3/22【Backtest Evaluation Method】We explain all items of the backtest image21
I am the Reiwa-era dabu-ru i-i of EA developers.
First of all
EA developers gauge the quality of an EA by
“backtesting.”
So as a user,
to “judge the quality of an EA,”
you need the skill to evaluate backtests.
With that in mind, in this series we will
write about “how to find profitable EAs from backtests.”
【About win rate (%) and loss rate (%)】
Personally, I don’t care about win rate at all.
It is naturally determined by the type of EA.
Also, even with a high win rate, one can still fail with a small, steady loss,
in short, an EA makes profit by managing the net point difference.
The following are
two manuscripts that are planned to be posted in the future.
【Manuscript 1】
To get straight to the point,
an EA only needs to be positive in the net point difference,
so win rate does not need to be considered.
If the win rate is low,
you might think, “This EA will never win!”
For example,
9 wins and 1 loss with a 90% win rate,
average win amount: 1000 yen
average loss amount: 10000 yen
Then
9 wins × 1000 yen = 9000 yen net gain
1 loss × 10000 yen = 10000 yen net loss
A net gain of 9000 yen and a loss of 10,000 yen results in a net decrease of 1,000 yen in capital.
In other words,
the same one win and one loss
will have different win/loss amounts depending on the moment,
so you cannot judge the EA’s advantage from win rate alone.
【Manuscript 2】
I am the Reiwa-era dabu-ru-i of EA development.
When you are about to purchase an EA,
you may wonder about“win rate.”but
however,
choosing an EA by win rate is actually a big mistake.
The win rate of an EA is naturally determined by the type of EA.
--------------------------------
For example
As the performance of an EA improves,
the win rate naturally settles as follows.
-----------------------------
Scalping or contrarian
Win rate… around 70%
Average win:loss = 1:2
-----------------------------
Trend following or swing
Win rate: around 40%
Average win:loss = 2:1
-----------------------------
Anomalies (timed trading)
Win rate… in the 50% range
Average win:loss = 1:1 (slightly more wins)
-----------------------------
By combining these win rates with average wins and losses,
you get a positive profit or loss in terms of net point difference.
-----------------------------
Therefore,
in order to achieve good performance,
trying to increase win rate is inherently difficult due to design constraints.
Scalping or contrarian strategies
enter at well-timed edges,
thus their win rate tends to be around 70%,
trend following or swing
aim for profits during longer trends, so
their win rate tends to be in the 40% range,
anomalies (timed trading) unlike the above, enter around 50%-50%,
so the win rate tends to be in the 50% range.
From my experience so far,
when the win rate is near 70% and the average win is larger,
the number of trades in a period is small,
and there is a high likelihood of overfitting.
That said,
win rate alone does not determine whether an EA is good or bad.
On the other hand, if used properly,
“include EAs with various win rates in a portfolio”
which leads to incorporating various types of EAs,
so you can expect diversification of risk.
Well then!
━━━━━━━━━━━━━━━
【Reiwa-era dabu-ru-i EA Listing】
https://www.gogojungle.co.jp/users/112481/products
━━━━━━━━━━━━━━━
【Reiwa-era dabu-ru-i’s series articles】
https://www.gogojungle.co.jp/finance/navi/series/1700
--------------------------------
https://www.gogojungle.co.jp/finance/navi/articles/64703
https://www.gogojungle.co.jp/finance/navi/articles/64738
https://www.gogojungle.co.jp/finance/navi/articles/65619
https://www.gogojungle.co.jp/finance/navi/series/1701
https://www.gogojungle.co.jp/finance/navi/articles/64723
https://www.gogojungle.co.jp/finance/navi/series/610
【Reiwa-era dabu-ru-i’s Philosophy】
