Page 1 of 1

Has Trend Following Really Gotten Harder?

Posted: Wed Mar 16, 2005 7:53 pm
by fkienast
I have corn price data (reverse adjusted) going back to 1970. I decided to do some tests on the years 1970-1979 as compared to 1990-1999 to see if trend-following systems worked better in the past than presently. The system I tested was a breakout system with stops. When corn makes a new n-day high (n optimizable), the system goes long, setting a stop at an (optimizable) percent of the current price. The position is reversed when corn makes a new low. Such a system can be optimized so that out-of-sample data gives around an 18% annual profit, with a maximum drawdown (multi-year) of around 50% (not great I know). That is for the 1990's. I tried the same system in the 1970's and was suprised that the parameter values optimized similarly and performance was not a lot better. Granted, I could improve the 1970's performance by building in assumptions based on hindsight of the trend (i.e. giving long positions more favorable treatment than short positions). But I was not able get significantly better performance for the decade or individual years of the 1970's than the 1990s given equal treatment of long and short positions.

Obviously corn is just one market, and only one type of system was tested. However if the markets were a lot different then than now I would have expected to see a difference in just about any trend-following system. I am curious if anyone else has backtested systems on old historical data. Did you notice systems performing better in the past than now?


Posted: Thu Mar 17, 2005 5:47 am
by Old European

I completely agree with your conclusion.

I have been doing numerous tests with multiple moving average cross-over type of systems over almost 30 years of historical data and on a generic market portfolio of about 50 futures markets. Time and time again I was surprised to see that out-of-sample testing didn't lead to deteriorating results. The parameter values (robustly) optimized over say the first 15 years turned out to only slightly different from those optimized over the other 15 years or over the entire 30 years.

Although very many people (including Richard Dennis, see recent message from D. Hoffman on this board) state that the markets are nowadays very different than they were in the 70s or 80s, I haven't seen this anywhere in my testing. Many systems have been performing as well (or even better!) over the last 10 years than over the preceding period. Of course 2004 for example was a tough year, but for long-term trend following we should have at least a 10 year time frame in mind.


Old European

Posted: Thu Mar 17, 2005 9:54 am
by Dean Hoffman
I think a good exercise for a VT owner who is curious about this would be to test a portfolio of say 25 markets over the following periods.

1975-1985 then 1985-1995 and then 1995-2005

Obviously try to choose markets that traded during all periods for an apples-to apples comparison. Compare the various performance ratios and see it there is any noticeable difference in decades (Or five year increments etc.)

Posted: Thu Mar 17, 2005 11:51 am
by kianti
Have hedge funds eroded
market opportunities?
- JP Morgan

Posted: Thu Mar 17, 2005 4:43 pm
by JAM
Kianti, thanks for a good read, but I'm going to suggest that it might not be a good idea to post a proprietary research piece up on a public forum such as this without JP Morgan's permission, given the copyright.

Posted: Thu Mar 17, 2005 5:36 pm
by verec
JAM wrote:[...] but I suggest that it might not be a good idea to post a proprietary research piece up on a public forum such as this without JP Morgan's permission, given the copyright.
As someone working for one of JPM competitors, I can attest that such "research papers" are meant to be read by the broadest possible audience, and are freely available on the premises, to anyone caring to grab a copy.

It is, indeed, the best interest of most investement banks, to promote their "unique" research in order to attract "savy investors".

Posted: Thu Mar 17, 2005 7:02 pm
by fkienast
Maybe corn would be more likely than most commodities to still trend due to position limits? Position limit I think is 5 million bushels which would only be about $10 million. Maybe too small for hedge funds.

I did find that the grains seem to trend the best of any commodities, which is why I am concentrating on them. As per the report, I do indeed notice that it is difficult to design a decent trend-following system using stock indexes. In fact, when I first started testing systems, I only had access to stock data (from Yahoo). So I set up a moving average crossover system, trying values between 2 and 100 days for the two moving averages, and ran it against indexes such as the Dow Jones and Nasdaq indexes. I found that most of the profitable values had the short-term moving average with a bigger number than the long-term moving average. In other words, one would need to go long when the short-term moving average was <i>below</i> the long-term moving average! Even so, profits were neither robust nor consistent. One thing I did find with the stock market that was profitable in testing was to fade opening gaps of volatile stocks (such as technology issues). If a gap was of the "right" size (neither too small nor too large), fading the gap and then closing the position at the end of the day was profitable. This however did not work very well with the indexes (even QQQQ). I imagine more institutional trading (as opposed to simply buying by mutual funds) is done with indexes than with individual stocks. Individual stocks may not be liquid enough for institutional investors, therefore profit opportunities could still remain for smaller individual players.


Posted: Thu Mar 17, 2005 8:15 pm
by Bondtrader
Maybe your universe of stock indexes is too small. I just ran a quick test on a portfolio of 9 stock indexes: STOXX, Russell2K, CAC-40, Nikkei, S&P500, DJ30, Nasdaq100, EuroStoxx, Dax. With Veritrader's long term trend following system "Triple moving average", using parameters (200, 160, 36), I got the equity curve shown here. It isn't earth shattering but it does make a profit. You can see how looooong term are the trends it follows, by calculating round trips per market per year.

Posted: Fri Mar 18, 2005 5:51 am
by AFJ Garner
Although I posted this also under "New Richard Dennis interview" I thought it worth repeating .........for those who like to follow the gurus.

Here is an interesting quote from Ed Seykota dated 24th Feb 2005 (taken from his website) when asked the question whether he is still trading the same systems he traded 20 years ago:

"In 1972, when I start trading, successful investors cite these rules as ones that they remember their predecessors citing:

Trade with the Trend.
Ride Winners
Cut Losers
Manage Risk

I see nothing to change."

Posted: Fri Mar 18, 2005 7:39 am
by JAM
This is off topic, so I will get off it, but in regards to dissemination of proprietary research, different firms (including the one I work for) take a different view towards the dissemination of their product to non-clients or non-prospective clients, as research is a revenue driver (direct and/or indirect). I can't comment on JP's policy, just wanted to throw out a caution.

Posted: Fri Mar 18, 2005 8:45 am
by AFJ Garner
Hmm, I must say I wrote instituational brokerage research for Swiss Bank Corporation way back in the 80's. I have to tell you we were always only too pleased to have that chucked around in every way possible. It was our advertising, our calling card. That was its purpose. My less kind friends used to claim they had spotted large mounds of my remaindered resarch piled up around the dustbins in Central, Hong Kong. Chuckle.