Trend following not working on my 7,000 backtests !

This is a discussion on Trend following not working on my 7,000 backtests ! within the Trading Systems forums, part of the Methods category; Without making any backtests , just by looking at naked charts i can say trend following works amazingly fine , ...

Reply
 
LinkBack Thread Tools Search this Thread
Old Nov 1, 2017, 11:10am   #33
tar
 
tar's Avatar
Joined Nov 2006
Without making any backtests , just by looking at naked charts i can say trend following works amazingly fine , just dont over leverage .
tar is offline   Reply With Quote
Old Nov 1, 2017, 11:19am   #34
Joined Jan 2008
qwertyuiop1 started this thread Yes. Its interesting to get peoples feedback. Thank you for the feedback.

I guess the main point of my post was that if you read any book on trend-following, broadly speaking you get the impression that if you manage your risk, cut your losses, let profits run etc. that really it is a psychological challenge in handling drawdowns after that.
i.e. the system will certainly be profitable - but can you the trader actually trade it consistently?

And I guess my very comprehensive back test (if I may say so myself !) proves otherwise.
To be fair - the rules I used aren't bad and do adhere to all the basic principles of trend-following.

It used ATR for a SL placement (like the turtles), used moving average crossover instead of ATR trailing like the turtles did - but really...there's no reason why one should be better than the other over the long run.

Even this guy http://www.thetrendfollower.com/p/be...rformance.html
His system is very similar to mine and he has a positive expectancy of 0.32R - (albeit he enters on a breakout rather than mine which entered on a moving average crossover)(
He then pushes his stop-loss to break even after 24 hours if trade in profit or else closes (Please note as per my original post I also back-tested this Stop loss adjustment but it didn't help my results)

May I also add (and this bit is very disconcerting) - I originally manually back tested my system over a 8-10 uncorrelated instruments (mainly commods rather than FX) over 300 trades over the last year only.
This had a positive expectancy of 0.22R per trade.

This gave me full confidence on the system - it was only by pure chance I decided to follow up with outsourcing a more thorough back test like I did on FX pairs.
(I think most people would have reasonably ascertained that 300 trades is enough of a sample size. I'm sure many have brought a system live on a smaller sample sized back test)
But in this instance it wasn't !

Now of course you could say maybe commods have a 'personality' more suited to my system. But I think my system should be more robust than that.


The main point of my thread is this. Sticking to the basic simple ideologies of trend following is not enough - unlike what the books may make you think.

Even in Curtis Faith book he emphasises simplicity. But i'm questioning that. Although in saying that the turtles rules were very simple and had a lot in common with my own rules. SO why did their rules work and mine didn't?
Yes - they entered on a breakout which I am not doing. But they also emphasis that the entry is of little relevance anyway and trade management and sticking with the system is the key!
In fact - if I remember correctly Richard Dennis himself said in market wizards book he thinks anyone could come up with rules that are at least 80% as effective as his but the problem is they wouldn't stick to the rules.
I believe mine would be in that 80% - but still doesn't work. Have I just proved Richard Dennis wrong ??

And also - in Michael Covells book he gives a number of different systems that are effectively moving average crossover systems - and he states in the book that his comprehensive back tests provide a positive expectancy.

SO I'm just really confused really. I feel I've done nearly everything right - or at the very least I haven't done anything obviously wrong. And it doesn't work !
I do note a couple of posters suggested my entry was not good enough - but if you buy into the theory that the entry is least important, (which is broadly considered conventional wisdom - logic that I too would agree with) then I cant put too much weight on the notion of changing my entry will be the difference of turning the entire system into a positive expectancy.

Definitely back to the drawing board it seems !

Last edited by qwertyuiop1; Nov 1, 2017 at 11:29am.
qwertyuiop1 is offline   Reply With Quote
Old Nov 1, 2017, 11:20am   #35
Joined Jan 2008
qwertyuiop1 started this thread
Quote:
Originally Posted by tar View Post
Without making any backtests , just by looking at naked charts i can say trend following works amazingly fine , just dont over leverage .
Over what sample size is that out of curiosity and over what length of time are the results from?

Also - may I ask what is your expectancy per trade in terms of R? (Not sure if you're familiar with the R concept)
qwertyuiop1 is offline   Reply With Quote
Old Nov 1, 2017, 2:58pm   #36
 
dbphoenix's Avatar
Joined Aug 2003
Quote:
Originally Posted by qwertyuiop1 View Post

Even in Curtis Faith book he emphasises simplicity. But i'm questioning that. Although in saying that the turtles rules were very simple and had a lot in common with my own rules. SO why did their rules work and mine didn't?

Yes - they entered on a breakout which I am not doing. But they also emphasis that the entry is of little relevance anyway and trade management and sticking with the system is the key!

In fact - if I remember correctly Richard Dennis himself said in market wizards book he thinks anyone could come up with rules that are at least 80% as effective as his but the problem is they wouldn't stick to the rules.

I believe mine would be in that 80% - but still doesn't work. Have I just proved Richard Dennis wrong ??

And also - in Michael Covells book he gives a number of different systems that are effectively moving average crossover systems - and he states in the book that his comprehensive back tests provide a positive expectancy.

SO I'm just really confused really. I feel I've done nearly everything right - or at the very least I haven't done anything obviously wrong. And it doesn't work !
I do note a couple of posters suggested my entry was not good enough - but if you buy into the theory that the entry is least important, (which is broadly considered conventional wisdom - logic that I too would agree with) then I cant put too much weight on the notion of changing my entry will be the difference of turning the entire system into a positive expectancy.

Definitely back to the drawing board it seems !
What you've done is to "prove" that others' systems are ineffective, which is not the same thing as proving that trend-following is ineffective. You must first learn just what trend is and how it plays out. This is not difficult given that the market has only two states available to it: trending and ranging. If you don't understand the characteristics of each and how to distinguish between them, you're stuck at Go.

Can all these people be wrong? Yes.

As for trade entry being unimportant, the "conventional wisdom" can also be wrong. If after all you are immediately in a loss position, what does that do to you in terms of being able to manage the trade? Before taking someone's word for it that entry doesn't matter, I'd want to see a hell of a lot of verified trades that aren't anecdotal.

Giving up at this point is not wise. You're not required to learn how to trade both trends and ranges, but why not? They're not difficult. But first you have to forget about everything you "know" that isn't true.
__________________
Trading Price
dbphoenix is offline Training literature vendor (e.g courses / spreadbetting guides / eBooks)   Reply With Quote
Old Nov 4, 2017, 9:30am   #37
Joined Oct 2014
I read through and don’t think I’ve seen any mention of different exit strategies. In my tests I found the exit can be more important than the entry with massive difference in outcomes. If you can, test what closing at 2r vs 3r vs moving stop at entry at 1r or 2r then close when candle closes through a 9 ema .... just some ideas to play with. Also, how do you plan not to get killed in ranging markets? Tomorton above had a condition that could help you a little. Only trade if the EMAs are stacked and add a 200. Alternatively, only enter a trade in the direction of the trend in the higher time frame to a multiple of 4. Again, just ideas for you to mess about with, best of luck.

Edit: Not sure how you do your tests but could you set it so you dont enter longs within a set distance of previous day’s high and wait for a break first and viceversa for the shorts? I dont use crossovers but use EMAs and I find that significant levels such as pivots, daily highs/lows play a big part and dont think you allow for that. You might find forward testing with little money or even a Demo more useful as you could add filters that make sense.


Btw, english is my second language, if all I did was confuse, I apologise.

Last edited by carryontrading; Nov 4, 2017 at 9:45am.
carryontrading is offline   Reply With Quote
Thanks! The following members like this post: tomorton
Old Nov 4, 2017, 4:59pm   #38
Joined Jan 2008
qwertyuiop1 started this thread
Quote:
Originally Posted by carryontrading View Post
I read through and don’t think I’ve seen any mention of different exit strategies. In my tests I found the exit can be more important than the entry with massive difference in outcomes. If you can, test what closing at 2r vs 3r vs moving stop at entry at 1r or 2r then close when candle closes through a 9 ema .... just some ideas to play with. Also, how do you plan not to get killed in ranging markets? Tomorton above had a condition that could help you a little. Only trade if the EMAs are stacked and add a 200. Alternatively, only enter a trade in the direction of the trend in the higher time frame to a multiple of 4. Again, just ideas for you to mess about with, best of luck.

Edit: Not sure how you do your tests but could you set it so you dont enter longs within a set distance of previous day’s high and wait for a break first and viceversa for the shorts? I dont use crossovers but use EMAs and I find that significant levels such as pivots, daily highs/lows play a big part and dont think you allow for that. You might find forward testing with little money or even a Demo more useful as you could add filters that make sense.


Btw, english is my second language, if all I did was confuse, I apologise.
You've made some interesting points for me to think about. I guess I want to avoid tweaking parameters too much to avoid curve fitting. But if the tweaks make sense and if successful over a similar sample size then the tweaks may well be valid.

You may have missed it on my original post but I did try adjusting after 24 hours depending on whether trade in profit or not but it made no noticeable difference.

Out of curiosity do you trade a successful trend following strategy yourself?
If so what is your expectancy cy per tade do you have in terms of R? And over whstvsampke size?
qwertyuiop1 is offline   Reply With Quote
Old Nov 4, 2017, 5:42pm   #39
Joined Apr 2016
Quote:
Originally Posted by qwertyuiop1 View Post
to avoid curve fitting.
Why would a curve fitter avoid curve fitting ? This makes no sense. But I don't suppose it matters, the outcome is exactly same when real money is involved.

Same like when you try to take someone's money on the street and you do a curve fit before you try to take someone's money on the street, you will end up in exactly the same place: in a police cell or laying on the ground clutching your crotch from pain.

Making money isn't about curves, it's about why should someone willingly give your their money. Money in the market don't grow on trees that you can just pluck off at will.
EnlightenedJoe is offline   Reply With Quote
Old Nov 4, 2017, 6:43pm   #40
Joined Jan 2008
qwertyuiop1 started this thread
Quote:
Originally Posted by EnlightenedJoe View Post
Why would a curve fitter avoid curve fitting ? This makes no sense. But I don't suppose it matters, the outcome is exactly same when real money is involved.

Same like when you try to take someone's money on the street and you do a curve fit before you try to take someone's money on the street, you will end up in exactly the same place: in a police cell or laying on the ground clutching your crotch from pain.

Making money isn't about curves, it's about why should someone willingly give your their money. Money in the market don't grow on trees that you can just pluck off at will.

But I'm not a curve fitter. Nor should anyone be a curve fitter.
All successful systems need to be robust.
qwertyuiop1 is offline   Reply With Quote
Reply

Thread Tools Search this Thread
Search this Thread:

Advanced Search

Similar Threads
Thread Thread Starter Forum Replies Last Post
Differences between backtests done by several software kojinakata Trading Systems 8 Mar 29, 2015 2:51pm
$1 to $1,000,000,000 Tribute to S1quash Hotch Trading Journals 7 Aug 6, 2011 7:22pm
Metatrader backtests flawed oildaytrader Trading Software 19 Nov 4, 2009 9:57pm
Long chart history for backtests.. dbinvest Forex 0 Dec 20, 2007 1:01pm

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)