Diversified Trend Trading Approach

Non-linear systems that allow for non-deterministic (intractable) outcomes necessitate probabilistic heuristic mind sets to address them. To have any attempt of matching your mental models with the reality you need to embrace bayesian thinking, fat tailed distributions and asymmetries. So let's have a look at spies, crime and lightning strikes to see how.

Ways to distinguish between market convergent and divergent phases.

I have a lot of time for John Ehlers who with his electrical engineering background approaches market interpretation using methods to distinguish between signals and noise and has developed numerous adaptive indicators.

Below is an intriguing line of inquiry.




...and here is more info on MESA and what it is attempting to do.

.....and the obligatory podcasts.

 

1. Indicators, DSP & Cycles with John Ehlers
2. Adapting to market conditions with John Ehlers
3. A brand new, more responsive indicator with John Ehlers

Enjoy

C

On Complex Systems such as our Markets

Complex systems seem to display thresholds of complexity which, when reached give rise to sudden non-linear jumps in the complexity and emergent structural relationships. The superposition principle of simply additive construction does not apply to nonlinear components. This is the same as saying that you can never create a complex system with any number of linear components. You need the threshold steps of non-linearity to achieve this. It is the reason why we need neurons and transistors to create a brain or AI. The neurons and transistors apply threshold constraints to signals allowing for non-linearity and emergent complexity.

Here is an example of what we are talking about applied to electrical circuit boards from the book "Hidden in Plain Sight 9": by Andrew Thomas









So what causes non-linearity in the markets. Answer: the ability to match buyers and sellers with different trade volumes. Result is that the market is dynamic and possesses emergent qualities like any complex system. To respond to this prevalent market feature just like any other complex system where chaos and complexity prevails, you need to adopt adaptive systems to tackle them. That is why market conditions change and why conditional changes are non-predictive in nature as we simply cannot anticipate when the thresholds for new emergent complexity will arise due to its intractable nature. The fat tails are a direct result of these threshold steps which dramatically alter market behaviour.

Monday morning and already +4% ?



What is happening?



USD and CAD strength came back. Also, TRY weakness.



USD CHF, that was a close call, we almost lost you:



Gold new lows:

Less movement in my universe but still doing ok. It would be nice to lock away these small gains. I would prefer portfolio heat to be +3% as opposed to -4% as when the UK and US sessions open....the sumos like to come out and play. Let's hope they are full of saki and having a bit of a snooze at the moment :-)

NZDUSD on KAMA

Ouch.....but we knew it was going to be bad from EDTT :-(

16 consecutive losses makes you seriously start questioning your holding power.

Yes....it was also a loser on EDTT over the same period...but not to the same degree due to the ATR offsets.

Still.....you can't win them all. You can clearly see where market conditions changed in the equity curve.

More work needs to be done however as the Donchian break entry is a red rag to a bull to mean reverters.

I am going to wait for the EA for any further testing. This manual method is painstaking.

This reminds me:



Physicist see markets as waves, complex systems etc.
Electrical engineers see markets as resistors....
Dancers say market is dancing etc. but still some made money (https://www.amazon.com/How-Made-000-.../dp/1614271690)

Perception tainted by previous conditioning.

LOL :-)

Transformers are here

https://www.reddit.com/r/ThisIsGeniu...rmers_cosplay/

and view as if from another planet:

Calbuco Volcano, Chile - Francisco Negroni [3500 x 2231]

I have a lot of green heat floating....good problem to have:



System diversification in action:



Kama has different static portfolio internally non-correlated.
EDTT is on 30M and still using dynamic portfolio.

Christ MP. Not sure how I missed this gem of a post. I think in the euphoria of testing DJI30 that a myopic focus tends to make my mind miss things.

Yes mate....trading the equity curve is a great idea and something that we definitely will be considering. There is lots of info about this by the CTA's who use this concept particularly for trend following and it is very succesfully deployed by Funds of Funds who rebalance their fund portfolios by only allocating to FM's who are experiencing a drawup as opposed to a drawdown. We also can use it as we trade live in parrallel with our walk forward testing. Now that it looks like we are on a path eventually to full automation, this is something we can definately test with our instrument selections in our portfolios.

We are of the opinion that there are two broad phases of market condition.....namely divergent conditions versus convergent conditions.....and each of these can have various volatility profiles.

These market conditions can stay in force for many years and we feel that without the benefit of hindsite....that it is virtually impossible to predict when they convert from one to the other. The best we can do is *assume* that markets can lock into this behaviour for extended periods and therefore one of the best tools we have to identify when these phases are in place is to use our equity curves to identify them. Guys like John Ehlers may have methods to indicate when these conditions change....however they always will be lagging as you need an extended data set to identify past market behaviour in attempt to identify when the *change occurs*. We however can use our equity curve in a similar manner.

For example....take this recent example of NZDUSD



We believe that we could deploy either three mechanisms to respond to this issue.

1. A filter that turns off the instrument after a succession of losses and turns on that instrument when drawup is detected;
2. Have a ranking system that selects from a list by virtue of the equity performance of the walk forward test. Poorly performing instruments get progressively filtered out of the ranking; or
3. Have the instrument on all the time and drown it's meager performance with the results of the balance of the portfolio.

At the moment we are using option 3....thereby not having to 'predict' when to turn off or on the instrument....however now that we are moving to a dynamic as opposed to a fixed portfolio where we need ranking systems for deployment of a potential infinite selection of instruments....the ranking system is going to be tested.

One of the biggest drawbacks of the EDTT is its discretionary nature where analysis such as this becomes very subjective and difficult......however a priority of mine is to get rid of any discretion so we can be objective in our assessment and dig deeper into what the 'reality' is telling us.

Cheers C

Now it is time for me to goggle :-)



Nice one R.

With KAMA we are going to be able to extract a vast number of options with correlation benefits. We are well on the way now with system diversification.

Still need to address a few small loose ends but I am loving our direction in KAMA progress at the moment.

In the hustle and bustle of daily life we often fail to consider how an appropriate environment can super charge our mental horsepower. The modern office is perhaps the very worst environment for 'coherent thinking'. Time to go back to the cave and sit around the fires while we think clearly.

We have the EA and now we can blast through testing....but note that at the moment data on 5 min is limited so back tests are running from 1 Aug 2017 to 30 Apr 2018 as we compile data for portfolio crunching later.

Trade risk $500 fixed for position sizing across all instruments.

Working off KAMA settings of 10,3,30 and timeframes M30 and M5. Down the track I will also test across the longer timeframes of say H4 and M30 as T suggested.

All trade activity operates off the open of the bar only.

Lets start with the euphoric DJI for the period,

Karma on DJI30



Total Return for 9 month period: 43%
Drawdowns: To be determined later in analysis phase.

Comparison against EDTT return over same period of 26.28%

Notes: Equity curve demonstrates that when the nice and steady trends are on....this technique nails it. Despite the euphoric returns we need to dial down volatility of the return stream through position sizing. We are too leveraged at current levels.

ATR position sizing is bogus and responsible for a lot of the volatility of the return stream given variation in lot size. Need a new measure such as that applied by EDTT.


Karma on ASX200



Total Return for 9 month period: -9%
Drawdowns: To be determined later in analysis phase.

Comparison against EDTT return over same period of -1.24%

Notes: Equity curve demonstrates the impact of mean reversion. We need to incorporate a stepped entry off the Donchian Break in a future version of this EA to push entries away from major support/resistance zones where false breakouts prevail....like we did with EDTT. As consecutive losses build we need to progresively push away from range bound conditions.

We also need to dial down return stream volatility as discussed previously. Note comparives against EDTT. We are too leveraged with a $500 position sizing risk multiplier.

Also note the lot size problem using ATR position sizing. Use another method such as R suggested (eg. Clenow vol adj position sizing etc.)


Karma on AUDCAD



Total Return for 9 month period: -20%
Drawdowns: To be determined later in analysis phase.

Comparison against EDTT return over same period of -3.36%

Notes: Clearly demonstrates that there is a lot more work to do in responding to whipsaws from mean reversion and significant volatility associated with position sizing. Currently we need to dial down leverage by about 3x - 5x to address return stream volatility but in doing so we significantly dilute the power of KAMA in trends (eg. DJI30). If we cannot address these issues on this timeframe combination then consider also testing longer timeframes to see if trending with KAMA is better applied as T suggested.

Karma on AUDJPY



Total Return for 9 month period: -6%
Drawdowns: To be determined later in analysis phase.

Comparison against EDTT return over same period of -1.08%

Notes: Refer to comments above. Don't lose heart as there is a blue light at the end of this tunnel. KAMA has what it takes....but steps need to be taken to address the current issues. On a pound for pound basis using appropriate vol adjusted scaling I can see a strong similarity in return profile to EDTT with correlation benefits. Any trend following system applied to these timeframes would see similar return profiles.....but where the difference lies in the steps taken to manage adverse volatility.....and we have lots of tricks up our sleeve yet.....and remember this period we are testing has been a dog of a time for trend followers in general. Just refer to the CTA stats over this period. Most FMs were experiencing their highest drawdowns for the last 20 years during this period.


Karma on AUDUSD




Total Return for 9 month period: 2.4%
Drawdowns: To be determined later in analysis phase.

Comparison against EDTT return over same period of 1.84%

Notes: Refer to comments above. Also note the unrealised equity curve of the charts above suggesting that at times significant profits can be left on the table with exits.

Karma on BRENT




Total Return for 9 month period: 7.7%
Drawdowns: To be determined later in analysis phase.

Comparison against EDTT return over same period of 2.16%

Notes: Refer to comments above.

Karma on CAC40




Total Return for 9 month period: -2.5%
Drawdowns: To be determined later in analysis phase.

Comparison against EDTT return over same period of 3.60%

Notes: Refer to comments above. Noticeable impact on whipsaws off Donchian without offsets applied like EDTT. Big impact of a few losses....Disadvantage of no stops in this particular test.

Karma on CADJPY



Total Return for 9 month period: 3.1%
Drawdowns: To be determined later in analysis phase.

Comparison against EDTT return over same period of -3.52%. Now that's a nice surpise

Notes: Refer to comments above. Swings and roundabouts guys. One giveth and one taketh away. Good for correlation in the portfolio


Karma on EURAUD



Total Return for 9 month period: -11.45% Ouch!!!!!!!!!
Drawdowns: To be determined later in analysis phase.

Comparison against EDTT return over same period of -1.80%.

Notes: Refer to comments above. Swings and roundabouts guys. One giveth and one taketh away.

That's it for the moment. I will continue to crunch away and then post the portfolio results for comparison recognizing that this is still a work in progress :-)

C

The Need for Benchmarks as a Guiding Reference tool

Attached Image

The system development project we are currently working on is all coming together slowly and it appears we are on the right path.

One of the exercises that we do on this thread which is not a worthless exercise is the monitoring of trend following fund managers. One of the crucial pieces of info we gather from studying these benchmarks is an understanding of performance realities of trend following over the long term.

If we didn't have these benchmarks to guide us and I produced those equity curves in the prior post.,.....to those that had no benchmark reference point to work on.....following their perusal of those equity curves they would have abandoned this system and system hopped towards a better looking suite of equity curves....that are inevitably curve fit.

This entire process we have used to-date has avoided any form of curve fitting and the logic we use is allowing us to generate meaningful and realistic equity curves that allow us to assess whether the system has legs or not in trending and non trending regimes.

The period we are using at the moment is definitely a non-trending regime which is evidenced by the performance of the diversified fully systematic trend following CTA's, however what we are finding is that our performance metrics are keener than the competition.

I am looking forward to putting this all together in terms of portfolio analytics....as this blood sweat and tears is starting to take shape.

The good thing to also consider is that there are many things we can address here to possibly further improve the metrics considering the generic form that the EA is in now.

Cheers C

To show performance of DTT, I made an explore. Maybe new comers will gain some confidence

Very nice L. Great to see you reaping benefits from it:-)

Would it be possible for you to post these trades in some excel/csv format?