Dear Colleagues,

Over the past few months, I have been trading on the MNQ using candlestick patterns. During this period, I identified a recurring pattern that prompted me to undertake a comprehensive study and testing process. This pattern was developed exclusively based on open values without additional indicators, using a combination of up to six bars.

Objective

The primary objective of this article is not to present a detailed analysis of the pattern or develop specific trading code. Rather, it seeks to illustrate the susceptibility to overfitting and the consequent potential for incurring significant losses.

Methodology

Initially, I downloaded the relevant data for the specified time frame and conducted tests on the identified pattern using machine learning to expedite the process. The results of this analysis are as follows:

• Equity Curve: The equity curve showed a favorable upward trend.
• Profit Factor: A respectable profit factor of 1.64.
• Trade Volume: A significant number of trades over the two-month analysis period.

However, it is crucial to note that the pattern demonstrated no predictive ability and, if traded, would have resulted in substantial losses.

Overfitting Concerns

The intent of this exercise was not to exhaustively analyze millions of combinations with machine learning but to assess whether my initial observations were influenced by overfitting. This highlights a key point: overfitting is a common challenge that can impact both advanced algorithms and manual analysis. With experience and ongoing learning, individuals can develop strategies to avoid or mitigate these issues.

I have included the full equity curve, derived from backtesting and out-of-sample (OOS) validation, to provide a comprehensive overview of the findings.