Library stepwise forward optimization for Meta

Library stepwise forward-optimization for MetaTrader

This is the main page of the documentation library for WalkForwardOptimizer MetaTrader 4/5.

introduction

Step by step forward optimization (Walk-forward optimization, WFO) is designed to verify that the optimized trading adviser (expert adviser, EA) will maintain its profitability in the future unknown data. This is achieved by running the advisor forward-test period with the so-called "data is sample" (Out-of-sample, OOS), which were not used during optimization. In order to optimize other use typically the previous period, the data from which is formed a so-called sample (in-sample, InS). Process consecutive optimizations and forward test is repeated many times with shift work window in the future. Details can be found in an article in English about the walk-forward optimization in Wikipedia.

The documentation uses the following terms:

= Data sampling optimization window

Data is sample = step forward testing

If the average profitability of the adviser to the forward-tests is acceptable, it proves his ustroychivost. Otherwise, you need to change any settings to optimize or advisor strategy as a whole, while the system does not acquire a sufficient ability to generalize, you need to operate in the future.

In MetaTrader no built-in mode step forward testing (MT5 allows for a single forward-testing, but it is not suitable for full-step analysis). To address this a couple of tools you can use problem: WalkForwardOptimizer library and WalkForwardReporter script (WFR).

Library stepwise forward optimization for Meta

User guide

The header file and the case study (page WFO MT4 in the Market)

The header file and the case study (page WFO MT5 in the Market)

Options

functions

formula parameters

Creating reports (WFR MT4 page in Market)

usage scenarios

FAQ

Product Overview for the Walk-Forward Analysis

Related posts

Next posts

Leave a Reply

Your email address will not be published. Required fields are marked *