Self-trained MA cross! - page 4

New comment
 

A new idea, kind of...

Although I have been lurking in these forums for awhile, this is my first post so I hope its a good one.

The basic idea of my approach, is to define an ideal ma cross based on the market conditions, market conditions defined by the gearing function (using the language established by snow leopard).

I would do this by importing data to excel(csv data works well) and use a vb program to cycle through the data placing trades based on a variety of ma crosses. The program would record the trades as profit, ma's involved, and the value of the gearing function when the trade was placed.

Hopefully at this point, if you were to graph the profits of a particular ma against the gearing function, there would be a nice a bell/gaussian curve. This would be a good check, but the next step would be to run a program through this data and select the best ma cross for some given range of values of the gearing function.

The decision function(we're trading now) would work as any ma cross ea would, except that the values of the slow and fast ma's would be looked up in the data provided above, according to the current value of the gearing function.

Now, what should we make the gearing function? A good example of what we should expect, although perhaps not a practical choice, would be volume. While volume itself is very volatile, you could calculate a sort of moving average taking data from previous bars, the same time on previous days, the same time on previous same day(current time - n * 1 week), or some combination of the three.

While the conditions of any given ma cross would stay the same, they are applied to a moving picture of the market. Plus those conditions can be refreshed whenever new data turns up and the range of data you use for that optimization can be however long you want it to be.

^^^ That is alot of writing, my apologies. See the attached text file for a discussion of alternative options for the gearing function.

Files:
gearing_function.txt  3 kb
 

Thanks Boss !!

igorad:
Hi humble_trader,

I think you can try this indicator https://www.mql5.com/en/forum/174228

Also try to use new one.

Thank you Igorad.

 

Self Adaptive indicators and Parallel Functions

Hello CodersGuru,

Take a look at Clayburb's website he talks about this and gives sample tradestation code on this subject plau there is a nice powerpoint presentation that he gave at the Tradestation's world Expo

http://www.clayburg.com/

Enjoy

EK

 

Thanks!

Emerald King:
Hello CodersGuru,

Take a look at Clayburb's website he talks about this and gives sample tradestation code on this subject plau there is a nice powerpoint presentation that he gave at the Tradestation's world Expo

http://www.clayburg.com/

Enjoy

EK

Thanks for the announcement EK!

 

Codersguru,

Can you please look at your PM? I need your help. Thanks.

 

maybe this can help the trainer... ATR Values using multiple ''yesterday close ''DAILY BAR'' and todays ect ect.. if your system can compute what values coincide with the winning ma cross setup, if you are running multiples eventually after enough time youll get a good average for the EA to decide which cross to use depending on range values of atrs.. as of now i do look at daily closes and 28 and 5 day atrs on the same chart level ie they look like overlapping MA 5 pasted over 28 like you would do with MAs on say on a macd ''112 works great on macd''. ill repost after i can figure out a good way to describe what im getting at.

 

Hello codersguru,

I am new in this forum, but I like this idea about "Self-Trained MA cross" EA very much! If this really can be done, this must be the way to go for MA cross EA in the future.

Are you still considering this idea? What can we do to help you?

 

Just found this and thought I may add my own approach.

Firstly - I use some phase detector to determine frequency (periods)

This then can be used as an input into a MA (I sometimes use EMA's but mostly SMA's because then it's easy to determine the lag.)

I also use a phase shifted wavelet (on the same above periods) and add this to the above MA.

In this way, I create some kind of "MA" that is phase shifted onto market price/position (or with whatever percentage lag I may decide - for instance to create MACD's etc.)

Main point is - because the frequency (periods) can be determined on every bar, the MA is controlled by the market data directly...

 

That's can help.

Hello,everyone!

I get this EA from www.mql4.com.It's named as self-trained adviser.I think, it's does'not use MA cross, but imitating buy and sell, stored it and used it for determining real buy and sell. If you interested, I can translate manual in the code.

Best regards!

 

A self trained MA cross expert doesn't seem like it would be that hard to do. you would have to run two terminals one for optimizing and then the other to run the EA. You can optimize every number of set bars, or every bar if u use a higher time frame like 4h, maybe 1h if you limit the range. Or you could optimize it everyday at a specific time.

I like crossover strategies too Codersguru but I disagree with whoever said u have to keep reoptimizing in order for a crossover strategy to continue to work, its usually not the case.

and self training might actually be a bad idea in the end.

123456789
New comment