Is the built-in Donchian system too good to be true?

Discussions about the testing and simulation of mechanical trading systems using historical data and other methods. Trading Blox Customers should post Trading Blox specific questions in the Customer Support forum.
mgdpublic
Roundtable Knight
Roundtable Knight
Posts: 117
Joined: Tue May 12, 2009 1:35 pm

Is the built-in Donchian system too good to be true?

Post by mgdpublic » Mon Jul 06, 2009 3:24 pm

I've run the Donchian system on a relatively large portfolio of futures and got a Sharpe Ratio of 1.3 over 18 years. That just seems to good to be true. I would think if it were that easy everyone would be doing it. Does the fact that a BO system like that is standard with a testing package mean that it's days are numbered? Will this go the way that tight stops did starting in '96?

Asamat
Roundtable Knight
Roundtable Knight
Posts: 175
Joined: Fri Jun 03, 2005 7:50 am
Location: Walldorf, Germany

Post by Asamat » Mon Jul 06, 2009 4:32 pm

You ask us to predict the future, don't you?

We can only tell - as you did - that in the past it worked. Also I think you will find that to run it in reality is not as easy as it seems in the backtest.

Chris67
Roundtable Knight
Roundtable Knight
Posts: 1046
Joined: Tue Dec 16, 2003 2:12 pm
Location: London

Post by Chris67 » Tue Jul 07, 2009 3:11 am

I think there are big differences between the Donchonian b/o system and the original style pure break-out system such as that run by the Turtles in the 1980's - obvioulsy that is for you to work-out.
It may well be that the Donconian system's days are numbered - and then again it maybe the Wolrd's most profitable trading system for the next 10 years.
I run a version of it and have done for a number of years. I'm currnetly in a draw down that started in December 2008 and its not the first draw down of that length - if you think thats easy - try runing it !
Still averaging an MAR > 1 and compounding an average of 20%+ for the last 4.5 years mind ?
Chris

LeviF
Roundtable Knight
Roundtable Knight
Posts: 1428
Joined: Mon Dec 22, 2003 12:24 pm
Location: Des Moines, IA
Contact:

Post by LeviF » Tue Jul 07, 2009 9:18 am

A Donchian channel system will basically accomplish the same task as a volatility breakout system, as a MA crossover system, etc. For any of these systems to work, there must be trends. If there are trends, then all of these systems will get you aboard at some point or another. If there are no trends, than these systems will not work. It is as simple as that.

If you think there will be trends in the future than trade them. If there will not be trends than I dont know what to do because I have not been able to develop a profitable, non-trendfollowing system. Thankfully, since the dawn of the markets, there have been trends.

Algonquin
Roundtable Knight
Roundtable Knight
Posts: 158
Joined: Sun Jan 28, 2007 9:38 pm
Location: Greenwich, CT

Post by Algonquin » Tue Jul 07, 2009 4:43 pm

Try running the Donchian system with realistic transaction costs and you will have answered your question. The performance results are very different.

As already mentioned, the Donchian system is a basic trend following system that will profit from trends in roughly the same way as an ATR breakout system, a moving average cross-over system, etc. In my personal experience, there are definitely better systems to be discovered.

Incidentally, if you would like a fair estimation of what results you can expect from a top-notch real world system (over time), a quick perusal of iasg.com or autumngold.com should be insightful.

Algonquin

mgdpublic
Roundtable Knight
Roundtable Knight
Posts: 117
Joined: Tue May 12, 2009 1:35 pm

Post by mgdpublic » Tue Jul 07, 2009 8:04 pm

What do you all consider realistic transaction settings in TB? Thanks.

Asamat
Roundtable Knight
Roundtable Knight
Posts: 175
Joined: Fri Jun 03, 2005 7:50 am
Location: Walldorf, Germany

Post by Asamat » Wed Jul 08, 2009 3:39 pm

Algonquin wrote:In my personal experience, there are definitely better systems to be discovered.
What are the better systems you suggest are possible to discover?

LeapFrog
Roundtable Knight
Roundtable Knight
Posts: 695
Joined: Mon May 17, 2004 4:18 pm
Location: Boston, MA

The Holy Grail System

Post by LeapFrog » Thu Jul 09, 2009 10:35 am

Here is what a holy grail system looks like. I believe B. Madoff offered something like it but that didn't work out too well in the long run.
Attachments
The Holy Grail.PNG
A Holy Baloney Grail System
The Holy Grail.PNG (18.9 KiB) Viewed 8943 times

Algonquin
Roundtable Knight
Roundtable Knight
Posts: 158
Joined: Sun Jan 28, 2007 9:38 pm
Location: Greenwich, CT

Post by Algonquin » Fri Jul 10, 2009 8:21 am

Transaction costs in the real world will vary greatly based on the markets being traded, the entry mechanism, the frequency of trading, the size of the traded account, the specific contract being traded, the skill of the trader, the trading session, and so on. So it is hard to generalize. However, you should infer from the public, real-world results of CTAs that it is not as simple as plugging 5% slippage into your backtest parameters. If you step your various transactions costs on the Donchian system you will see marked differences in performance.

As for better systems to be discovered, they are out there. There are long term trend following systems whose performance did not degrade over the past several years. While I cannot go into the specifics of those systems, I will say that if you using popular trend following strategies, incorporating well-known techniques, and following all of the conventional trend following wisdom, your results will be equally commonplace.

Algonquin

AFJ Garner
Roundtable Knight
Roundtable Knight
Posts: 2040
Joined: Fri Apr 25, 2003 3:33 pm
Location: London
Contact:

Post by AFJ Garner » Fri Jul 10, 2009 11:18 am

8)
Last edited by AFJ Garner on Mon Jul 20, 2009 7:21 am, edited 1 time in total.

Algonquin
Roundtable Knight
Roundtable Knight
Posts: 158
Joined: Sun Jan 28, 2007 9:38 pm
Location: Greenwich, CT

Post by Algonquin » Fri Jul 10, 2009 12:43 pm

AFJ Garner,

Yes, you can find "a version of the Donchian system whose performance has not degraded in recent years." But this is true for almost every well-known, simple trend following system! If you try enough parameter combinations you will most certainly find a system that took precisely the right trades in order to sidestep the enormous volatility that beset most trend followers for the last 6 or 7 years. But is that really the criteria for "a very good system?"

When I have examined the Donchian system over the past several years of market data what I have found is that the parameter sets are not stable. Whereas you could once vary parameter sets generously with the Donchian system and not expect large variances in your results, that is no longer the case. Most parameter sets for the Donchian system have degraded quite substantially, in my testing at least. And I have found this to be the case for every simple trend following system included in TB. Other, lesser-known strategies to which I have referred do not even come close to suffering the same levels of degradation. They are simply more robust, in my opinion, which is my personal definition of a very good system.

Why? My theory is that in certain markets today virtually all hedger risk premia is being extracted by momentum funds. The only premia left to be gotten for momentum players is from other trend followers and speculators. As the marketplace gets more and more crowded, those competing with the same strategies will encounter overcrowded trades, and with it increased volatility, less meaningful price information, and degraded risk adjusted returns. And I expect this Red Queen Effect to continue its inexorable progression. Well known systems in use today -- even those that have not degraded as much as others -- are at great risk of sudden failure as the populations of those trading strategies reach their respective tipping points. Just my humble opinion.

Algonquin

AFJ Garner
Roundtable Knight
Roundtable Knight
Posts: 2040
Joined: Fri Apr 25, 2003 3:33 pm
Location: London
Contact:

Post by AFJ Garner » Fri Jul 10, 2009 1:24 pm

8)
Last edited by AFJ Garner on Mon Jul 20, 2009 7:22 am, edited 1 time in total.

Chris67
Roundtable Knight
Roundtable Knight
Posts: 1046
Joined: Tue Dec 16, 2003 2:12 pm
Location: London

Post by Chris67 » Sat Jul 11, 2009 4:47 am

I have to say I totally Agree with Mr Garner on this subject
Unfortunately what is being effectively said here is that markets wont trend anymore .. if they dont then no TF systems will work .. if they do the Donchian system will work very well.
The argument that more players coming into the mrket will degrade systems is utter nonsense as regards to TF ... and lacks an awful lot of logical thought .. the places where systems will degrade as more players come into the market are shorter term systems and convergence systems / mean reversion / arbitrage. There is actually a very intelligent argument to suggest that the more trend followers that enter the market the more markets will actually trend !!!
However and I once again subscribe to a mantra that I have rattled on about for the best part of 10 years ... more and more players will not come into trend following .. for the simple reason that investors dont like draw downs/volatility and don't understand simplicity still maintaining to you need to be a ph.d wizzkid to make money in the markets - all the eveidence suggest that its these types who blow up most frequently
I could name numerous examples of one off funds that pursue these "holy grail strategies" that have more AUM that the entire amount of money that has come into TF in the last 5 years.

All in all evidence speaks louder than words .. a basic Donchonian system made easily more than 100 % last year ... markets moved and trended ... there is no degredation whatsoever .. what there is an overflow of eop-le trying to design low volatility low draw dwon high frequency trading systems (look at the hedge fund job adverts on efinancialcareers.com) .. which tells you that the next few years will be excellent for trend followers and that holy grail systems will continue blowing up as they always have and always will.

Algonquin
Roundtable Knight
Roundtable Knight
Posts: 158
Joined: Sun Jan 28, 2007 9:38 pm
Location: Greenwich, CT

Post by Algonquin » Sat Jul 11, 2009 9:51 am

Chris67,

Since the early 90s critics of trend following have been proclaiming the imminent death of the strategy. Some of its adherents have been responding that markets will forever trend so the strategy will remain viable. Who was correct?

Neither. Some trend followers prospered over this time period. Others closed their doors. Why?

Because trend following isn't as simple as just "following trends." If it were that easy we'd all be stinking rich. In order to successfully trend follow a trader must choose the right portfolio, select a winning bet size, find a trend, enter at an advantageous point, manage risk during the trade, and exit at the appropriate time. Some trends allow this process to occur in a profitable manner. Many do not, as highlighted over the past several years.

It is beyond debate that more money has entered trend following. Just run a search on the internet for the total assets under management for managed futures. And similarly there is no debate that simple trend following methods have seriously degraded. Run a simple moving average crossover system and see what the risk adjusted returns have looked like over the past several years.

Will markets always trend. Of course. Will simple, well-known systems always be able to profit from the trends of the future? I don't know, but the accumulating evidence would suggest not.

Algonquin

Chris67
Roundtable Knight
Roundtable Knight
Posts: 1046
Joined: Tue Dec 16, 2003 2:12 pm
Location: London

Post by Chris67 » Sat Jul 11, 2009 11:57 am

Algonquin

You fall into the common trap I hear so much in this industry of Quoting the term "manged futures'" .. a generic and meaningless term tht has nothing to do with anything .. its a bit like saying that cars are crashing so therefore the "transportation" industry is more dangerous ...
The whole financial industry is full of lose meaningless terms .. Trend following does not equal managed futures. Managed futures are funds that trade futures ... there are hundreds if not thousands of ways to trade futures
I have spent years building a definitive data base of just about every trend follower/ trend following program out there ... its a huge database .. and the results are absolutely startling
The amount of new funds to have emerged in the last few years is very small , the AUM they have is also small.
How many funds out there trade equities ... 100,000, 500,000 , 10 million ??? There are not 200 trend followers on the planet ...

Anyway I digress your message states that you think markets will always trend ... how therefore can trend followers not make money if they have a simple robust trend following system ? Looking at the results of Henry and Dunn over the last 5 years doesnt give too many answers ... you and I do not know what goes on in their organisations .. the majority of tf funds have made small fortunes in the last 5 years ?
The rough period for trend following is exemplified all over the place by funds like Tulip trend fund $100,000 with them in 1994 is today worth $8.7 Million ? There are many of these around and many use simplicity

Two points I would add are this

1. The majority of poorer performing TF's have inevitablly bowed to the cravings of the investor community for lower draw downs and reduced volatility
2. As I said in my first point I have been using a simple TF system for 4.6 Years .. have annualised Returns of 26 % (Net) , max draw dwon of around 20% and last year the simple Donchonian system made me 48 % Net ?

I dont think the investor community will ever really get it !!!

Algonquin
Roundtable Knight
Roundtable Knight
Posts: 158
Joined: Sun Jan 28, 2007 9:38 pm
Location: Greenwich, CT

Post by Algonquin » Sat Jul 11, 2009 12:32 pm

Chris67,

I appreciate what you are trying to say. Compared to the hedge fund industry as a whole, it is true that trend following funds are a relatively small asset class. But it is beyond dispute that on a percentage basis the growth in managed futures over the past 20 years has been dramatic. And the majority of managed futures funds are trend following in nature.

I applaud you on your performance results over the past five years. What you accomplished is no easy feat. All I am saying is that my research results show a continual decline in risk adjusted returns for the Donchian system, as well as a number of well-known systems. Not for all parameter sets, mind you. There are a handful of parameter sets that seem to have weathered the increased volatility of recent years. You, for one, seem to have been prescient enough to choose a set of parameters that profited. Nevertheless, based on my studies it does appear that the Donchian system is trending toward reduced profitability and increased instability. If your research shows something different please let me know.

At any rate, my main point in this dialogue is that for the TB user there are other systems to be discovered that are not evidencing the same decline. In short, there are better ways to profit from today's trends that can be discovered with a little creativity and hard work.

Algonquin

AFJ Garner
Roundtable Knight
Roundtable Knight
Posts: 2040
Joined: Fri Apr 25, 2003 3:33 pm
Location: London
Contact:

Post by AFJ Garner » Sat Jul 11, 2009 2:05 pm

8)
Last edited by AFJ Garner on Mon Jul 20, 2009 7:22 am, edited 1 time in total.

MichaelCovel
Full Member
Full Member
Posts: 11
Joined: Fri Mar 28, 2008 1:03 pm
Location: Henderson, Nevada
Contact:

Post by MichaelCovel » Thu Jul 16, 2009 2:31 pm

AFJ Garner wrote: I can only assume that complacency is what led to a few very bad years for JWH and Dunn. Judging by the way AUM evaporated, both firms looked to be very close to closing their doors.
I don't think that is a wise assumption. Dunn's performance for 2008 was outstanding (+51%). Same with Henry. A recent article discussed Merrill Lynch's decision to pull assets from Henry at the bottom of a DD. Assets leaving doesn't always mean there is a problem with the trader or the approach. Dunn told me recently that for his signals he is still using his approach set forth in 1974.

alp
Roundtable Knight
Roundtable Knight
Posts: 309
Joined: Mon Aug 27, 2007 8:09 pm

Post by alp » Thu Jul 16, 2009 8:29 pm

Algonquin wrote:As for better systems to be discovered, they are out there. There are long term trend following systems whose performance did not degrade over the past several years. While I cannot go into the specifics of those systems...
With the due respect, I don't see any evidence in your posts that you have discovered some of these "better systems... out there". If all you want is belief and approval you can likely give them yourself.

techtrade89
Senior Member
Senior Member
Posts: 28
Joined: Tue Dec 30, 2008 5:36 pm

Better systems

Post by techtrade89 » Thu Jul 16, 2009 9:37 pm

A few questions:

1. Algonquin says: "At any rate, my main point in this dialogue is that for the TB user there are other systems to be discovered that are not evidencing the same decline. In short, there are better ways to profit from today's trends that can be discovered with a little creativity and hard work." My question is: Are you saying that these systems can be discovered using the TB software but they are just not the same as the prepackaged systems/parameters that come with the software? Or are you suggesting a preference for an approach such as writing your own source code? My current approach is centered on custom source coding for relatively short term trading (1-3 weeks holding period on average), but perhaps the TB software provides advantages regarding discovery of these other systems in the long term context?

2. Algonquin says: "Other, lesser-known strategies to which I have referred do not even come close to suffering the same levels of degradation." In other words, the statement is that these strategies are not proprietary/confidential to a small group at a private fund but simply "lesser-known." Is this the case?

3. There seems to be, quite reasonably, a great deal of emphasis on maximum drawdown. However, my investigations into trend-type systems suggest that long periods of relative stagnation, even if not technically in drawdown, might be more of a challenge. Have others encountered this?

Thank you all in advance!

Post Reply