New Approaches to Creating EAs | Forex Factory

Options

Permalinks
Print Thread

Similar Threads

How can I learn (for free) to create EAs for this new build? 1 reply

MT4 - Creating a new price stream 0 replies

Help with creating a New Indicator 5 replies

Creating New Indicators 10 replies

New Approaches to Creating EAs

  • #1
  • First Post: Edited 12:13pm Nov 25, 2012 12:18am | Edited 12:13pm
Hi All,

My intro - skip to next paragraph for the question.
As my name suggests I'm a discretionary trader, and I have never actually used an EA. My personal opinion is that far too many traders in retail spend too much time looking for killer systems and not enough time actually trading. Some of the comments I read in these threads (I mean all over FF) truly astonish me. Having said that, I sometimes get insights into the market that I wish I could code as a signal, but alas, I am no coder and don't have the time/patience for it. I have a lot of patience for the forex market however, and I absolutely fell in love with it over three years ago, which gave me the patience to trade demo accounts for over two years before going live. In the beginning I focused on pure tech., but before I went live I developed approaches based on data outside of price action as well.


My question to start this thread is about different approaches to risk-management in trading systems. What about only taking trades where a certain risk/reward ratio (SL/TP levels) is statically probable according to a further algorithm, and build that as a sub-routine for a wide variety of signal generators? Or just a few based around the same logic as the algorithm. I'm thinking something based around ATR and volatility indicators (indexes).

Also, does MQL4 have the ability to crunch data by any time-frame? For example, if you wanted to count the difference of H minus L price, including spread, every 10 seconds, and maybe include volume into the mix? Would you need some kind of external script like Perl for this?

And here's a truly wild idea: feed data from a very successful manual trader and use the computer to find correlations and consistencies within all known data points, and code them. Update the data every 28 trading days and compare the new results, correlate them further, etc.
That was a joke. For a trader, I have a very good sense of humour.


My hope for this thread is it becomes a space where people can express their more abstract ideas about the behaviour of the market, and how we might make it relevant, and to generate ideas to share with other, more specific threads.

Michael
  • #2
  • Nov 25, 2012 7:20am Nov 25, 2012 7:20am
Quoting DrDiscretion
And here's a truly wild idea: feed data from a very successful manual trader and use the computer to find correlations and consistencies within all known data points, and code them. Update the data every 28 trading days and compare the new results, correlate them further, etc.
That was a joke. For a trader, I have a very good sense of humour.


My hope for this thread is it becomes a space where people can express their more abstract ideas about the behaviour of the market, and how we might make it relevant, and to generate ideas to share with...
http://www.forexfactory.com/showthread.php?t=393968

You may want to read this thread to get an idea of what is possible, but more important what effort is involved. This wasn't done in MQL4, but that is not to say that it couldn't be or that his approach couldn't be adapted to the retail forex market.
  • #3
  • Nov 25, 2012 6:24pm Nov 25, 2012 6:24pm
Quoting DrDiscretion
Hi All,
My question to start this thread is about different approaches to risk-management in trading systems. What about only taking trades where a certain risk/reward ratio (SL/TP levels) is statically probable according to a further algorithm, and build that as a sub-routine for a wide variety of signal generators? Or just a few based around the same logic as the algorithm. I'm thinking something based around ATR and volatility indicators (indexes).
Some like fixing the level but adjust over time to compensate to recent trading conditions.

Others use a variable to dynamically adjust the levels to suit the current market. ATR is one popular variable for this... but remember, you run the risk of that variable later becoming the weak link in a system, since now not only do you have to worry about your signal's edge fading away as the market changes, but the variables ability to adjust risk fading away as well.

In either case, the best way is always what proves to be best over thousands of trades and a time period stretching into years. Regardless of what you think is a better way of doing it, prove it first through statistics before committing to any one method.

Quoting DrDiscretion
Also, does MQL4 have the ability to crunch data by any time-frame? For example, if you wanted to count the difference of H minus L price, including spread, every 10 seconds, and maybe include volume into the mix? Would you need some kind of external script like Perl for this?
Short answer, it's not easy, but yes.

Longer answer, you can create an array and fill it with all kinds of data streamed from the quotes, then doing whatever kind of organizations or operations you can dream of, all from within MQL4.... However, this isn't as easy as just referencing bars and price levels on the chart your EA runs atop directly, so expect to roll up your sleeves and get messy coding here.

Quoting DrDiscretion
And here's a truly wild idea: feed data from a very successful manual trader and use the computer to find correlations and consistencies within all known data points, and code them. Update the data every 28 trading days and compare the new results, correlate them further, etc.
That was a joke. For a trader, I have a very good sense of humour.


My hope for this thread is it becomes a space where people can express their more abstract ideas about the behaviour of the market, and how we might make it relevant, and to generate ideas to share...
Good luck and good trading!
FXGears.com
  • #4
  • Nov 27, 2012 4:42am Nov 27, 2012 4:42am
Thanks both above for responding so generously. Codemeister, thanks for the link, I'll read in depth when I get time tomorrow, it looks exactly the sort of think I'm talking about.
As I say, I'm no coder, but I'm sure there must be ways for independent traders to make more effective EA's than I've seen around forums so far.
Like trading itself, it's all too easy to get bogged down trying to solve specific problems, when a more general and intuitive response might be called for.
One thing I'm sure of is that making good EA's simply using MT4 is damned hard!
While every indicator must have some kind of back-test, wouldn't it be better to back-test the known correlations of the real market first, before trying to filter signals with standard indicators?
Everywhere on this excellent site I see traders with ideas, desperately trying to find coders, and EA nuts desperately trying to crack the old nut!
  • #5
  • Nov 27, 2012 12:06pm Nov 27, 2012 12:06pm
Micheal, you do need to read the article to appreciate the obstacles you are facing.

As far as back testing, there has been a number of good discussions in other threads about it. It is its own topic, but my feeling is that back testing has little value especially the way it is conducted using MT4. The data quality is not good enough to expend much effort on and acquiring good data is an expensive proposition. Your approach is a little different than I have seen discussed, but I think it still requires quality data. Enough said.
Quote
Everywhere on this excellent site I see traders with ideas, desperately trying to find coders, and EA nuts desperately trying to crack the old nut!

Let me give you my perspective on the ideas for automated trading systems I see. I have read lots of poorly written, (not just lack of english skills) half thought out ideas. Some of them look like they took a few hours of research and threw their conclusions into a request that took a few minutes to write. The rest are bunch of ideas that are recycled variations of ideas that have been proposed many times (instead of combining an RSI with a Stoichastic, they propose using Heiken Ashi with an RSI). Some of the few, good proposals I have read come from traders who don't realize the difference between a manual (discretionary) system and a truly mechanical system. Perhaps we move in different circles or I have missed the wheat amoung the chaff. If you can provide links to a few of the good proposals you have seen lately, it would give me some relief from the usual dreary stuff I see.
  • #6
  • Last Post: Dec 13, 2012 5:11am Dec 13, 2012 5:11am
Quoting CodeMeister
... bunch of ideas that are recycled variations of ideas that have been proposed many times (instead of combining an RSI with a Stoichastic, they propose using Heiken Ashi with an RSI). Some of the few, good proposals I have read come from traders who don't realize the difference between a manual (discretionary) system and a truly mechanical system. Perhaps we move in different circles or I have missed the wheat amoung the chaff. If you can provide links to a few of the good proposals you have seen lately, it would give me some relief from the...
Well you certainly hit the nail on the head there, but I'm afraid I can't refer to any good ideas from recent memory either.

The problem I'm having lately is where to find good market feeds for certain kinds of data, like Non-Commercial positioning, and the ability to aggregate different markets with - not back-testing - but a regressive analysis with correlation results etc etc.

I realise that I could buy a Bloomberg Box, but they're expensive!

The online forex community has come so far, and there is more good info around than ever before, but I haven't found anywhere that will take that next step.

ps. As a long time fan of Google I have to say: Google finance sucks. These guys are DATA experts, that's what they DO! They've squeezed out email, online video, Microsoft, data storage, online advertising, mobile apps, you name it, but their financial data is almost useless. That's my rant.
0 traders viewing now