Backtest Evaluation
Using AI to create EAs and strategies.
Since it runs in the free tier, the focus of backtesting is on MT4 EAs.
I would like to evaluate the backtest itself by leveraging the detections I have worked on with discretion.
Table of contents
Backtest results
First, note that the settings are not the same
What I want to convey
In conclusion
Even if a long-term backtest for 10 years shows good results, it does not mean you can trust it.
I still tested the strategy I created by changing the settings year by year, for four years of backtesting.
Backtest results
Lot size fixed at 0.1
Trailing stop: changed in three stages
Only the maximum and minimum values will be written.
Number of trades
151–294
Win rate
57%–63%
Total profit
-$149 to $370
First, note that the settings are not the same
There were years when no matter how you tweak the settings, it did not become profitable.
Mostly, some settings were good.
And dates are quite important; when there are Mondays around year-end or New Year, spreads widen and gaps open wider.
In January, by excluding the first week and in December the last two weeks, performance changed quite a bit.
Even when total results were positive, half of them were losses.
ect….
Performance worsened simply by excluding New Year holidays.
Possible reason: when gaps opened, there was a high likelihood profits could have been earned there.
What I want to convey
Being profitable in the long term does not mean you can rest easy at all.
I think the most important thing is to understand the logic (when to take a position) and be confident in its operation.
In conclusion
EAs and discretion can coexist.
When volatility, carryover, or other sudden changes are expected, whether to keep a position or not, or to avoid trading for that period, requires a certain amount of experience and varies by preference.
Therefore, I think the results do not vary greatly based on the user's mindset or personality.
Having done discretionary trading, automation has many advantages: reduced monitoring time, significantly lowered stress, and decreased development of bad habits like overtrading, among others.
In backtesting, the aim is not to achieve strong results, but to identify which parts of the settings cause significant changes in performance, how wide the range of settings is, and to estimate overall performance and trade count. I intend to use it as a valuable tool for understanding those aspects.
Since it runs in the free tier, the focus of backtesting is on MT4 EAs.
I would like to evaluate the backtest itself by leveraging the detections I have worked on with discretion.
Table of contents
Backtest results
First, note that the settings are not the same
What I want to convey
In conclusion
Even if a long-term backtest for 10 years shows good results, it does not mean you can trust it.
I still tested the strategy I created by changing the settings year by year, for four years of backtesting.
Backtest results
Lot size fixed at 0.1
Trailing stop: changed in three stages
Only the maximum and minimum values will be written.
Number of trades
151–294
Win rate
57%–63%
Total profit
-$149 to $370
First, note that the settings are not the same
There were years when no matter how you tweak the settings, it did not become profitable.
Mostly, some settings were good.
And dates are quite important; when there are Mondays around year-end or New Year, spreads widen and gaps open wider.
In January, by excluding the first week and in December the last two weeks, performance changed quite a bit.
Even when total results were positive, half of them were losses.
ect….
Performance worsened simply by excluding New Year holidays.
Possible reason: when gaps opened, there was a high likelihood profits could have been earned there.
What I want to convey
Being profitable in the long term does not mean you can rest easy at all.
I think the most important thing is to understand the logic (when to take a position) and be confident in its operation.
In conclusion
EAs and discretion can coexist.
When volatility, carryover, or other sudden changes are expected, whether to keep a position or not, or to avoid trading for that period, requires a certain amount of experience and varies by preference.
Therefore, I think the results do not vary greatly based on the user's mindset or personality.
Having done discretionary trading, automation has many advantages: reduced monitoring time, significantly lowered stress, and decreased development of bad habits like overtrading, among others.
In backtesting, the aim is not to achieve strong results, but to identify which parts of the settings cause significant changes in performance, how wide the range of settings is, and to estimate overall performance and trade count. I intend to use it as a valuable tool for understanding those aspects.
× ![]()