I have been writing EA's for about a year now, after nearly trading my first account into the ground, I thought automated trading must be the way to go. So I set about learning MQL & the many pitfalls of backtesting...

Anyway, to the point, after backtesting every possible indicator & combination I can think of I just can't seem to find a indicator(s) which gives a signal hit rate of more than about 30%, it does not matter what I do, reverse the signals, use limit orders, sliding losses, breakouts, hedging etc I just can't come up with anything that can make more than about 5-10% a month. I wrote EA's to trade randomly and I got pretty much the same results.

I watch the people in this forum also trying to come up with indicator driven EA's, after a couple of weeks the thread will die as people realize they is no real profit to be made. There is the odd exception like Firebird or the Hans breakout, but as you can see in the MQL trading contest, Firebird is starting to lose as the effectivness of the curve fitting that was done wears off.

I am driven to the conclusion that traditional indicator TA just does not work that well.

There are examples of people making good money in automated trading, but one would have to assume they are using methods which in some way diverge from the 'indicator combo' mentality. So we know that it can be done, but what are we missing...

Is it possible that people writing the EAs or whatever just arent good traders to begin with? They may be excellent programmers but that by no way means they can effectively code an EA that will be able to make a profit. I mean you said you have almost traded your first account into the ground, so wouldnt it make sense that no matter what combination you have tried you havent found anything in backtesting that works?

You may be right, but I doubt it. During the initial trading attempt I was clueless, I mean really clueless. I am now trading automatically and making money, just not huge amounts, so I like to delude myself that I've made some progress. Being an effective programmer does not preclude somebody from learning to trade. I think your missing the point of my post though, is tranditional TA effective in the context of EA trading? if it is effective, why are not more people making money?

i've only been on this site a little while and i know what you mean about threads dying off after a while. perhaps if indicators dont work you can code according to price action, such as candle patterns.

__________________
There is no candle.

If your hitting these levels consistently by autotrading, and your system has relatively low drawdowns, and you can back up these claims with solid reliable back tests and an audited track record over a couple of years then youve probably just become a very rich man.

In answer to your question, I think TA is applicable, my own manual methods are entirely TA based, and they work, but there is an element of discretion, and general awareness of market conditions that are hard or if not impossible to capture into simple logic. I think the main problem with most EA's that I see posted on the forums is that they have very rigid criteria driving them for example IF RSI > 70 AND EMA5 < EMA8 THEN SELL.

When trading manually, for me at least, TA provides more of a reference framework, Im generally looking for a confluence of signals, and whilst I may be looking for a "high" RSI value, If there's strong supporting evidence from other indicators or price action or Im trading off the back of economic data etc I'll use discretion and take the trade, whereas the EA wouldnt as RSI perhaps only read 68. The failure to address this idea of confluence from a cross section of indicator types really dosnt help EA's perform to their full potential. In my experience the tradable patterns that the market provides arnt crisp and sharp enough to capture using simple if then else statements.

I also think the fact that most EA's focus on TA signals from a single timeframe reduce the chances of success quite drastically.

In my experience the biggest single thing that prevents most people developing successful EA's is that they cant actually trade manually, so theyre spending time trying to automate something they know very little about !, the most basic of things we do as traders arnt even considered by the majority of EA's.

Ive seen quite a few people trade EA's successfully, but those people where all traders, and thats probably the real answer as to why more people arnt making money from EA's. Being blunt, what are the probabilities of someone producing a ridiculously simple peice of software consisting of a few hundred lines of code, that can be used by absolutely anyone to double or tripple their accounts every few months. Its unlikely to happen.

"the most basic of things we do as traders arnt even considered by the majority of EA's."

Could you give some examples?

Its always hard to answer questions such as this, because everyone’s styles are different, so what I say may not apply to techniques used by others, however I’ll give it a try.

Probably the biggest differences I see between how a manual trader thinks, and an EA works, involves initial stop placement. Although I tend to use a statistical basis for stops, I also consider volatility, and support and resistance. So for example, let’s assume that statistics have shown that my optimum stop is 16 pips from entry, and I’m taking a short. Most of the time I'll stick to my 16 pip rule, because it works, however If there’s a strong area of resistance 25 pips above my entry, or the markets particularly choppy, I may well increase the stop, and reduce position size to take advantage of this. Most traders I know tend to employ any technique that would improve the probabilities in their favour and do this sort of thing all of the time. Whilst its technically feasible for an EA to do this, Ive not seen many that do.

I’ll also often adjust both stops and targets on the quality of the entry signal, particularly if I’m getting confirmation from other timeframes, I may be aiming for anywhere from a 1:1 to 5:1 reward to risk ratio, and I''ll change these dynamically on the fly as additional information comes to light.

The whole business is about identifying high probability set ups, and good risk reward opportunities. I’ve yet to see an EA that even considered the concept of risk reward, and by that I don’t mean blindly risking X% of equity, most are basically based on taking an entry, which might or might not be a decent entry, and then hoping things turn out OK.

Stop management is another area where EA’s are generally quite hopeless and generally don’t emulate the actions of a manual trader. Most of my strategies are based on set risk reward ratios; let’s assume that I’ll risk 20 to gain 40. However, occasionally I’ll be lucky enough to enter into a strong move, and by trailing stops sensibly, I could take 120 pips from a large intraday move. A few of those each month really do make a huge difference to performance. As a manual trader I can react to good fortune, change exit strategies on the fly as these sorts of opportunities present themselves. In fact for around 18 month in my own manual trading 80% of my trades resulted in mostly small gains offset by small losses with most of my profits earned entirely through those sorts of scenarios where I capitalised on nothing other than good luck, simply by understanding how to trail a stop.

Don’t get me wrong here, I wasn’t trading one of these crazy trend strategies with a low probability of winning, with lots of small losses and the occasional big win, I was trying to do the complete opposite, my entries where really quite good, but because I was a crap trader, with all of the usual faults and weaknesses I closed way too early, and had a tendency to let stops get taken out rather than bail out. However it demonstrates how even someone relatively inexperienced making a complete mess of the strategy I was supposed to be trading can capitalise on situations as they occur. So a manual trader will tend to trail stops based on price action as it unfolds.

Now look at how most EA’s trail stops, they impose crazy rules such as if the trades +30 in profit set the stop to break even, if +50 then set stop to +20, at +70 set stop to lock in 40 etc. It’s clunky and mechanical, and it takes no account whatsoever of the dynamics at play in the market and it’s not particularly effective. I’m not saying that the approach doesn’t work; I’m just saying someone with a couple of years experience will generally outperform an EA with regards to stop placement.

Personally I tend to avoid trading around news events; most EA’s have no consideration whatsoever of the effects of economic releases. Although I don’t take much notice of fundamentals I’m at least aware of when the news is coming and the likely effect on the market, an EA in the hands of a novice is going to place a 50 lot position with a 15 pip stop 2 minutes before NFP.

I spend a lot of my day looking for information, reading news feeds, getting info on bids and offers, constantly updating the map of the territory, assessing how price is behaving.

I’m running out of examples now but add to that a general awareness of support and resistance, and a familiarity with price action obtained through watching millions of one minute candles forming over the years, daily interaction with other traders etc etc etc and you start to get something an EA cant copy.

I consider myself to be quite disiplined, and Ive done a lot of work in defining strategy, but even so as a manual trader there are times when Im just completely useless, indisiplined, and not really in tune with whats being offered, and when this happens I cut back, and take a break, there's other times when Im hot and I know it, and I know I should be stepping on the gas. Im not talking about going crazy, or breaking fundemental money management rules, but I am talking about taking more marginal set up's, or being slightly more agressive. I suppose that an EA could do something similar by measuring its equity curve, to determine when its in phase.

I suppose what Im saying is unless someone's experienced these sorts of issue's its unlikely that they'd adressed them in the EA design, and perhaps they shouldnt as they are hard to impliment. Perhaps the secret is to keep the EA simple, but from experience Id say that although simple EA's can be profitable, they do tend to suffer drawdowns that are much graeter than experienced with manual trading.

Probably dosnt help much but the best I can come up with.

Thanks for the anwer, I think you are correct in that an EA fails to synthesize the many forms of information available to the trader. See, this is where I see a contradiction in the standard advise given to beginners, which is 'stick to your system no matter what' yet it would seem the best traders take a more holistic approach. So is the question really is, TA any good in isolation?

Thanks for posting this. It articulates something I've been struggling with for several months. I've stopped pursuing EAs and instead focus on developing meaningful indicators that alert me when certain objective criteria are met. Then I can add the human analysis to that and take action as appropriate.