Diversified Trend Trading Approach

The worst thing is that my weekend is rapidly approaching a close while yours is just beginning, How about we trade places? :-)

Donchian Suite Tests continued....

Refer to system definition in Post 640

Strategy: D4T
Type: Medium Term Trend Trader
Timeframe: D1
Test Period: 1/1/2000 - 31/12/2015
Donchian Entry: 150 Period
Donchian Exit: 75 Period
Initial Stop: 75 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:



Notes: Of the 40 instruments in the portfolio, 27 were profitable with the balance being unprofitable. High impact drawdown periods were between 2004 - 2007. A total of 546 trades taken with a 25% win rate and positive expectancy of 0.53 x R. Where R = $500.

Results are improving all the time as we are stepping out into longer timeframes which is indicative of most trend following strategies. Trade activity is more restricted to significant price moves thereby filtering directionless price action. Interestingly Bill Dunn (a well known and successful CTA) is categorical in his statement that over the last 10 or so years, alpha in trend following is found in progressively longer timeframes. In the 1980's to 1990's trend following was a game played in the medium term timeframes, but with increased participation in the market and greater market efficiency, combined with greater market intervention since 2008, trend followers in general have had to extend their trend following timeframe to avoid the whipsaws. ​Of note is the concentration of the drawdown between 2004 and 2007 which needs to be overcome through portfolio blending (as a later exercise in this thread).


Unscaled annual results are as follows.



Scaled result with profits reinvested: To achieve a max drawdown over the period of 30%




Well that's it for tonight. Will continue tomorrow.

Time to unwind. Night all ..... :-)

C

Gladly ! I love this job and I cannot wait till the Wellington open. ( Well I would gladly switch unless you live in Norway, Sweden or Iceland)

Edit, but it's nice and warm in Australia right now, right?

Warm is the word. Great if you are a lizard. I really need a change with a freezer and some snow shoes. :-)

Guys

I am going to have to go into hibernation for a while to adjust the spreadsheet model to reflect performance in association with Net Asset values of the portfolio as opposed to the current basis which only refers to realised gains.

You will have noticed that at the end of each test, we have considerable un-realised profit retained in the hypothetical portfolio which is closed out at the end of the test and leads to those significant final month profit results. I have realised today that the models I have been testing at a consolidated system level significantly outperform the TF Index but this is only reflected in the final profit result. I was wondering why the going was so tough and what fictional system I would need to include to get the damned equity curve mirroring the big boys....... and then suddenly the lightbulb hit me. We are actually outperforming the CTA's with this hypothetical modelling but you just don't see it reflected in the equity curve until the very end of the test when the entire portfolio is closed down. Other factors such as the drawdown also cannot be directly compared as the drawdown in the hypothetical modelling is based on drawdown from realised equity...as opposed to unrealised equity.

To make any sense or comparison between equity curves from the CTA's and this hypothetical exercise, it will be necessary to evaluate the NAV of the hypothetical portfolio on a monthly basis and also determine drawdown in relation to drawdowns from peak unrealised equity. It won't take long to adjust the model but bear with me.

Once this is done, I will return to this analysis and continue on with the process of finalisation and portfolio blending. At the same time I will be adjusting results to reflect management fees, performance fees, administration and custodian fees etc so we can get a closer comparison and something that makes sense going forward.

I hope this makes sense. Signing off for now but will return to finish off what I started. There is a light at the end of this tunnel. :-)

Cheers

C

It certainly depends on the meaning you give it. Any strategy is essentially a filter or set of filters that discriminates between being long, short or flat based on market conditions.

Its not uncommon to notice that a strategy performs better when you apply a trend filter. I could say that a system of entry via coinflip suddenly becomes better when you have a filter that only takes signals in the direction of the trend. Still doesn't relate to what I meant.

The idea being challenged is that more is better. The notion that when ten indicators generate a signal the quality is somehow better simply due to the number of indicators used. Some people describe this spurious concept as confluence.

All things equal, less is preferable. On Forex Factory you see the negative example of this being demonstrated with great exaggeration.

Donchian Suite Tests - Revised to directly compare with the Fund Managers monitored in this thread.

Refer to system definition in Post 640

Right folks we are now back on track..........

Following on from post 665 of this thread, I have now adapted the spreadsheets to be able to more effectively directly compare performance results of the Donchian systems being currently tested against the fund performance of the professional fund managers monitored in this thread. You will see a significant difference in the equity curves from previous posts that now, much more closely follow the TF Index equity curve and possesses far better correlation. Given the improved comparative measures now adopted, I will be revisiting the performance of Donchian systems already assessed commencing from post 643 to post 662.

.......but to get in the mood.....you guessed it :-)




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


Un-scaled Result - No profits reinvested - Gross Performance (before FM fees):



Un-scaled result with profits reinvested - Net Performance (after FM fees):


Note: Rescaling to reduce drawdown impacts will be conducted during the portfolio blending phase of this exercise.



More to come :-)

Regards

C

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


Un-scaled Result - No profits reinvested - Gross Performance (before FM fees):




Un-scaled result with profits reinvested - Net Performance (after FM fees):




Note: Rescaling to reduce drawdown impacts will be conducted during the portfolio blending phase of this exercise.



More to come :-)


C

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


Un-scaled Result - No profits reinvested - Gross Performance (before FM fees):



Un-scaled result with profits reinvested - Net Performance (after FM fees):




Note: Rescaling to reduce drawdown impacts will be conducted during the portfolio blending phase of this exercise.



More to come.

PS You have to laugh with Covel. He really doesn't like the CNBC predictive guru's and especially those from the buy and hold fraternity. Enjoy the podcast :-)
http://www.trendfollowing.com/2016/0...llowing-radio/



C

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


Un-scaled Result - No profits reinvested - Gross Performance (before FM fees):



Un-scaled result with profits reinvested - Net Performance (after FM fees):



Note: Rescaling to reduce drawdown impacts will be conducted during the portfolio blending phase of this exercise.


Thats the short term Donchian systems done. Tomorrow I will move onto the medium term systems.

Cheers


C

......time to put the feet up. :-)

Donchian Suite Tests continued.....

Refer to system definition in Post 640

Here we go with the Medium range trend following and swing systems.

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


Un-scaled Result - No profits reinvested - Gross Performance (before FM fees):




Un-scaled result with profits reinvested - Net Performance (after FM fees):





Note: Rescaling to reduce drawdown impacts will be conducted during the portfolio blending phase of this exercise.


More to come.

C

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


Un-scaled Result - No profits reinvested - Gross Performance (before FM fees):




Un-scaled result with profits reinvested - Net Performance (after FM fees):




Note: Rescaling to reduce drawdown impacts will be conducted during the portfolio blending phase of this exercise.


More to come.

C

Donchian Suite Tests continued.....

Refer to system definition in Post 640


Strategy: D4T
Type: Medium Term Trend Trader
Timeframe: D1
Test Period: 1/1/2000 - 31/12/2015
Donchian Entry: 150 Period
Donchian Exit: 75 Period
Initial Stop: 75 Period ATR x 3


Un-scaled Result - No profits reinvested - Gross Performance (before FM fees):





Un-scaled result with profits reinvested - Net Performance (after FM fees):




Note: Rescaling to reduce drawdown impacts will be conducted during the portfolio blending phase of this exercise.


More to come.

C

Donchian Suite Tests continued.....

Refer to system definition in Post 640


Strategy: D4S
Type: Medium Term Swing Trader
Timeframe: D1
Test Period: 1/1/2000 - 31/12/2015
Donchian Entry: 150 Period
Donchian Exit: 30 Period
Initial Stop: 75 Period ATR x 3


Un-scaled Result - No profits reinvested - Gross Performance (before FM fees):




Un-scaled result with profits reinvested - Net Performance (after FM fees):




Note: Rescaling to reduce drawdown impacts will be conducted during the portfolio blending phase of this exercise.


That's it for the medium term traders. Next up come the long term trend and swing traders and then we are done and can commence blending portfolios to gain benefits from system diversification.

Be back soon.


C

Donchian Suite Tests continued.....

Refer to system definition in Post 640

Sometimes you just have to slow it down.....




Here we go with the longer timeframe Donchian traders.

Strategy: D5T
Type: Long Term Trend Trader
Timeframe: D1
Test Period: 1/1/2000 - 31/12/2015
Donchian Entry: 200 Period
Donchian Exit: 100 Period
Initial Stop: 100 Period ATR x 4


Un-scaled Result - No profits reinvested - Gross Performance (before FM fees):



Un-scaled result with profits reinvested - Net Performance (after FM fees):



Note: Rescaling to reduce drawdown impacts will be conducted during the portfolio blending phase of this exercise.


More to come.


C

Donchian Suite Tests continued.....

Refer to system definition in Post 640

Lifting the tempo a notch.....


Strategy: D5S
Type: Long Term Swing Trader
Timeframe: D1
Test Period: 1/1/2000 - 31/12/2015
Donchian Entry: 200 Period
Donchian Exit: 50 Period
Initial Stop: 100 Period ATR x 4


Un-scaled Result - No profits reinvested - Gross Performance (before FM fees):



Un-scaled result with profits reinvested - Net Performance (after FM fees):



Note: Rescaling to reduce drawdown impacts will be conducted during the portfolio blending phase of this exercise.


Only two more to come and then we are on to the exciting phase of portfolio blending where things hot up.




C

Donchian Suite Tests continued.....

Refer to system definition in Post 640

Strategy: D6T
Type: Long Term Trend Trader
Timeframe: D1
Test Period: 1/1/2000 - 31/12/2015
Donchian Entry: 300 Period
Donchian Exit: 150 Period
Initial Stop: 150 Period ATR x 5


Un-scaled Result - No profits reinvested - Gross Performance (before FM fees):



Un-scaled result with profits reinvested - Net Performance (after FM fees):



Note: Rescaling to reduce drawdown impacts will be conducted during the portfolio blending phase of this exercise.


Only one more to come before we commence portfolio blending.

Cheers

C

Some Preliminary Simple Blends

While waiting for the final Donchian Test to finalise, I thought I would post a few simple blends to get the appetite working.




Portfolio Blend: D5S_D3S
Blend details: D5S (50%)/ D3S (50%) - Scaled x 8 (Trade Risk = 2% of Equity)

Performance Results (after Fees)


Notes:
A powerful simple system combination producing a CAGR (after fees = Net) in line with the TF Index. The Gross Trading Profits provide a guide to the trading performance outside of a fund structure. However......to achieve these Net results requires you to sit through drawdowns of up to 50%. This unfortunately is the name of the game for trend trading. It is important to note that drawdown is directly proportional to time. The greater the trading history, the greater the anticipated drawdown that will be experienced. A number of the CTA's such as Dunn Capital are prepared to tolerate this level of drawdown and treat a drawdown as a necessary evil that is required if you plan to stay in the game a long time and desire performance returns that significantly exceed the buy and hold investment community.

Just be aware that the S&P500 (Total return) index over the same period also experienced similar levels of drawdown of 40% (Gross)-50% (Net) yet its Gross performance returns were approximately 1/3 that of the TF Index.

Portfolio Blend: D5T_D3S
Blend details: D5T (43%)/ D3S (57%) - Scaled x 7 (Trade Risk = 1.75% of Equity)

Performance Results (after Fees)




Portfolio Blend: D5T_D1T
Blend details: D5T (43%)/ D1T (57%) - Scaled x 7 (Trade Risk = 1.75% of Equity)

Performance Results (after Fees)



More to come tomorrow.

Cheers

C

Donchian Suite Tests continued.....

Refer to system definition in Post 640

Here we go. The final Donchian test in the suite.

Strategy: D6S
Type: Long Term Swing Trader
Timeframe: D1
Test Period: 1/1/2000 - 31/12/2015
Donchian Entry: 300 Period
Donchian Exit: 75 Period
Initial Stop: 150 Period ATR x 5


Un-scaled Result - No profits reinvested - Gross Performance (before FM fees):





Un-scaled result with profits reinvested - Net Performance (after FM fees):






Note: Rescaling to reduce drawdown impacts will be conducted during the portfolio blending phase of this exercise.




Now we will be able to move onto some further more comprehensive blending exercises.

C

Blended Portfolio Examples

Below are 5 different blended portfolio examples. You will note that there is little variation between the options which is highlighting that each option is capturing the majority of alpha to be had across the portfolio over the timeframe. It is also symptomatic of the robustness of the Donchian breakout technique as all contributing systems (from short term to long term) possessed a positive expectancy over the test period. This is most unusual but augers well that we are dealing with statistical validity as opposed to random luck.

Pay little attention to the initial drawdown achieved on commencement as this is a feature of the slow start up of the portfolio as opposed to market conditions during this period. You will note that there is a departure in the equity curve from the TF Index between 2000 - 2002 and this is attributed to the fact that the CTA's tracked in this exercise had been in operation well before 1 Jan 2000 and their portfolio's were already operating at optimum efficiency. On the other hand, the hypothetical portfolio commenced trading short term trades shortly after commencement, but it took between 1 to 2 years before medium term and long term traders were operational whereby system diversification benefits started taking effect. You will visually observe how similar the equity curve profiles between the hypothetical system and the TF Index were from 2002 to 2015 which in reality would have translated to a far higher correlation.....however I am not going to go back to the drawing board to commence the portfolio earlier and track performance from 1 Jan 2000 with the system operating at maximum efficiency. You will just have to take my word for it.

The important drawdown period was between 2011 - 2013/2014 which was symptomatic of all trend following CTA's and reflective of generally poor trend trading conditions during this period. If we want to reduce this impact we will need to diversify into other systems that were thriving during this period. The key to a smoother equity curve is therefore going to be found in balancing trend/counter-trend systems with mean reverting systems.

Despite the feature of ugly drawdowns which is common to all medium to long term trend following systems, the performance returns (in reality land) over a 16 year period are impressive and effectively triple the returns available to conventional buy and hold investment logic.

DAll EW Net - Donchian Suite - All Systems (Equal Weighted allocation)



DAll LT Net - Donchian Suite - All Systems (Long Term bias allocation)


DAll MT Net - Donchian Suite - All Systems (Medium Term bias allocation)


DAll ST Net - Donchian Suite - All Systems (Short Term bias allocation)


Test Notes, Findings and Opinions:
1. The entire test avoided any propensity for curve fitting. No selection bias was made in relation to the choice of instruments traded provided that they were liquid and had a sufficient data history to be included in the tests. All variables were chosen prior to commencement and were not a result of any optimising. This illustrates the robustness of very simple trend following methods.
2. Entry and exits play little part in the success of the models. Predominant contributors to alpha were from instrument diversification, system diversification and money management (position sizing). IMO a focus on technique is simply tilting at windmills and you are more likely to be the subject of random luck as opposed to a definitive edge.
3. As much alpha was squeezed out of trending market conditions as possible using this technique across multiple time-frames including trend trades, swing trades both in-trend and counter-trend. The similarity of results between this hypothetical test and the successful FM's in the trend following space suggests that the reality is that a retail trader is unlikely to outperform the best FM's out there as the alpha made available is already being captured by the big boys.
4. Discretionary traders may be more nimble than the automated system traders and possibly could squeeze a bit more alpha out of the market, but is it realistic to assume that a discretionary trader can trade as many as 12 discreet systems with the same degree of fidelity as systematic traders?
5. If you expect to escape drawdowns when seeking high performance, you are simply dreamin'. You must become familiar with the long term performance history of your system and it's historic drawdowns as it is extremely likely you will be putting a potentially good system into the dustbin thinking that it's broke....when in fact the drawdown is just part and parcel of life as a trend trader.
6. Learn from the Big Boys, as they really know what they are doing and the best out there are performing at peak trading efficiency. Strive to understand their processes and logic as this is where the real edge in this market lies buried.
7. Trend following has been a tough game since 2000 and offers far lower returns and higher volatility than that experienced in the previous 20-30 years. In part this can be attributed to increased market participation, government intervention and a large part can also be contributed to the lack of free returns offered by far lower interest on unutilised equity (which in the 80's to 90's could mean an addition of 3-5% CAGR than what was available between 2000 to 2015. Despite the difficult patch, trend following is in no shape or form 'dead' and offers excellent wealth creation ability provided of course you are prepared to ride the roller-coaster with the variability of returns.
8. The great returns for trend following are inevitably experienced when markets experience 'crisis'. So when the proverbial hits the fan in the markets, at least you know who will be taking your money.

The systems described above will be included in my monthly performance monitoring and we will be able to compare how these models fair in walk forward mode against the CTA's. I will re-post December performance results in my next post with these systems included so we can see how they ranked against the FM's.

Cheers C