Diversified Trend Trading Approach

No probs about the questions :-)

Mate..I can't see what instrument that is? The positioning of my SDC's tends to go in a heirarchy from a short term swing on M30 to progressively extend the SDC on longer timeframes as possible. Ideally I am trying to obtain different SDC's on every timeframe and therefore I tend to go to the shortest swing move form M30 and step out for H1, H4 D1 etc......I hope that makes sense.

C

Nice entry where you have it R. No probs with that one.

C

Nice Tunes Carp. Particularly liked 'I can see clearly now". It made me reminisce about 'ye old days' in Melbourne in flares and a full head of hair sucking on a Sunnyboy






Now onto the serious stuff.

In regards to your question, what I am finding from my testing across 6 different asset classes (forex, equities, treasuries, energies, softs, metals) is that in general (apart from a few individual exceptions) all are experiencing extended phases of mean reversion at the moment. I attribute this to the fact that they are becoming more and more correlated. The only single factor that I can tie this down to is the glut of printed money post GFC which has had to find a home somewhere and no asset class has been spared. It is not the fundamentals driving these asset classes at the moment but participant behaviour supported by the behaviour of the central banks. Everyone and their dog are selling the spikes and buying the dips with the assurance that the central banks will step in if volatility takes off.

I am thinking that the bottom will occur when interest rates have to ultimately rise......but I am hopeless at this predictive stuff. It could go on for a long time...so I just press on regardless following price and not anticipating it.

The Fundie guys on this forum will know much more about this :-)

It was WTI as XTIUSD. So you draw longest SDC on M1 and shortest on M30. Are there any rules? Standard deviation of 1 should encompass 68% of data points:

 

1. [quote]By definition, one standard deviation includes about 68% of all data points from the average in what is referred to as a normal distribution pattern, while two standard deviations include about 95% of all data points.[/quote]

Do you take into account that SDC which is containing 100% of the data in trend may be wrong and should be redrawn? Or less than 68%? Probably I am overcomplicating....

Thanks!

Dear C.

How would you use all or any your systems if you would have 5 000 USD instead of 200 000 USD and wanted to use it for compounding for later and not living off it?

I am humbled by your detailed and clear answers, Mr Copernicus. Thank you!

There are no hard and fast rules aside from the need to select an obvious swing/high/low as a commencement point for the SDC and use a consistent SDC width. For me I use 1.0 std dev. You can use whatever setting you like. Some people like a wider SDC as it is more representative of price movement. For example 2.0 will encapsulate price movement more comprehensively however you will pay a price in reducing your position sizing given the basis of fixed fractional position sizing used to calculate lot size. We calculate position size according to the distance between our entry and stop (being the outer band of the SDC). If the SDC is wider the distance increases and our position size reduces if we want to risk the same % of trade capital. [/quote]

C

Good question R.....but IMO there should be no difference apart from the practicalities of your brokers lowest order sizes allowed. With microlots, you theoretically should be able to play exactly the same game. Check out different brokers and their minimum lot size offerings. I remember a podcast where a guy said that these days you can trade like a professional fund on $10K given the offerings of some brokers.

With regards to compounding (reinvesting profits) .....I would highly recommend doing it if you are live. For testing purposes I adopt non-compounding simply to remove the effect of compounding to see what is going on with the equity curve. Adopt fixed fractional position sizing of say 0.25% or 0.5% per trade on your trade capital and reinvest profits to build the capital base.

C

Mate it is my pleasure. There are not many folk interested in the SDC, which does frustrate me a bit as it is such a great analytical tool. I hope you stick with it and develop a great system from it. :-)

Cheers

C

GREAT..Stuff

Cheers D. Good to have you on board :-)

Unfortunately this is because society is more and more dumb. I see more and more people offended by "wall of text" which consists of only 10 sentences or less. People don't like to read...

http://www.educationdive.com/news/ar...teracy/421401/

Your posts are scientific and for serious, rational and logical people. And they require thinking. I was looking for such knowledge on FF and lost hope. But somehow I found your thread recently by chance.


I learned not to expect anything from anyone. What a relief that is!

I am in this place of the thread

Now I understand the broad scope of the idea of diversification. And I am asking about entries in one puny system

But if I am individual trader with limited resources I cannot run 10 strategies as standard CTA does.
Is diversification of less than 10 strategies worthwhile?

What do you think about such diversification for 1 person (as a side project):

1. DTT (4h and 30m)
2. DTTR (4h and 30m)
3. 20 instruments (currencies, futures, 1-2 blue chip stocks or indexes)

I don't think there is a way to back test this.
Only forward testing is possible (because of discretion)?

I assume you have some EAs running in your automated portfolio for every possible market condition? How is it going?

Crumbs....I pity you R. There is some heavy duty reading in some of those posts. :-)

As you get further into the thread you will see where I started getting really excited about the principles of diversification. This to me seems to be the key for longevity and as you get deeper into it you will see that you can construct purpose built models that can weather almost any storm using very simple trading models specifically designed to avoid any form of curve fitting that, when blended together, can produce superior performance metrics than what I have been able to achieve through discretionary systems. Unfortunately this road will ultimately require you to have to form a decision....whether to stay with your discretionary systems....or choose the fully automated systematic path.

Mate your proposed diversification into DTT (4hr and 30m) is a good choice as the two timeframes are sufficiently separated to produce benefits from timeframe diversification. Also you are going to get a lot of action on the M30 which is fun and keeps you active. The only issue I have with the 30 min timeframe is that if you plan to be loyal to the strategy and trade every signal you are going to have many sleepless nights.

Have you thought about D1 on EDTT and D1 on DTTR? You will lose the benefits of timeframe diversification but retain the benefits of system diversification. But importantly for practical reasons, this will enable you to focus on your alternative occupation but still participate in these strategies? You will find that there seriously is not much difference in the performance metrics of M30 versus D1 apart from far fewer trades and sleepless nights. Also that will enable you to possibly widen your investment universe to include say 20-30 instruments in the following markets with say 4 instruments from each asset class below to obtain a balanced exposure:
1. Forex;
2. CFD commodities (eg. softs, energies, treasuries, metals and indices)

With limited capital, futures can be an obstacle with lot sizes and it is probably better to trade CFD's of these futures such as that provided by Avatrade or IG markets etc. until the capital builds. Futures tend to have lower frictional costs than CFD's but you do need greater levels of capital to trade them. Futures contracts have an expiry and you need to merge discreet series through data adjustment to create continuous futures contracts if you want to effectively back-test them which takes a bit of effort.

I would also tend to shy away from individual equities and commence with indices that through CFD's offer 24 hr coverage avoiding any gaps between market close and open. You can get into individual equities later when your capital for trading builds.

You are right about not being able to back-test with a discretionary system....but what you can do as a fix for this is use the strategy tester on MT4 and place your SDC's on the bars, entry points and trailing stops as you walk through the history. I have spent many long days doing this in the past and it is better than nothing. You however will have to assume a spread and swap rate to apply to your testing but you can conservatively pitch this to give you an idea. If you adopt a standard convention to spread and swaps you at least will then be able to compare performance results of runs across different markets.

The automated portfolios I am testing is a work in progress mate. The good thing however is that with a portfolio the game is one of continuous improvement on the same portfolio through market, timeframe and system addition as opposed to trading a single strategy that may work for some time and then you are left scratching your head looking for a new strategy to trade.

I have simplified things to categorizing each system into whether they are a momentum trader or a mean reversion trader. All systems can effectively be categorised into these two buckets. Within each bucket I then look at strategy variations through system variation that help to 'plug' the drawdown gaps of the current portfolio. As I progress with system development I am slowly able to plug these deficiencies and improve my historic risk weighted returns. The key name of the game is to totally avoid any form of curve fitting and you slowly develop your skills in being able to build specific systems to address particular market conditions.. Interestingly, the systems I develop now are getting simpler and simpler as opposed to becoming more complex with filters etc. so that they can navigate a broader range of market conditions.

You would be amazed at the positive expectancy systems guys on this forum throw away that can be effectively used in a portfolio. I have often thought I should be a system horder of these failed systems on Forex Factory...... as once you understand the equity curve and performance of these supposed failed systems over a long timeframe (that actually produce a positive expectancy), they can always be put to use in filling a specific drawdown gap in the portfolio.

R....if you are interested spend some time listening to the Better Systems Trader podcast or the Chat with Traders podcast. It is right down your alley and will give you heaps of ideas to mull over. These podcasts have seriously helped me lift my game.

Two very interesting podcasts below dealing with what we have been discussing.

1. How to think about strategies like a quant and diversify like a boss
2. Daniel Fernandez On How You Trade 2000 Strategies With Only $500

Thank you, I am feeling as I am moving in light speed in my development. So much to research, so little time.

I have some history with robots long time ago with the use of (then free) https://forexsb.com to create the one, perfect, robot. Now I understand the flaw in the thinking.

If I would start tinkering with simple EAs to build robotic portfolio, do you think MT4 is enough for that? I have some robots from the past, but I can't see a way in MT4 to test them together in portfolio as you did to test their DD and CAGR collectively. Do you use custom software? I read about Access database you used.

This is interesting gimmick: https://forexsb.com/expert-advisors-generator
I used to do it all by hand.... now it is a button on website...

Hehehehe. Welcome to my disease. You can see why I need to stick to my path and avoid getting side-tracked with other projects. There is just not enough hours in the day. :-)

When people say that a single EA can't handle every market condition they are right, so they tend to shy away from EA's altogether and trust in their brain to address variable market conditions....but unfortunately they don't extend the logic and look at what happens when you combine multiple EA's which are configured for different market conditions. The ability to blend systems is virtually unlimited and goes well beyond what the brain can process. That's why I ultimately have succumbed to systematic trading. The good ol' brain just couldn't handle it.


That's a great link R....Wow....it blows me away. I would love to get into the back-end of their website to see how they are generating algo's. I have seen something a bit similar with strategy quant and a few other programs that generate algo's on the fly and associated code for MT4 or tradestation etc.....however despite their attractive lure, this is not the path I have chosen to take with my trading. Some of the EA's they generate have great results but understanding the logic of them can be very testing. I am trying to use my own logic to generate systems that respond to various market conditions and specifically respond to weaknesses in my portfolio. To do this I really need to get under the hood and understand the logic of the systems and why they perform/underperform in varying market conditions.

Yep you can do this R on MT4 but it is a hard slog. I started this process on MT4 and RLF from this forum (a very talented but grumpy bastard/tic) was kind enough to embed a bit of code into his EA's which I used that generated csv files which could be uploaded into an Access database. This works well but used to take about 3-5 days to process about 45 market instruments as one portfolio.

Inevitably I got frustrated with this and because I am familiar with MS Excel, I and Fred decided to have a crack at using this as a preferred solution with Fred producing the VBA components. Now I can punch out a new system and test it in a portfolio in a few hours all within the Excel environment. I store the results and the market data in MSAccess to build up a history and find and reuse systems that might help plug the holes in my portfolio.

Of course you could use Tradestation, Ninjatrader, Amibroker, Trading blox etc. to do this as they are much better geared for portfolio analysis than MT4...but I am so glad I chose my own solution as it gives you amazing power being able to develop things from scratch and truly understand how everything ticks.....but this won't be for everybody.

If you are familiar with MSExcel, I started a thread that provides some samples that you might want to get stuck into.....but it requires a major mind shift and a strong desire to go down this path as the learning curve is steep. My recommendation to anyone is to first become more familiar with diversification using discretionary systems such as those described in this thread before considering venturing down the systematic route.....as it is a real drain on your time and if you have commitments elsewhere, it can drown you.

I hope this helps R. Cheers C

PS You might want to subscribe to Bass's great thread. He and I are on a similar journey. There is so much excellent info in his thread particularly in regards to systematic trading. It is a valuable go to source of info for me :-)

Thank you!

Since my personality type is INTP (Architect/Logician) my talent is to create, improve, simplify systems (I own a production company where for 8 years I was doing just that).
https://www.16personalities.com/intp-personality

So it is only natural I should at least try robots diversification and choose the method which is most effective and simple. I will still run discretionary systems till automatic solutions will have better parameters. However looking at your other thread 3% CAGR for last 17 years is not enough to pull me to the dark side...It may make sense for someone with 1 million capital, but not for someone with 10 000 USD capital. I am not even mentioning 20% DD. Did the results of robot army improved lately?

I lost a lot of time using beta version of https://forexsb.com 6 years ago to design EAs. Simple ones performed the best. There was hundreds of indicators and
conditions (time, hours, if <> then etc.)

I tend to be very careful to focus on one working method and not be distracted until many months of active trading pass and I can tell if the method suits my personality and lifestyle. Now I am testing SDC and diversification.

Do you have data what was the longest losing streak of different discretionary systems in this thread?

Concerning your suggestion of 1D timeframe, it creates too large SL for me....and not enough signals. 4h and 30m creates just enough signals. Since I am in Europe, that is no problem for me.

So if Tharp's definition of 6 market states is true, then minimally you need only 6 robots for the portfolio?

I wonder what DD they had there:

I think I am digging into nuances of DTT. I use 4h and 30m and in the original post you say you draw on lower tf the same channel as on larger tf from the same point, but then you say in other post that you draw on the smaller tf from last significant low/high.

Could you please comment on EUR JPY below? I was stopped on channel drawn from latest significant low on smaller tf. I would be not stopped if I was using larger tf channel, but the SL would be gigantic.



Same here on USD SEK, but now at BE with second entry: