Discussing the article: "Strategy Optimization and Forward Analysis (Part 1): The Pardo Method — A Basic Model"

New comment

Check out the new article: Strategy Optimization and Forward Analysis (Part 1): The Pardo Method — A Basic Model.

The article explains how to establish a reproducible process for developing and testing trading systems in MetaTrader 5: from formalizing entry and exit rules and risk management to post-optimization validation. It is based on the "Pardo Method": splitting historical data into in-sample and out-of-sample periods, forward testing, multiple markets/timeframes, and selecting stable parameter "plateaus" instead of isolated peaks. Using the examples of PardoSystem and the PardoEA / Breakout_Bounce Expert Advisors (EAs), this article demonstrates a practical test plan for the MetaTrader 5 Strategy Tester.

You have written a strategy, built an indicator and an Expert Advisor (EA), and gotten a nice curve in the backtest — but the main question remains: can you trust it in real trading? This article focuses not merely on implementing ideas in MQL5 code, but on building a reproducible workflow pipeline: from formalizing a trading idea to objective post-optimization validation. We describe the specific artifacts and steps that every practicing algorithm developer should end up with:

  • clear formalization of entry/exit rules and risk management (an indicator plus one or two Expert Advisors (EAs));
  • selection of relevant parameters and determination of scan ranges;
  • a scheme for dividing the history into in-sample and out-of-sample (forward) periods, with criteria for window sizes;
  • a method for finding “plateau” (stable) settings instead of fragile best runs;
  • a protocol for forward testing, multi-environment testing (across multiple markets and timeframes), and decision-making thresholds.

The purpose of this article is not to provide a set of magic filters, but rather a practical test plan and procedure that will help you understand which parameters are truly important, how to assess the overall reliability of a model, and how to determine its “shelf life” and the rules for re-optimization. The examples show the PardoSystem indicator and the PardoEA / Breakout_Bounce Expert Advisors (EAs) as illustrative implementations of this methodology.


Author: Roman Shiredchenko

It says:

Существует два подхода к созданию торговой системы. Первый подход использует логику...

Where is the second attempt? I haven’t even seen a mention of it.

Stanislav Korotky #:

Written:

Where’s the second attempt? I haven’t even seen a mention of it.

In the second part, I’ll explain: the second approach is an empirical search, without any theoretical basis or explanation – like looking for a needle in a haystack. Although the second approach can lead to profitable trading models, it is characterised by a poor understanding of why the model works...
So where’s the nice picture showing the result of the test run/tuning?))
Dmitriy Skub #:
So where’s the nice picture showing the result of the test run/tuning))?

) (that’s the key bit)

Let’s keep this in the discussions – it doesn’t fit the narrative, but these are positive findings for real trading!

New comment