Backtest period is too short for the EA
Geko is here.
This series explains how to create a portfolio.
This time, I will explain my own criteria for excluding portfolios.
The exclusion criteria are as follows.
- Items with negative forward performance
- Items without a backtest
- Items with a short backtest period
This time,
2. Short backtest period
is the topic.
Gogo-jan listed EAs have specified backtest criteria.
（See below）
There are various opinions on backtest periods.
- From around 2007, including the Lehman Shock volatility, up to the most recent
- Gogo-jan method (over 7 years)
I currently base it on more than 10 years up to the most recent period to align with the current market environment.
What viewpoint one holds may depend on the developer, but if a backtest period is shown that is too short, I will not include that EA in my portfolio options.
What is "too short"
is
- backtests such as one year or one month
- or only a few months excluded from the most recent data
such backtests.
I believe these backtests aim to show only the good parts and hide the bad parts.
Whether or not the presenter has that intention, I become skeptical.
If doubts arise about the data, you have two options: either make efforts to dispel the doubts about the data, or exclude the data's underlying program and seek a new program.
Which option would you choose?
