Diversified Trend Trading Approach | Page 33 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Is Grid Trading (combined trend following and counter trend) Profitable? 3 replies

Making a good diversified portfolio 10 replies

Diversified EA Money Management Code 10 replies

HAMA PAD - A Simple Trading Approach 7 replies

Trading a (semi) automatic approach...but want more 11 replies

Attachments: Diversified Trend Trading Approach
Exit Attachments

Diversified Trend Trading Approach

  • #641
  • Feb 12, 2016 10:18pm Feb 12, 2016 10:18pm
Quoting Atokys
{quote} {quote} lol, hope your testing goes well. My gripe with filters is the implicit suggestion that they add value with no drawback. People tend to think of filters in the context of "filtering out bad trades". After having spent much time wrestling with these ideas I came to the conclusion that simplicity prevails for two reasons: 1. More filters increase the likelihood that a good historical performance is the result of curve-fitting. 2. The inescapable fact of trend trading is that profits come from being exposed to market...
I think it depends on what the actual 'filter' being utilised is.

I previously developed a trend following strategy for trading equities. By the introduction of one 'key' filter the bad trades did in fact reduce, as well as overall drawdown to the system. This was then back tested over 10 years of clean historical data ensuring that each block of new data backtested was out of sample. I even ran monte carlo analysis on the backtested data and found the overall profit & max DD distributions to be extremely tight.

The lesson learn (for me anyway) is that they shouldn't be discounted but there does need to be real logic behind why you think they might work.

Just my 2cents.

CCH
  • #642
  • Feb 12, 2016 10:34pm Feb 12, 2016 10:34pm
Quoting Chorlton
{quote} I think it depends on what the actual 'filter' being utilised is. I previously developed a trend following strategy for trading equities. By the introduction of one 'key' filter the bad trades did in fact reduce, as well as overall drawdown to the system. This was then back tested over 10 years of clean historical data ensuring that each block of new data backtested was out of sample. I even ran monte carlo analysis on the backtested data and found the overall profit & max DD distributions to be extremely tight. The lesson learn (for me...
Interesting Chorlton. Thanks for sharing.

I am thinking that a possible use for filters is as a method to reduce trading in market noise. The Donchian systems I am testing now obviously side-step this issue as entries and exits are positioned at distal positions to normal market volatility hence probably do not require any filter additions. The system already has simplicity built into it's entries and exits.

If however I was trading a MA crossover system, I would be looking for a device to avoid trading within normal volatility and would find it difficult to not at least consider utilising some sort of filter to push the entries and exits out of the 'noise'. You can get into some very ugly drawdowns with MA crossover systems unless you engage some form of filter to supplement entry/exit conditions such as an ATR filter or other alternate method such as Perry Kaufmann's efficiency ratio which is an effective signal to noise ratio measure.
  • #643
  • Edited 11:53pm Feb 12, 2016 11:15pm | Edited 11:53pm
Donchian Suite Tests

Refer to system definition in Post 640

Strategy: D1T
Type: Short Term Trend Trader
Timeframe: D1
Test Period: 1/1/2000 - 31/12/2015
Donchian Entry: 20 Period
Donchian Exit: 10 Period
Initial Stop: 10 Period ATR x 2
Note that all unrealised profit taken at the end of the test hence the significant profit taken on final bar of test.

Note that when comparing against other methods drawdowns usually are experienced upon portfolio commencement as it takes time for the portfolio to get up and running and achieve diversification benefits.

Un-scaled Result - No profits reinvested:
Attached Image (click to enlarge)
Click to Enlarge Name: D1T Unscaled.JPG Size: 167 KB


Notes: Of the 40 instruments in the portfolio, 25 were profitable with the balance being unprofitable. High impact drawdown periods were between 2000 and 2002 and between 2010 and 2011. A total of 2047 trades taken with a 30% win rate and positive expectancy of 0.19x(R). R = $500.

Unscaled annual results are as follows.
Attached Image


Scaled result with profits reinvested: To achieve a max drawdown over the period of 30%
Attached Image (click to enlarge)
Click to Enlarge Name: D1T Scaled.JPG Size: 69 KB


Now onto D1S. Stay tuned.

C
  • #644
  • Edited Feb 13, 2016 12:20am Feb 12, 2016 11:58pm | Edited Feb 13, 2016 12:20am
Quoting Copernicus
{quote} Interesting Chorlton. Thanks for sharing. I am thinking that a possible use for filters is as a method to reduce trading in market noise. The Donchian systems I am testing now obviously side-step this issue as entries and exits are positioned at distal positions to normal market volatility hence probably do not require any filter additions. The system already has simplicity built into it's entries and exits. If however I was trading a MA crossover system, I would be looking for a device to avoid trading within normal volatility and...

Agreed but how does one define market noise in advance? That's one of the many questions that I haven't found a satisfactory answer to yet.

As a side note, based on my own research and understanding of the equities market I am a strong believer in following trends as I believe they do offer an edge. In addition, the logic in keeping losses to a minimum and taking maximum advantage of those outlier trades (which provide the massive R:R) is something that I have always coded into my systems.

However, since making the transition from equities to developing systems for trading FX, the underlying issue for me at the moment is that I do not yet fully understand what really drives price on a FX chart from one moment to the next. Although I understand some of the more basic theory behind it, I am not at a stage yet where I have been able to formulate a robust trading plan in order to consistently trade. For example, on an equities chart, I can be reasonably confident whether a breakout is really the start of a real new trend or whether there is a good chance that it will fail. Unfortunately if I apply the same kind of analysis to an FX chart I don't have no where near as much success.

My personal opinion thus far (caveat: As I am constantly learning my opinions may change over time) is that price movement in FX is largely random. By this I mean that although on a granular level there are clearly obvious reasons why each individual buy and sell order is being placed at any one time, due to the shear volume of trades and differing opinions in the market at that same time it creates a situation that in essence looks like random movement. However, there are also those specific times where price has a clear direction and is no longer random. Unfortunately, I am yet to consistently identify and capitalize on these specific moments. One area which I initially investigated in order to define this was to look at volatility. But to date, my findings has varied greatly from looking at FX charts vs equity charts. This may be due to the fact that the volume which I analyse comes from my broker via MT4 so is probably not very accurate. Alternatively, there are probably other factors at play which I am yet to understand.

As I can definitely relate to your trading approach, I will continue to follow this thread with interest.
  • #645
  • Feb 12, 2016 11:59pm Feb 12, 2016 11:59pm
Quoting Copernicus
Donchian Suite Tests Refer to system definition in Post 640 Strategy: D1T Type: Short Term Trend Trader Timeframe: D1 Test Period: 1/1/2000 - 31/12/2015 Donchian Entry: 20 Period Donchian Exit: 10 Period Initial Stop: 10 Period ATR x 2 Note that all unrealised profit taken at the end of the test hence the significant profit taken on final bar of test. Note that when comparing against other methods drawdowns usually are experienced upon portfolio commencement as it takes...
How do you do your testing? Really interested...
  • #646
  • Edited 7:09am Feb 13, 2016 12:38am | Edited 7:09am
Quoting Chorlton
{quote} Agreed but how does one define market noise in advance? That's one of the many questions that I haven't found a satisfactory answer to yet. As a side note, based on my own research and understanding of the equities market I am a strong believer in following trends as I believe they do offer an edge. In addition, the logic in keeping losses to a minimum and taking maximum advantage of those outlier trades (which provide the massive R:R) is something that I have always coded into my systems. However, since making the transition from equities...
Hi Chorlton. Great to have you on board.

Obviously in determining the why's for a trend is a fairly subjective exercise. My own trading philosophy therefore plays a big part in the methods I use. As such, the following rationale pretty well sums up my approach.

1. I believe that in general markets are very efficient leading to significant random price action attributed to the different motivations of the huge number of different participants. Obviously the very mature markets like the indices, major currencies etc. are dominated by this feature. Emerging markets however are not quite as efficient and as a result, you see some new currency pairs, indices etc. displaying significant linear trending tendencies.
2. Within this random price action however lies price action bias which only shows up over time in the way random patterns such as random trends, etc. are directionally extended. Visually it is exceedingly difficult to filter what is effectively a random trend versus one with 'real' non-random significance. What IMO causes this directional bias in otherwise random price movements is a phenomenon referred to as 'autocorrelation'. This is where we get a feature referred to as 'levy drift' rearing it's head in otherwise random data. Driving this autocorrelation in my opinion is overall fundamental factors that are prevalent drivers of the major secular trends (you see on the MN and W1 timeframes). The repetitive actions of news events over time on the longer term timeframes impose a serial repetition in this bias which translates to the lower timeframes. It is difficult visually interpreting this on a price chart so the best tool I have in achieving this is applying a linear regression channel over a price series to determine which way the non-random bias is playing out. The following article is a great way to view things in my opinion. The effect of a small bias in an otherwise random price series is pretty powerful.
3. With all this in mind, I therefore see price action comprising a significant random behaviour (normal volatility) with a small kernel of non-random price behaviour that is extending this price action in a particular way. In trading trends therefore I aim to determine the volatility of price action and set my entries and exits at the distal edges of this volatility to avoid the whipsaws associated with normal volatility. If you have a similar opinion then Perry Kaufmann's signal to noise ratio pretty well sums up a method to employ to differentiate between a valid non-random signal versus noise. You will find details of this on pp10 of his 5th edition book 'Trading Systems and Methods". Effectively he measures an efficiency ratio which is the net change in price from say A to B (as an absolute value) and divides this by the sum of individual price changes of a series (as absolute values). The result seeks to remove the volatility associated with a price move from A to B and isolates that which doesn't contribute to that which does....if that makes sense.

....anyway, looking forward to your input mate.

Regards

C
  • #647
  • Feb 13, 2016 12:44am Feb 13, 2016 12:44am
Quoting Chorlton
{quote} How do you do your testing? Really interested...
I have only recently moved off the MT4 platform into a Microsoft Excel environment to get a bit more serious with the testing. There are some great portfolio testing software solutions out there but I find there is nothing better than developing the system yourself as it forces you to become very familiar with all elements of the system you are trading. Only the other day Bass on this thread pointed me towards Trading Blox software which is a very powerful solution which is probably going to be a fail-safe for me if my progress in Excel starts to slow me down.....but so far so good. If you know Excel and get familiar with DLL's and linking to 3rd party API's it can be a really powerful solution.

Cheers mate.

C
  • #648
  • Feb 13, 2016 1:16am Feb 13, 2016 1:16am
Hi Copernicus,

Firstly I don't want to interrupt the flow of your thread in anyway so if my questions are doing this please just let me know and I'll refrain from asking too many of them

Assuming you are OK, then I do have a question regarding your point below. Apologies if this has already been covered in past threads. I am only now going through these.

Quoting Copernicus
{quote} In trading trends therefore I aim to determine the volatility of price action and set my entries and exits at the distal edges of this volatility to avoid the whipsaws associated with normal volatility.
If I understand correctly you are waiting for sufficient price movement in one direction in order to satisfy yourself that price is no longer trading within the normal volatility range before taking a new trade. If this is true, have you done any previous testing in order to determine on average how much further price would continue to move in the same direction after you place your entry trade?

Also, do you set your trades to break-even at some point?

PS. Thanks for the link to that article... very interesting....
  • #649
  • Feb 13, 2016 1:21am Feb 13, 2016 1:21am
Quoting Copernicus
{quote} I have only recently moved off the MT4 platform into a Microsoft Excel environment to get a bit more serious with the testing. There are some great portfolio testing software solutions out there but I find there is nothing better than developing the system yourself as it forces you to become very familiar with all elements of the system you are trading. Only the other day Bass on this thread pointed me towards Trading Blox software which is a very powerful solution...
Thanks for the response.

I am 'OK' with Excel but have never been involved with DLLs or APIs so not sure how I'd go with it. I am still trying to get my head around MQL4 !! Having come from Amibroker and Tradesim which between them had great backtesting functionality, MT4's strategy tester is nothing short of a joke!! Hence my question.
  • #650
  • Edited 7:02am Feb 13, 2016 1:36am | Edited 7:02am
Quoting Chorlton
Hi Copernicus, Firstly I don't want to interrupt the flow of your thread in anyway so if my questions are doing this please just let me know and I'll refrain from asking too many questions Assuming you are OK, then I do have a question regarding your point below. Apologies if this has already been covered in past threads. I am only now going through these. {quote} If I understand correctly you are waiting for sufficient price movement in one direction in order to satisfy yourself that price is no longer trading within the normal volatility range...
No probs C with the questions. :-)

Depending on the method I use, then I have a few tools that help me in placing entries and exits. In my discretionary methods DTT, EDTT and EDTT1 that are described in this thread I use standard deviation channels (SDC's) which extend to previous major swing highs/lows to determine the trend direction, momentum and volatility range. What this does is define recent historic price action using objective measures as opposed to subjective line draws. The volatility of price action falls within this channel to give me an idea of the substance of any directional bias that may be lying there. I then place predictive entry lines in clear blue sky that forces the price action to extend to levels where I am confident that true directionality exists and then I hop on board using a breakout from these predictive entry lines. I use multi-timeframes on entry so I scale into trends from M30, then D1 then H4 then D1 etc. all the way up the timeframes.

I define the end of trend when price touches the outer standard deviation channel and this linear line is also used as a trailing stop on the trend. That way, it forces a result in the direction of prevailing trend immediately on entry so you will rarely ever cop a full loss. Using the SDC as a trailing stop, it naturally moves you to breakeven when volatility leads it there. I only enter a new timeframe when at least breakeven is achieved on the lower timeframe to avoid excessive risk to the portfolio. It is a method that is similar to pyramiding but is scaled to each timeframe accordingly.

With the Donchian automated system I am trialing now, the exit is determined by the donchian high/low so in effect ratchets you towards breakeven in accordance to price moving in-trend. If you are wrong on entry, the most you get hit with is the initial trade risk defined by the ATR stop. The ATR initial stop ensures position sizing is adjusted to recent volatility upon entry so when things get boisterous, position sizes reduce correspondingly. Using the Donchian as a breakout ensures you are outside normal volatility as to achieve this level you must be reaching a new high/low over an extended price series.

In essence all my methods ensure that prices have to stretch significantly directionally before I get triggered. Trading a diverse portfolio allows me to make things tough for price and have to work hard to get there. That way I keep the powder dry and only get interested on the more major price movements. Apples being apples, while I never know if the trend is a real one or simply a random trend, by making price have to work hard for entry I am naturally biasing my trades to more durable trends....if that makes sense.
  • #651
  • Edited 2:59am Feb 13, 2016 1:44am | Edited 2:59am
Quoting Chorlton
{quote}I am still trying to get my head around MQL4 !! Having come from Amibroker and Tradesim which between them had great backtesting functionality, MT4's strategy tester is nothing short of a joke!! Hence my question.
hehehehe. Mt4 is a real pain in the butt for backtesting and is very limited in that capability. I would stick with Amibroker if I were you and possibly consider using Interactive Brokers if you want to test comprehensively in this forex universe. One of the main reasons I threw up my hands with MT4 in testing is that to test simple systems across a diversified portfolio, would take me a week when now it takes me a few hours or so with the Excel solution. As I am writing this post Excel is coming to the end of a test that has been running across a portfolio of 40 instruments between 2000 to 2015 for about an hour. It is making my life far simpler. :-)
  • #652
  • Feb 13, 2016 2:07am Feb 13, 2016 2:07am
Donchian Suite Tests continued....

Refer to system definition in Post 640

Strategy: D1S
Type: Short Term Swing Trader
Timeframe: D1
Test Period: 1/1/2000 - 31/12/2015
Donchian Entry: 20 Period
Donchian Exit: 5 Period
Initial Stop: 10 Period ATR x 2
Note that all unrealised profit taken at the end of the test hence the significant profit taken on final bar of test.

Note that when comparing against other methods drawdowns usually are experienced upon portfolio commencement as it takes time for the portfolio to get up and running and achieve diversification benefits.

Un-scaled Result - No profits reinvested:
Attached Image (click to enlarge)
Click to Enlarge Name: D1S Unscaled.JPG Size: 168 KB



Notes: Of the 40 instruments in the portfolio, 22 were profitable with the balance being unprofitable. High impact drawdown periods were between 2001-2002, 2010-2011, 2012-2013 and 2014-2015. A total of 2633 trades taken with a 30% win rate and positive expectancy of 0.11x(R). R = $500.

Unscaled annual results are as follows.
Attached Image



Scaled result with profits reinvested: To achieve a max drawdown over the period of 30%
Attached Image (click to enlarge)
Click to Enlarge Name: D1S Scaled.JPG Size: 70 KB



Two down now only 10 more to go.......*gulp* :-)

C
  • #653
  • Feb 13, 2016 3:50am Feb 13, 2016 3:50am
A musical interlude to pass the time....in funk. :-)

Inserted Video
  • #654
  • Feb 13, 2016 4:22am Feb 13, 2016 4:22am
Donchian Suite Tests continued....

Refer to system definition in Post 640

Strategy: D2T
Type: Short Term Trend Trader
Timeframe: D1
Test Period: 1/1/2000 - 31/12/2015
Donchian Entry: 50 Period
Donchian Exit: 25 Period
Initial Stop: 25 Period ATR x 2
Note that all unrealised profit taken at the end of the test hence the significant profit taken on final bar of test.

Note that when comparing against other methods, drawdowns usually are experienced upon portfolio commencement as it takes time for the portfolio to get up and running and achieve diversification benefits.

Un-scaled Result - No profits reinvested:
Attached Image (click to enlarge)
Click to Enlarge Name: D2T Unscaled.JPG Size: 242 KB



Notes: Of the 40 instruments in the portfolio, 26 were profitable with the balance being unprofitable. High impact drawdown periods were between 2001-2007. A total of 1375 trades taken with a 23% win rate and positive expectancy of 0.22x(R). R = $500.

Unscaled annual results are as follows.
Attached Image




Scaled result with profits reinvested: To achieve a max drawdown over the period of 30%
Attached Image (click to enlarge)
Click to Enlarge Name: D2T Scaled.JPG Size: 69 KB



More to come :-)

C
  • #655
  • Feb 13, 2016 5:34am Feb 13, 2016 5:34am
Donchian Suite Tests continued....

Refer to system definition in Post 640

Strategy: D2S
Type: Short Term Swing Trader
Timeframe: D1
Test Period: 1/1/2000 - 31/12/2015
Donchian Entry: 50 Period
Donchian Exit: 10 Period
Initial Stop: 25 Period ATR x 2
Note that all unrealised profit taken at the end of the test hence the significant profit taken on final bar of test.

Note that when comparing against other methods, drawdowns usually are experienced upon portfolio commencement as it takes time for the portfolio to get up and running and achieve diversification benefits.

Un-scaled Result - No profits reinvested:
Attached Image (click to enlarge)
Click to Enlarge Name: D2S Unscaled.JPG Size: 242 KB




Notes: Of the 40 instruments in the portfolio, 25 were profitable with the balance being unprofitable. High impact drawdown periods were between 2001-2007. A total of 1769 trades taken with a 27% win rate and positive expectancy of 0.16 x R. Where R = $500.

Unscaled annual results are as follows.
Attached Image



Scaled result with profits reinvested: To achieve a max drawdown over the period of 30%
Attached Image (click to enlarge)
Click to Enlarge Name: D2S Scaled.JPG Size: 71 KB



That pretty much wraps up the short term timeframes. Now onto the medium term timeframes. More to come :-)

C
  • #656
  • Feb 13, 2016 6:51am Feb 13, 2016 6:51am
Donchian Suite Tests continued....

Refer to system definition in Post 640

Strategy: D3T
Type: Medium Term Trend Trader
Timeframe: D1
Test Period: 1/1/2000 - 31/12/2015
Donchian Entry: 100 Period
Donchian Exit: 50 Period
Initial Stop: 50 Period ATR x 3
Note that all unrealised profit taken at the end of the test hence the significant profit taken on final bar of test.

Note that when comparing against other methods, drawdowns usually are experienced upon portfolio commencement as it takes time for the portfolio to get up and running and achieve diversification benefits.

Un-scaled Result - No profits reinvested:
Attached Image (click to enlarge)
Click to Enlarge Name: D3T Unscaled.JPG Size: 241 KB



Notes: Of the 40 instruments in the portfolio, 26 were profitable with the balance being unprofitable. High impact drawdown periods were between 2002-2007. A total of 702 trades taken with a 27% win rate and positive expectancy of 0.42 x R. Where R = $500.

I've hit a sweet spot here. The medium term trend traders show quite some promise and will be valuable contributors when we finally blend the portfolio.

Time to party!!!!!

Inserted Video


Unscaled annual results are as follows.
Attached Image



Scaled result with profits reinvested: To achieve a max drawdown over the period of 30%
Attached Image (click to enlarge)
Click to Enlarge Name: D3T Scaled.JPG Size: 69 KB



More to come :-)

C
  • #657
  • Feb 13, 2016 8:02am Feb 13, 2016 8:02am
Donchian Suite Tests continued....

Refer to system definition in Post 640

Strategy: D3S
Type: Medium Term Swing Trader
Timeframe: D1
Test Period: 1/1/2000 - 31/12/2015
Donchian Entry: 100 Period
Donchian Exit: 20 Period
Initial Stop: 50 Period ATR x 3
Note that all unrealised profit taken at the end of the test hence the significant profit taken on final bar of test.

Note that when comparing against other methods, drawdowns usually are experienced upon portfolio commencement as it takes time for the portfolio to get up and running and achieve diversification benefits.

Un-scaled Result - No profits reinvested:
Attached Image (click to enlarge)
Click to Enlarge Name: D3S Unscaled.JPG Size: 239 KB



Notes: Of the 40 instruments in the portfolio, 23 were profitable with the balance being unprofitable. High impact drawdown periods were between 2003 - 2004, 2010-2013 and 2015. A total of 984 trades taken with a 30% win rate and positive expectancy of 0.26 x R. Where R = $500.


Unscaled annual results are as follows.
Attached Image




Scaled result with profits reinvested: To achieve a max drawdown over the period of 30%
Attached Image (click to enlarge)
Click to Enlarge Name: D3S Scaled.JPG Size: 70 KB




More to come..... :-)


C
  • #658
  • Feb 13, 2016 8:20am Feb 13, 2016 8:20am
Quoting Copernicus
Donchian Suite Tests continued.... Refer to system definition in Post 640 Strategy: D3S Type: Medium Term Swing Trader Timeframe: D1 Test Period: 1/1/2000 - 31/12/2015 Donchian Entry: 100 Period Donchian Exit: 20 Period Initial Stop: 50 Period ATR x 3 Note that all unrealised profit taken at the end of the test hence the significant profit taken on final bar of test. Note that when comparing against other methods, drawdowns usually are experienced upon portfolio commencement...
To highlight this last post. All you need is very slightly greater R:R and you will make money. It is practically unavoidable.
If you are good at something, never do it for free--Joker
  • #659
  • Feb 13, 2016 8:27am Feb 13, 2016 8:27am
Quoting jmn5611
{quote} To highlight this last post. All you need is very slightly greater R:R and you will make money. It is practically unavoidable.
Evening J. I must admit the simple Donchian break is a robust system. No matter what variables I am using (caveat within logical guidelines) it just oozes alpha when it is trending. Just spending my Saturday night back-testing and dancing on my chair with a bit of background music. How's things your end? :-)
  • #660
  • Feb 13, 2016 8:31am Feb 13, 2016 8:31am
Quoting Copernicus
{quote} Evening J. I must admit the simple Donchian break is a robust system. No matter what variables I am using (caveat within logical guidelines) it just oozes alpha when it is trending. Just spending my Saturday night back-testing and dancing on my chair with a bit of background music. How's things your end? :-)
It is Saturday morning here in the States. I am doing the same thing essentially. Tweaking things and getting ready for next week.
If you are good at something, never do it for free--Joker
3 traders viewing now