Forward performance is negative EA
Geko is here.
I am explaining portfolio creation in this series.
This time, I will explain my own criteria for excluding portfolios.
First,
- EA with negative forward performance
This may seem obvious, but there is one thing to be careful about.
That is the period of forward measurement.
To be abrupt, have you ever seen Gogojan’s “System Trade FX” top page?
It looks like this on the page.
https://www.gogojungle.co.jp/systemtrade/fx
On the upper part of this top page, there is something called “Real-time Operation Ranking,” right?
Some people use this ranking as a reference when selecting EAs, but the ranking display has an initial period value of 1 month, doesn’t it?
In other words, the ranking is based on the measurement period of the most recent one month.
Whether to refer to it or not is a matter of preference, but from the perspective of an EA developer, I believe the period should be measured for at least one year.
The reason is connected to my own EA development process.
In my case, when developing an EA, I use a criterion of finding strong performers from over 10 years of backtesting.
Even if an EA that performs well is completed smoothly, it is realistic that it cannot continue to profit on a monthly basis throughout the backtesting period.
By running day-by-day and month-by-month profits and losses and operating for over 10 years, I aim for a reasonable level of profit, and I build EAs through repeated validation in backtesting.
From the perspective of this development process, short-term, tournament-like operation rankings are not very useful for reference.
A good-performing EA in the past few months, and whether it will continue to perform well in the future, I want to determine by comparing more than 10 years of backtesting with over a year of forward results.
