA) or Walk Forward optimization (WFO) is a sequential optimization applied to an investment strategy. The name of the analysis is called “walk forward” because we have a moving window that progressively traverses the whole period of the data history with a pre-established step.
Algorithmic traders apply such analysis to decrease the over-optimized parameters used in the investment strategy as we don’t want only a great looking backtest result, we want also a system that don't fail in live real money account.
As you already know WFA consist on multiple optimization stages and to perform the analysis we need to do this within a time window, where we first have the optimization in a larger window (in-sample) and a test in a smaller window (out-of-sample, unseen data).
During segment 1 all parameters combinations for the trading strategy will be tested (in-sample) and the parameter combination that is considered to be the most robust by the optimizer criteria will be used in the first ouf-of-sample segment.
The process will be repeat until the optimizer has traversed the whole data history with a pre-established step selected in the Options and results of the out-of-sample results will be at the Statistics page.
Once the optimization has completed you can analyze the equity chart and the statistics to understand if the strategy has performed better or survive market dynamics changes.
The example above was a successful result of a WFO. The strategy still performing on unseen data (out-of-sample).
You can analyse the result on the tested segments and most relevant metric's as start date, end date, profit, return / dd, trades, sqn, win/loss and more.
Change the inputs for the analysis as: number of segments, out of sample, numeric values range and optimization goal.
Apply a performance metric in each segment to validate the success of a strategy.