Do you believe in EA trading?

Hello all

I seen some performance for EAs is exceptional in the Automated Trading Championship 2008, some have made too much money from a starting $10,000. But I am still feeling unsure about the full thing, I don't know if I should trust EAs after all . The trading contest I think last for 2 or 3 months, but is this enough time for you to trust that an EA is good enough to trade for you for along time? Do you think the contest should run longer?. I don't know too much about the winning EAs in the past contests but it should really be interesting to know if they are still performing as good as they did in the past. Tell me what do you think and any experience you had with EAs that was successful. Do you believe in backtesting? I don't anymore. Personally I tried few EAs in the past that failed me. some of those I also paid money for and now I don't really believe in paying any more . I am now trying to learn more and I think this place has too much to offer.

On the MQL boards there have been messages from a person/company that has invested in some of the past winners. Interestingly enough, the end result was that even the winning EAs couldn't make money over the long term. As I recall, the drawdowns were just too much and they all crashed & burn. One of the EAs, supposedly, lost well into the six-figures in a single day and completely burned out.

Bottom line - winning an EA contest is equal parts skill & luck. If you can design an EA that happens to avoid the bad periods or drawdowns during the contest, then you might just win a very short contest.

It's interesting you say that, so after all we still need to keep an eye on how the EA is performing, and maybe do monthly optimization and checks on the method. if possible maybe write EAs that also watches the news calender and avoid trading during important news,
Thanks

I also think that the contest is won more through a bulldog attitude than a truly crafty EA. It's like the concept of the Matrix. It is a system, and within the system there are certain limitations. If you know the limitations you also know how to defeat them. Otherwise, EA's I don't have any faith in.

Now, a properly sorted stand-alone box running "black box" style on a dedicated connection and custom built platform underneath, definitely. Even then though, they often times require the human element in order to truly excel. To me, I see EA's as the half-way mark towards the true top.

definitely trade ea's ...would not have it any other way. The trick; however, is to have an ea that can shake rattle and roll... believe me, I have been through thousands and programmed many more.......very, very few can do what you can see and even then the money management has to be at peak.
but... if you have the time...go for manual...be prepared to spend hours, and hours at a computer.
Also, learn to program. Its not rocket science. It takes a few months but the rewards are great.

4XWeezal,

I would certainly be interested in learning more about programming so I could dabble in the field of EAs. Could you please suggest a good place to start, especially for a beginner like me. Thanks.

I am still learning but I have picked up many skills from the articles below,
http://articles.mql4.com/404
http://articles.mql4.com/417
http://articles.mql4.com/466
http://articles.mql4.com/516
http://articles.mql4.com/536

Buff: Microsoft has free software and free tutorials on how to use it. Most everyone has a use for Excel, so learn some Visual Basic (VBA). Then get into C++ for the MT4 platform. No use learning C-sharp and dot-NET yet unless you are going to use Ninja.
ps:
training
http://msdn.microsoft.com/en-us/beginner/default.aspx

download
http://www.microsoft.com/express/product/default.aspx

one of the richest dudes in the business makes money from a black box system... obviously not a metatrader EA though...

forget his name but he's a Serb living in London and I think it is futures... nice pdf interview on the guy floating somewhere on the internet.

He made some technology to help exchanges or something improve their efficiency, they said no... so he decided to "use the power" himself.

ok, well searched my computer and whadda ya know? found the pdf... enjoy

buff, let me know if you need any help. (I am a self trained coder -- however----so a lot of folks can outthink me on the tecky side.) I read through the MQL development course when I started. There are bookoo sites...It just takes time to internalize. Also, I personnally never did learn to program indicators because there is a zillion on the market that can be added to ea's. I can't add to anyones thoughts on those. ......Also, I would suggest modifying developed ea's that have a number of bells and whistles built in. I went through about a thousand or two and could only find 1-2 that were close to what I wanted. -----You may want to do the same.
For example: I have 5 levels of money managment built in my ea. The great programs; probably have 8-10. The most I have seen developed on this forum are 2-3. Also, MQ4 isn't always perfect in monitoring the market...you need to take that into consideration as to what type of complexity you want to build in.

Look for the ones that are in the top 5 or 10 each month, for a year or 2. Then if they also work in a different kind of year like low-volatility 2006, then you found yourself a winner

But... EAs can be fun... they can make you a lot of money quickly... as long as you treat all initial capital as gambling money.. i.e. you don't care if you lose it. Then you should pull your money once you made a killing (if your account hasn't been wiped out yet), because if you leave it in any longer, it probably will get wiped out soon... I'm talking about fun overleveraged EAs... of course you can make a really conservative EA, but that wouldn't be much fun.

Like many things in trading, EA's are blown way out of proportion. Before I give my biased opinion of EA systems, let me state that I have second hand knowledge that informs me that there are large market participants out there who use EA's in the market, but it is not as simple as what you see portrayed on this site, and sometimes the goal of these EA's is not even to make money...but nonetheless, EA's are used successfully in some respect today.

Do I use an EA? No. You see, a successful manual system identifies and captures an edge in the market and capitalizes on it. Building such a system is extremely difficult, and therefore the percentage of successful manual traders out there is incredible small, and to my knowledge getting smaller. But, to create a successful EA, you must not only already have such a system perfected, but then proceed to create a very specific and detailed algorithm which directs and governs a very limited computer mind to trade for you. At this point, it is my belief that the success rate of such a system drops dramatically lower, due to the complexity of programming and the ever changing diversity of market influences. So in all honesty, it seems like trusting an EA to trade for you complicates and lowers your chance to succeed.


Even so, EA creation has grown dramatically lately, and new traders looking to make millions are attracted to the idea of EA's. The idea of an EA appeals to almost every psychological hinderance that new traders have to overcome to succeed. For example, it is appealing to think that a computer that trades for you takes away all of your responsibility, ie the EA recipeint is not responsible for a failed trade anymore, it is the computers fault. Or how great does it sound to not have to do anything and still make loads of money. I mean, why bother with dealing with time, emotions or discretion based trading when a computer can trade for you! Well, the alternative is to simply nut up and fucking take the time and discipline yourself to trade like a professional. In the end you will have gained more as a person than as a trader.

So, as mentioned above, I am biased in this area. I have never attempted to use an EA to trade, but I have heard COUNTLESS accounts of people putting all their faith in such a system and losing everything they had. Of course this can be said for manual trading as well. However, I would pretty bluntly state that it is impossible to succeed with an EA until you an successfully trade manually first. That is a prerequisiste. Only then could you even begin to truly understand the boundaries and pitfalls of the system being traded.

I would say the problem with most EAs is that compared to a skilled trader manually trading a similar way, the EA will usually have a larger drawdown because EAs can't use discretion.

Having a larger drawdown, means that each trade the EA places at the same size as the manual trader would inherently be more risky. That said, the EA won't panic when it makes a bad trade... but then again, neither would a skilled trader

There is no universal trading strategy that would make profit in any market conditions. For example a trading system that makes profit in trending market will lose money in ranging market. Only a human can identify the difference in market condition and adjust the use of software accordingly. However, it seems like a holy grail since a machine does not have human emotions like greed and fear. I think the choice is just subjective, according to how you feel more comfortable and confident.

You might want to take a peek at a thread I started some time ago (I decided to shut it down but it's still available).

http://www.forexfactory.com/showthread.php?t=134827

I too, continue to be intrigued by the concept of EAs, if not the actual implementation thereof.

The conclusion I have come to so far is that EAs should be relied upon for what they appear to be ideally suited for - that is, the mindless slog of monitoring the minutiae of the march of market data. Something we as as humans don't do very well and get easily distracted from. But when it is time to exercise judgment based upon market and trading experience such as past decisions and results, humans seem to be better suited to do that task.

So my feeling is to use EAs to monitor the boring march of streaming data and then have them signal us when a decision has to be made, based upon a set of circumstances that we have predetermined by our design of the EA that we want to be notified of.

Turning over my trading and my account to an EA without my involvement in pulling the entry trigger is something I am loathe to do, if for no other reason than that the market conditions the EA was designed for may have now changed.

In short, I'm saying use the EA to supplement and enhance my powers, not to replace me at the trading desk.

Just stopped by, couldn’t resist posting on this topic.

"You see, a successful manual system identifies and captures an edge in the market and capitalizes on it. Building such a system is extremely difficult, and therefore the percentage of successful manual traders out there is incredible small, and to my knowledge getting smaller. But, to create a successful EA, you must not only already have such a system perfected, but then proceed to create a very specific and detailed algorithm which directs and governs a very limited computer mind to trade for you. At this point, it is my belief that the success rate of such a system drops dramatically lower, due to the complexity of programming and the ever changing diversity of market influences. So in all honesty, it seems like trusting an EA to trade for you complicates and lowers your chance to succeed."
Nicely put.

To automate my own approach, an EA would need to be able to intelligently identify key S/R areas, and analyze chart and candle patterns across multiple time frames. Not a trivial programming task. But not insurmountable: by repeatedly sifting enough chart based setups as “valid” or “invalid”, one could gradually construct and hone a fair approximation.

However, most existing EAs tend to automate indicator-based methods, which greatly reduce their probability of identifying and exploiting a real inefficiency. If their success is overly dependent on the way in which the parameters have been fit to the test data, then the likelihood of long-term profit is further reduced. It goes without saying that ease of programming has no bearing on profitability.

The trading competitions prove nothing. Over a few weeks the number of trades made is going to be statistically insignificant, making the outcome a lottery. It's simply a vehicle to allow MetaQuotes to showcase MT4.

EAs teach us that profitable trading requires much more than just MM plus mindset. Every EA employs strict MM rules, and trades a plan consistently, accurately, tirelessly, emotionlessly. But very few EAs are long-term profitable. Hence profitability must ultimately depend on whether or not a trader has a methodological edge.

Interesting opinions about EAs, so to be even near trusting an EA we need:
1- An EA that doesn't trade during news release. although some traders only trade the news this means if the EA trades during news time it need some algorithm to understand how a human trader would trade during news release. ultimately if we want a sophisticated EA we need to even feed it some codes to make it analyse a news event before trading which is very difficult. Yes we have access to calendar events this could somehow be fed in the EA to stop it from trading through a news event.
2- If we trust an EA we need to keep an eye on it, I don't think I can follow the rule that says "Set and Forget", I think the second part of this statement need changing .
3- EA must also analyse Support and Resistance points for a currency.
4- It must also analyse different time frames before taking a trade.

I am not sure If I missed anything else.

Thx for the article - do you by any change know when it was published?

Regarding the trading competitions then forget about them. The results suffers from "survivorship bias" as you have thousands of trading scripts trading without any real risk. It is almost certain that a random algorithm will show an outstanding performance during the competition period if given enough trials (that is, if thousands of random algorithms are submitted to the contest). Another thing is that those who have really effective algorithms will probably not show off in a competition where a rather limited amount of money can be won (compared to what the algorithm can rake in itself).

Regarding the concept of "EA"'s, there are very successfull quant funds and propertiary trading firms which operates completely automated strategies, but they are certanly not using Metatrader and they don't trade "when the blue line crosses the red line" types of trading systems which are mostly found on forums like FF.

People naturally tends to relate to what they allready know. Therefore discretionary traders often states that "you must be able to trade dicretionary before you can automate" or "a computer can't trade as effective as a human". But this is biased by the mindset you develope as a discretionary trader. The arguments can certainly be turned around as there are automated strategies which would be impossible for humans to trade. And some quants don't even belive that discretionary trading is possible.

Go to places like http://www.nuclearphynance.com if you want some clues about how quants operates and what kind of strategies are being used. But be prepared for a different kind of world.

Your getting into a specific strategy there. Also time frames are to make charts useable for humans. The EA has no need of time frames.

An unattended EA needs to following IMO:

1. To continually monitor the status and quality of the connection to the server
You may want to set a threshold where it stops trading if the connection is unstable.

2. To continually monitor the quality of the data
Live Price Data. This needs to be checked for bad ticks, gaps in the data, changes in the spread etc. The data needs to be validated and corrected. For example, a short period of wide spread could distort analysis based on bid prices, mid prices may be better. Ideally the EA should have a second source of live data for reference.
Historic Price Data. On-line history files in MT4 can be full of gaps and errors. You may need to maintain off-line historic data.
External Data. Things like economic announcements and holidays can be imported into an EA. This data must be sanity checked before being acted upon.

3. To continually monitor the status of the trades
You can't just send an order and expect it to work in MT4. There are loads of possible errors, requotes and things that can go wrong. All orders must be checked to make sure they have executed as expected. In addition, open positions must be checked regularly to check for unexpected opening or closing of positions. All changes to positions must be logged with reasons.

4. To log everything
All events must be logged. Connection events, data events, trade events. Everything.

5. To alert via email/SMS when problems occur

6. To report statistics on all functions of the EA
What slippage to you get at different times of day?
What is the % uptime of the server connection?
What is the slippage on stop orders?
What is the typical delay in executing market orders during busy periods?
You need to make sure the EA gives you enough data to answer these sorts of questions fairly easily.