Idea/Strategies for Modular EA Architecture

Hi all,

First a bit of background information.. I'm a graduate software engineer who has taken an interest in Forex over the last couple of months.. Mainly reading Babypips, a few videos and playing with a demo account. I got bored after playing with a demo account after a couple of weeks and thought I'd be better off going live.

I've been going live for a few weeks now (only around $250 outlay, about $40 in profit) as I feel it's better to trade real money so your emotions come into play. My main issue is, I'll trade larger lot sizes (by large I mean 0.4) on trades I'm more sure of, and I'll keep losing money on the smaller 0.01 sized trades when I'm just 'goofing around', or not in the right mood.
This, along with an interest in programming, has given me the idea of writing an Expert Advisor which will strictly stick to logic without the anxiety and other emotions which affect trading decisions.

The most frustrating part is MQL4 only being procedural, when I'm used to Object Oriented Programming. The code syntax and constructs are all easy to pick up (I've had plenty of experience with Java and some C/C++), the biggest hurdle is all the MQL4/Metatrader specifics.

Anyway, my idea is to make a rather modular trading system which is adaptable and can change strategies as required.

At the moment, I've divided it into the following system components:

MyExpert - the main EA
EquityManagment - actually opens trades, position sizing, setting take profits and stop losses
StrategyManagement - Imports and uses various strategies to make trading decisions
Strategy - Makes use of indicator(s), candlesticks or some other data representation to give bullish/bearish signals.

I've got the basics set up. My idea is to build several strategies, each of which may have varying probabilities. Strategies may contain support and resistance levels calculated, for example. Support and resistance found on higher timeframes would have a higher probability. The StrategyManagement component will make use of varying Strategies and their probabilities in order to make an overall trading decision.

Each strategy has a set of state transitions. Initial state being None, second state being a bullish/bearish signal.. There then may be several other states which exist until the end state where a Strategy returns a confirmed signal. Strategy Management is then able to look at various signals, assess their probabilities together (bullish probabilities vs bearish probabilities), and make a decision. StrategyManagement will not consider a strategy's signal until the strategy CONFIRMS that signal.

If a Strategy gives a signal (within an internal state transition), it has a Time to Live (TTL) measured in candles, in which the Strategy must confirm its signal or else be invalidated. I'm hoping this will allow for testing of multiple strategies, and only making trades based on the probabilities of these strategies, which will be skewed by the timeframe in which the strategies occur. That is, if strategies give a bullish signal within a certain timeframe of one another, the trade has higher probability.

Eventually, it would be nice to store historical data of previous trades, the strategies used and their subsequent outcome such that the StrategyManagement component can alter the probabilities of each strategy in order to help the system become adaptable to the market, both with filtering signals, and equity management.

My biggest issue is deciding on strategies (I'm thinking so far maybe support and resistance for range-bound trading, DeMark trend lines, MACD..) and being a newbie to forex, I'm unsure whether this system sounds suitable.

Would any experienced traders out there be able to advise if this system sounds feasible? My biggest concern is receiving mixed signals and having issues filtering them out based on probabilities, using too many strategies, the EA not placing trades because it waits for 'the perfect trade'..

However, I'm trying to be careful with my trading. Initially I found it far too easy to experience 'analysis paralysis' and pile indicator upon indicator on my charts.. Now I'm trying to go back to basics an build from there.

Thanks,

Mark

Sorry to tell you this, but if you are using or thinking of using a EA, you are not doing it right.

Best to stick with a manual strategy and trade higher time frames..

No, developing an EA.
I didn't say anything about what timeframes?..

lets start with version 1.0 of your EA.
http://www.forexfactory.com/[IMG...v_25/begin.gifhttp://i1103.photobucket.com/albums/...v_25/begin.gif
try to make profit based on 30 minute chart. just top to bottom

lets back test it, would it be 100% sucessfull


with 1 LOT standard it can make a decent amount of profit.

-X-

I believe one simple strategy using H4 would be sufficient.

http://i1103.photobucket.com/albums/.../Nov_25/h4.gif

Thanks for the feedback.

I'm hoping by making it modular, I can add/remove strategies easily. This way, I can test different combinations, etc. I suspect relying on too many strategies (or perhaps the wrong combination of strategies) would result in the system waiting for 'the perfect trade', which may never or seldom happen.

Scalability is one of the most important considerations.. The ability to keep it simple, whilst allowing for increasing size and complexity.

Thanks again.

Is it possible to create the subroutine which you can program in object oriented and then it communicate with main program using DLL.

I allways want my own EA but I am not good at programming. The best I can do is I can help you with documentation or making the flowchart of all the code.

-X-

Yea, I was thinking of coding external DLLs, however my main concern was the extra overhead incurred by marshalling/unmarshalling.

So far I've taken a pseudo Object Oriented approach whereby I treat each include file as a class. Functions and private variables are marked as ClassName.variableName, or ClassName.SomeFunction().

It seems to work okay so far.. The worst part is attempting to find syntax errors, as the compiler will join all files together, and you get heaps of cascading errors which can be a pain to try and fix.

My biggest issue (and I suspect this has more to do with being new to forex as a whole) is developing a set of strategies (or maybe just one or two) to use and backtest.

If I were to use multiple strategies combined, my biggest issue would be attempting to consolidate the various strategies into a single bullish/bearish signal, which is where my idea above came from -- give each strategy a 'time to live'. After this period, the strategy is no longer valid and is deemed to be a false signal.

However, if multiple strategies occur within their time to live period, we (hopefully) will have an overall bullish/bearish signal. I'm not sure how well this approach will work, as it seems quite possible to me the possibility of mixed signals (strategies giving both bullish/bearish signals).