Trend following not working on my 7,000 backtests !

qwertyuiop1

Member
Messages
81
Likes
3
I have lost all confidence on trend following.

I back-tested the below system over 9 major FX pairs using 4 hour charts over 13 years which consisted of exactly 7,004 trades.

Plot 3 EMAS on chart: 5,21,55
Use a Stop loss od 2ATR(14)

Enter long/short when 5 EMA crosses above/below the 55 EMA
Exit when 5EMA crosses back over ANY of the 21 or 55 EMA (whichever comes first)

And that's it !

As you can see it adheres to all the fundamental rules of a good trading system:
- IT knows what instruments it trades
- Knows entry signal
- Knows exit signal before entering which facilitates both cutting losses and maximising profits
- cuts losses
- lets profits run
- knows risk in advance

It barely broke even before costs (It had an expectancy of 0.01R per trade). Since costs were about 0.02 R it had an expectancy of -0.01R per trade.

I then tweaked it slightly whereby I closed any trades that were not in profit after 24 hours. And for those that were in profit I moved the SL to breakeven.
Again- the overall results were very similar.

I'm not saying trend-following cant assist in trading. But based on my back tests - (which was a significant sample size in anyone's language) it definitely isn't a simple case of cut your losses and let your profits run and you are guaranteed profits in the long run like many sites lead you to believe.

It really needs to be fine tuned a bit better than that. I really don't know where to go from here to be honest.

Ya - you can tweak the parameters - but that should now make any difference in the long run.

Id be curious to get other peoples thoughts on it.
 
Last edited:
My only suggestion which is off the back of fundamental analysis is to look for catalysts as a trade indicator instead of indicators. All trends are driven by fundamentals and there is always one or more catalysts at the start (a news event) that starts traders piling in. Taking the 4h chart as your base, derive your signals off candles that exhibit a large enough move that would signify something has happened. This would filter out the stuff that eats away at your account.

Brad
 
Sounds like you're testing oscillations, not trend. Try drawing a "trendline" underneath your swing lows. At the very least, this enables you to take back control of your trading instead of leaving it up to indicators.
 
Sounds like you're testing oscillations, not trend. Try drawing a "trendline" underneath your swing lows. At the very least, this enables you to take back control of your trading instead of leaving it up to indicators.

Oscillations? Not sure what you mean?

And you mean drawing trend lines beneath swing lows for Stop-loss placement?

I appreciate the suggestion but I fail to see how logically that could be better than using 2ATR.

The fact is that in my backtests that most of the time I was stopped out from the EMAs crossing over anyway.
And a 2ATR placement still lets me assume large positions sizes in quiet times and vice verse. Effectively it should be no better or worse than using swing lows as stop -placement over a large sample size
 
How did you arrive at the values of the EMA's and ATR based stop loss, are they optimised for the data? (ie. curve fitted). Have you tried any other trend following strategies other than MA crosses? I say this because IMO they are far too simplistic. Please don't rule out trend following just because one very basic strategy doesn't work...
 
My only suggestion which is off the back of fundamental analysis is to look for catalysts as a trade indicator instead of indicators. All trends are driven by fundamentals and there is always one or more catalysts at the start (a news event) that starts traders piling in. Taking the 4h chart as your base, derive your signals off candles that exhibit a large enough move that would signify something has happened. This would filter out the stuff that eats away at your account.

Brad

Hmm.. again I genuinely do appreciate the suggestion.

But trying to identify catalysts that aren't false catalysts for movements on major FX pairs is nigh on impossible for the lay man on the street.
And you'll still need a mechanism to cut losses. Which means your back to using indicators or price action.

I'm suddenly very sceptical on trend following as a concept. Im not going to be so arrogant and state that people aren't making money from trend following principles.

But if you read all the books the overwhelming message is that once you stick to the basic principles of trend-following (cut losses and let profits run), then its really just a psychological challenge after that.

That is seriously under question in my mind right now.

The fact is my system obeys all the principles. To turn it into a positive expectancy system we would be getting into the world of very specific tweaks a.k.a. curve fitting.
My understanding is that any successful trend following system should be robust. Making specific tweaks geos against this principle.
 
Oscillations? Not sure what you mean?

And you mean drawing trend lines beneath swing lows for Stop-loss placement?

I appreciate the suggestion but I fail to see how logically that could be better than using 2ATR.

The fact is that in my backtests that most of the time I was stopped out from the EMAs crossing over anyway.
And a 2ATR placement still lets me assume large positions sizes in quiet times and vice verse. Effectively it should be no better or worse than using swing lows as stop -placement over a large sample size

If you're not wedded to indicators, this may be of interest. If you are wedded to indicators, then it won't be of any interest at all.
 
How did you arrive at the values of the EMA's and ATR based stop loss, are they optimised for the data? (ie. curve fitted). Have you tried any other trend following strategies other than MA crosses? I say this because IMO they are far too simplistic. Please don't rule out trend following just because one very basic strategy doesn't work...

The 2 ATR stop loss was my own. But its as good as any. Its a popular stop loss placement used by many people. It gives the price plenty of room to move. And as I stated in another post, the overwhelming majority of trades were exited due to crossing of EMAS. So its not like the SL is way too close that it impacts the expectancy of the system.


The EMAs used were taken from some other guys course that I bought for £100. In saying that I could have used any EMAs. It really shouldn't matter over a sample size of 7,000 trades.

I haven't backtested as comprehensively any other system. But in saying that - no matter what variances you put on the system, all trend-following strategies are largely similar in the basic principles.

IS there anyone out there that uses a trend-following system successfully?
If so what is your expectancy per trade in terms of R? And over what ample size?
 
Hmm.. again I genuinely do appreciate the suggestion.

But trying to identify catalysts that aren't false catalysts for movements on major FX pairs is nigh on impossible for the lay man on the street.
And you'll still need a mechanism to cut losses. Which means your back to using indicators or price action.

I'm suddenly very sceptical on trend following as a concept. Im not going to be so arrogant and state that people aren't making money from trend following principles.

But if you read all the books the overwhelming message is that once you stick to the basic principles of trend-following (cut losses and let profits run), then its really just a psychological challenge after that.

That is seriously under question in my mind right now.

The fact is my system obeys all the principles. To turn it into a positive expectancy system we would be getting into the world of very specific tweaks a.k.a. curve fitting.
My understanding is that any successful trend following system should be robust. Making specific tweaks geos against this principle.

Its simple, create a script that identifies a 4 hour candle that is on or larger than the average daily range. I highly doubt you will find many instances of these that turned out to be false positives. The only other suggestion I have is to drop automated technical trading forex and start using fundamentals instead.
 
The 2 ATR stop loss was my own. But its as good as any. Its a popular stop loss placement used by many people. It gives the price plenty of room to move. And as I stated in another post, the overwhelming majority of trades were exited due to crossing of EMAS. So its not like the SL is way too close that it impacts the expectancy of the system.


The EMAs used were taken from some other guys course that I bought for £100. In saying that I could have used any EMAs. It really shouldn't matter over a sample size of 7,000 trades.

I haven't backtested as comprehensively any other system. But in saying that - no matter what variances you put on the system, all trend-following strategies are largely similar in the basic principles.

IS there anyone out there that uses a trend-following system successfully?
If so what is your expectancy per trade in terms of R? And over what ample size?

Frankly I don't find moving averages cross overs to be a good indicators for stocks (I have no FX experience), maybe only as long term trend indicators for indexes...
I do admire your back testing, this is the way, so just go ahead and test another system and so forth until you find what works for you...
 
Its simple, create a script that identifies a 4 hour candle that is on or larger than the average daily range. I highly doubt you will find many instances of these that turned out to be false positives. The only other suggestion I have is to drop automated technical trading forex and start using fundamentals instead.

You mean when the 4 hour candle is larger than the daily ATR use this as an entry?

This could certainly be used a ns tested as a possible entry signal. No doubt about that.
But it all swings in roundabouts. IN waiting for that you could miss out on the beginning of a move whereas my system may have gotten you in to it earlier.

And if you are of the belief (which most seem to be) that the entry really is the least significant part of a trade, and that its all in the trade management then it shouldn't really make much of a difference what you use as your entry.

IN saying that - i'm not saying you're wring. Maybe with proper testing it could be a very successful system and it could have a very large percentage of winners.

But for me I would intuitively go along with the belief that its all about trade management rather than trying to perfect entry signals.
 
Frankly I don't find moving averages cross overs to be a good indicators for stocks (I have no FX experience), maybe only as long term trend indicators for indexes...
I do admire your back testing, this is the way, so just go ahead and test another system and so forth until you find what works for you...

Well...I suppose that is my concern. I really like the concept of mechanical trend following - and I just assumed once I adhered to the basic principles then it would naturally work....but alas that is not the case.

And at the end of the day, all trend following systems are broadly similar. There's only so many different way someone can cut losses and let profits run.
If you are using ATR or MAs or something else, over the long run its all much of a muchness.

Hence this thread. I am now questioning the entire concept of trend following.

Do my results mean I should abandon trend following completely? Id love if someone could persuade me not to - but my backtest results only point me in one direction./
 
You mean when the 4 hour candle is larger than the daily ATR use this as an entry?

This could certainly be used a ns tested as a possible entry signal. No doubt about that.
But it all swings in roundabouts. IN waiting for that you could miss out on the beginning of a move whereas my system may have gotten you in to it earlier.

And if you are of the belief (which most seem to be) that the entry really is the least significant part of a trade, and that its all in the trade management then it shouldn't really make much of a difference what you use as your entry.

IN saying that - i'm not saying you're wring. Maybe with proper testing it could be a very successful system and it could have a very large percentage of winners.

But for me I would intuitively go along with the belief that its all about trade management rather than trying to perfect entry signals.


if you want to get in sooner then you have to incorporate fundamentals. Here are some examples of the sort of bars i mean on EURUSD
 

Attachments

  • Capture.PNG
    Capture.PNG
    66.4 KB · Views: 574
Well...I suppose that is my concern. I really like the concept of mechanical trend following - and I just assumed once I adhered to the basic principles then it would naturally work....but alas that is not the case.

And at the end of the day, all trend following systems are broadly similar. There's only so many different way someone can cut losses and let profits run.
If you are using ATR or MAs or something else, over the long run its all much of a muchness.

Hence this thread. I am now questioning the entire concept of trend following.

Do my results mean I should abandon trend following completely? Id love if someone could persuade me not to - but my backtest results only point me in one direction./

Well, I do use a completely mechanical system for stocks (essentially I have a robot to do the trading), which works pretty good for me and it is sort of a trend following system, but I do not use ATR or MAs, so my advice is test everything that comes to your mind and sooner than later you'll find your system...
 
if you want to get in sooner then you have to incorporate fundamentals. Here are some examples of the sort of bars i mean on EURUSD

Yes - I do see those bars. But like any system they only provide the entry.
You still need to know
- where to place stop -loss.
- how to cut losses short
- How to take profit.


Basically you need to come up with a full trading plan - put the whole lot together and then backtest,. And until you've done that we really cant say one way or the other whether its any better or worse than my system.

Essentially you've only given an alternate entry, And as stated previously, most experts will agree that this is in fact the least important part to any any system.

Nut in saying that - I'm not suggesting your system shouldn't be developed and backtested. It may well prove successful. But for now we cannot say.
 
Well, I do use a completely mechanical system for stocks (essentially I have a robot to do the trading), which works pretty good for me and it is sort of a trend following system, but I do not use ATR or MAs, so my advice is test everything that comes to your mind and sooner than later you'll find your system...

May I ask (if not being too cheeky that is), given yours is a fully mechanical system, what are the rules you use?
And what is the expectancy in terms of R?
 
May I ask (if not being too cheeky that is), given yours is a fully mechanical system, what are the rules you use?
And what is the expectancy in terms of R?

Unfortunately I have worked very hard on my system and I am not about to publish my rules unless I am payed 8 figures :) But I can tell you, I found it by back testing any system I can think of or find for free (or small amount of money for books for example) and based on those failed back test and observations I came up with mine... Basically you are at the begging of your journey, you did test one, it didn't work, so you just have to go and do it again for another...
 
Unfortunately I have worked very hard on my system and I am not about to publish my rules unless I am payed 8 figures :) But I can tell you, I found it by back testing any system I can think of or find for free (or small amount of money for books for example) and based on those failed back test and observations I came up with mine... Basically you are at the begging of your journey, you did test one, it didn't work, so you just have to go and do it again for another...

In all your backtesting would you share ANY other system (that you don't use, and won't ever use) that you found to have even a modest positive expectancy? :idea:

Please :D
 
I have lost all confidence on trend following.

You are doing it wrong. Back testing is 100% guaranteed to make demo pips. Just try some different parameters until you get the magic numbers.

But personally, I don't see much point. Demo pips can't be used for buying lunch. With all that back testing, you will surely get hungry and need lunch. If lunch doesn't pay for itself, then what's the point of all that effort ?

Magic numbers don't work in the real market. Although I am willing to change my mind on evidence.
 
In all your backtesting would you share ANY other system (that you don't use, and won't ever use) that you found to have even a modest positive expectancy? :idea:

Please :D

Frankly I plan to use most of my systems in one way or another and I am always looking for more...

The main reason is they are either too close to one another or completely different (which is an edge on itself to have uncorrelated market strategies)

Also no offense guys and galls, but we are competing with one another on a daily basis, so I cannot tell you even the systems that do not work, just because this will save you the time and effort and by the process of elimination you might discover some of my systems (great minds think alike) and the increase money flow will disrupt my earnings...

Sorry about that... if you want to discus wild carp river fishing tactics, I have nothing to hide there since we are all competing on a different venues, so I can open another topic about it...
 
Top