I thought I created the ultimate EA, but...

Ahmet Metin Yilmaz 2022.01.20 17:10 #1

trevor88:

During the last month I've created an EA which is able to find the best historical combination between a pattern and all its relevant indicators/values.

Here is an example:

We give the EA all the "morning star" formations for the EUR/USD between 2010 and 2020 as input, and it returns which combinations (moring star + indicators/oscillators/prices on the graph) gives the best preformance.

As result we find that the "morning star" gives us a good profit if we buy when 1. the close price of the first candle is higher than the moving average at 48 period, 2. the moving average at 48 periods is lower than moving average at 288 periods, and 3. RSI is above 60.

I thought that since this combination works for the past prices it could have been working for future prices also... but reality is the results are not good.

Basically this Ea is not working... I feel i have no more ideas left. Do you have any suggestion?

then your strategy wrong ..

do you mean strategy is not working or codes of EA not working?

Alexandre Borela 2022.01.20 17:54 #2

trevor88:

During the last month I've created an EA which is able to find the best historical combination between a pattern and all its relevant indicators/values.

Here is an example:

We give the EA all the "morning star" formations for the EUR/USD between 2010 and 2020 as input, and it returns which combinations (moring star + indicators/oscillators/prices on the graph) gives the best preformance.

As result we find that the "morning star" gives us a good profit if we buy when 1. the close price of the first candle is higher than the moving average at 48 period, 2. the moving average at 48 periods is lower than moving average at 288 periods, and 3. RSI is above 60.

I thought that since this combination works for the past prices it could have been working for future prices also... but reality is the results are not good.

Basically this Ea is not working... I feel i have no more ideas left. Do you have any suggestion?

What you did is called curve fitting, you changed the inputs until it fitted the data. Any strategy will look good this way during backtest.

trevor88 2022.01.20 18:48 #3

Alexandre Borela #:
What you did is called curve fitting, you changed the inputs until it fitted the data. Any strategy will look good this way during backtest.

yeah... I tought I could uncover the underlying logic of a chart patter by doing this way. I didn't want to try every single strategy to find the one that works... I still think my idea could be somehow good(!?), maybe I need to take more variables in considerations?

I am open to suggestions :)

Questions from Beginners MQL5 Do you think EA's Machine learning in trading:

Renato Takahashi 2022.01.20 19:25 #4

trevor88 #:

yeah... I tought I could uncover the underlying logic of a chart patter by doing this way. I didn't want to try every single strategy to find the one that works... I still think my idea could be somehow good(!?), maybe I need to take more variables in considerations?

I am open to suggestions :)

Using more variables will overfit more your trade system. We need to have few variables and some criteria on optimization, such as walk-foward optimization.

Maybe if you wanna try your strategy, I sugest to make optimizations in small sample periods (1 month por example)... it is what I am trying with a trading system of mine. In this case, I optimized in one month and run in the next other, even in backtest... doing this settings, I had better results on the EA... I can tell you if real results will be good too later, but you need to optimize your EA every single month (work, work, work... I hope to result in success).

Walk-foward optimization - how Indicators: HI LO Indicator Why some great coders

Dominik Egert 2022.01.20 21:48 #5

Maybe a way to go is to create a strategy that does not require optimization.

Markets are fractal and random. They have a tendency to return. They have nothing to do with the price change you see on charts. - Well sort of.

If you can begin to see where the orders and how price reacts to these areas on the chart, then you will get a new view on what's going on.

Following to give an explanation:

When price turns around, why does it happen?

It's a shift in supply and demand.

Try imagining a market where they sell apples. Try to think the price on your chart is like the over all mean of the current markets activity.

So when does price of apples rise? When there is more demand than supply.

How can this happen? Well, someone bought a lot of apples at once.

Now try to see where would most traders place their stop loss, or exit their positions for profit.

That's the area you are looking for.

Hope this gives you a new view on price.

Take profit/stop loss Universal MA Cross EA Definition of a pin

trevor88 2022.01.22 18:12 #6

Renato Takahashi #:

Using more variables will overfit more your trade system. We need to have few variables and some criteria on optimization, such as walk-foward optimization.

Maybe if you wanna try your strategy, I sugest to make optimizations in small sample periods (1 month por example)... it is what I am trying with a trading system of mine. In this case, I optimized in one month and run in the next other, even in backtest... doing this settings, I had better results on the EA... I can tell you if real results will be good too later, but you need to optimize your EA every single month (work, work, work... I hope to result in success).

You actually gave me a good idea... I'll do the same thing but on smaller periods and try it on future prices