Page 1 of 1

Which futures markets trend the most?

Posted: Thu Sep 02, 2004 8:46 am
by jankiraly
If you think "market selection" is an important component of successful trend trading, maybe this 3-part article on the Omega List will stimulate your thinking. It has forced me to test my systems on some commodities that I've never tried before.
Part 1
Part 2
Part 3

It was reassuring to notice the markets I'd chosen using my own ad-hoc methods, were generally in the top 1/3rd of the chart. Not all of them but mostly. However it was alarming that several of the humongous trenders were commodities I'd never heard of, or (worse!) ones I'd known about but decided not to even try out, due to some feeble excuse or another.

Posted: Thu Sep 02, 2004 10:13 pm
by ksberg
Interesting read. Even Mark comments on some of the unexpected finds. However, it left me wondering if the average ADX ranking method was an appropriate measure of trendiness. For instance, I think the comments on S&P are illuminating since it appears slightly below average, yet is typically and consistently unsuitable for LTTF systems [e.g. try PGO, Thirteen, Turtle, etc on S&P over almost any period and see how it underperforms nearly all other markets].

Mark makes a comment that this was a first-pass attempt to understand historical trendiness, and that some elements are missing. I believe one missing element is volatility. A trend may be present, but it may darn hard to mount and ride the sucker.

With computing power easy to harness, why not combine a batch of LTTF systems and the their resulting performance into comparative measures? This may not be as "pure" as using an indicator like the ADX, but it's very pragmatic ... the result is meaningful in the context of actual trading.

Cheers,

Kevin

Posted: Thu Sep 02, 2004 11:24 pm
by TC
ksberg wrote: I think the comments on S&P are illuminating since it appears slightly below average, yet is typically and consistently unsuitable for LTTF systems [e.g. try PGO, Thirteen, Turtle, etc on S&P over almost any period and see how it underperforms nearly all other markets
.

Agreed that LTTF results on stock indices appear unsuitable when comparing these markets in isolation, however there true value becomes apparent when added to a diversified portfolio of markets. Trendiness of a market is therefore less important than its effects on the portfolio's performance of which it is part.

ksberg wrote: ... the result is meaningful in the context of actual trading.
and especially meaningful for multiple market traders. The low correlation of stock indexes with many of the most commonly traded futures markets is what can make a stock index a valuable part of a futures trader's portfolio.

Better to include a less trendy stock index that improves you portfolio performance than a better trending market that, due to its high correlation with other markets, may in fact be detrimental to overall performance.

Posted: Fri Sep 03, 2004 1:32 am
by ksberg
TC, undoubtedly correlation plays a big part in market selection and portfolio composition. MPT says as much. As you said "Trendiness of a market is therefore less important than its effects on the portfolio's performance of which it is part." ... Of course, we get the effects by looking at a system's outputs (trades), a time-series of profits and losses, not the underlying market as is the case with ADX.

That is not to say trend is not important. On the whole, one cannot take a collection of poorly performing markets and make a winning portfolio. On the other hand, it is often true that adding a lesser performing market into the mix can raise the entire portfolio performance if and only if the correlation serves the portfolio. It is a strange alchemy to be sure.

Kevin

Posted: Fri Sep 03, 2004 3:10 am
by Forum Mgmnt
ksberg wrote:On the other hand, it is often true that adding a lesser performing market into the mix can raise the entire portfolio performance if and only if the correlation serves the portfolio. It is a strange alchemy to be sure.
Ahh, now this is one of the most important truths for the advanced trading student.

Another, even less intuitive truth (and one the better of my fellow fund managers might wish I/we didn't discuss) is that a portfolio of systems that have been opimized to have the best performance characteristics in isolation as individual systems, most often does not perform as well as a portfolio of systems that have been optimized to maximize their noncorrelation with each other.

Analogously, adding a noncorrelated but relatively poor performing system to a mix of systems, can have a surprising effect on the risk-adjusted return of the entire portfolio of systems.

Perhaps even more surprising, sometimes adding systems that lose money can help, if they happen to make money when the other systems don't, and lose money when the other systems are making money.

It can be a bit like insurance, sometimes paying a little in fees to avoid the catastrophic losses makes sense; as long as those fees are not too high, and the amount recovered in times of disaster is sufficient to reduce the overall risk exposure. Just as the reduced risk afforded by insurance might let one take on greater overall commitments in a business settings which in turn might generate more than enough additional revenue to offset the costs; the reduced overall risk afforded by a noncorrelated system can sometimes allow one to trade the other systems hotter and make enough money to offset a negative expectation system.

Strange Alchemy indeed.

- Forum Mgmnt

Posted: Fri Sep 03, 2004 8:29 am
by Tim Arnold
Hi Forum Mgmnt,

This makes alot of sense to me, and I have included instruments in my portfolio that lose money over the test period, but raise my MAR by lowering the max drawdown.

I like the concept you raise about "optimizing the non-correlation with each other." I could imagine exporting trade files to excel and testing the correlation of the p&l, but my question is whether VT could give us some insight into the profit loss characteristic correlation of the instruments we are trading.

I'm thinking of testing several systems against the same basket of futures, and measuring the correlation of each system's performance against each others. Then I could bring together optimally non-correlated systems. Is this something VT can help analyze?

Thanks,

Tim

Posted: Fri Sep 03, 2004 8:06 pm
by Forum Mgmnt
Tim,

Yes, we'll eventually do quite a bit in this area.

We've still got quite a few basics to cover first, but I think there is much fruit to be found here.

- Forum Mgmnt

Posted: Sat Sep 04, 2004 3:44 pm
by garryboor
I think the S&P has not earned its bad reputation among trend following traders. With 0% T-Bill interest, $20 commission, 10% slippage, $50k starting account, and a 1.5ATR stop in Veritrader, I found some trends.

The trend is my friend

Posted: Sun Sep 05, 2004 2:38 am
by ksberg
garryboor wrote:I think the S&P has not earned its bad reputation among trend following traders. With 0% T-Bill interest, $20 commission, 10% slippage, $50k starting account, and a 1.5ATR stop in Veritrader, I found some trends.
Well, it could be that I'm living in a different universe. Maybe it's time I checked!

I do note that very few system vendors choose to include S&P in LTTF systems as followed by Futures Truth. I believe it's fair to say that most vendors addressing the S&P have systems specially designed to trade it, or the indices. Out of George's Top 10 list, two systems actually do include the S&P (Aberration and Basis II), of those two, Basis-II reports negative individual SP performance (a testiment to the earlier correlation comments).

In the broader list of tracked systems there are 170 that can even marginally be called portfolio-based. Only 11 of these include the S&P (be aware that variants like Aberration 1&2 get counted twice). That means at best, a total of 6% of portfolio system vendors choose to include the S&P in their systems. Looking over the 11 systems, I see several which are swing-trade vs LTTF systems. Unfortunately, the list doesn't provide an easy way to check day trade, swing, or LTTF (although some names make this evident).

Lars Kestner published an systematic overview of many common trading techniques in Quantitative Trading Strategies, including channel breakout, dual MA cross, momentum, volatility breakout, stochastics, and MACD. Of the LTTF strategies tested, SP was negative on CBO, momentum, volatility breakout, and MACD. It faired much better on dual MA cross, stochastics, and RSI. I already mentioned S&P performance with published LTTF systems such as Turtle, PGO, and Thirteen.

I see that I could have said things differently, so I'll eat my prior words ... I don't inherently think the S&P has a bad reputation. However, with over 90% tracked vendors reporting, I believe it's fair to say most LTTF system portfolios do not include the S&P. Based on the systems mentioned, I also believe it's fair to say many LTTF systems do not perform well on the S&P.

It's not a matter of finding trends, or something that will work on the S&P. There are systems that do real well on the S&P. In fact, I get pretty close to your equity curve using a shorter moving average system with trailing stops ... but it's not a LTTF (BTW: Garry, you didn't tell us which system and settings you used that would make it an LTTF).

Now, what makes sense to me is to mix systems designed for indices/equities with separate systems designed for commodities/futures. I know some people won't agree, but I believe the mechanics are different between the two markets. One place this shows up is options mathematics. IMO, Garry's S&P system would probably make a nice portfolio addition along side something like Turtle.

Cheers,

Kevin

Posted: Sun Sep 05, 2004 10:52 am
by TC
It's probably not coincidental that trendiness and volume are inversely related. My LTTFS performs much better on the MD,RL & DJ than the higher volume ND & SP (although it does make money in all 5 markets).

Unless you are daytrading and need the tighter spreads and faster fills high liquidity provides I fail to see the attraction of the SP. For a trend follower I think there is easier money to be made with the RL than the SP, IMHO !! For my system, trading a combination of DJ, ND & RL has given better results than any index traded alone.

Would like to hear from anyone with experience to the contrary.

The LTTFS I developed was optimised on a basket of 40 futures markets including the 5 stock indices. Interestingly, the parameters optimised on the futures markets were equally profitable when trading individual equities.

Even more interesting was the difficulty I had in improving performance by reoptimising on random portfolios of equities. My takeaway from this was that as time frame lengthens the quirks specific to individual markets diminish in significance.

Perhaps the only two constants across all the futures (40) & equities (600) tested were the system rules and trader psychology.

Tom

ADX ranks again ...

Posted: Mon Sep 06, 2004 12:03 am
by ksberg
TC, agreed ... no need to get fixated on the SP when there are other opportunities. Your comment on trend length agrees with my casual observation. BTW: returning to Mark's study, here's how the indices you mention stack up, and their current trade volumes. I hilited top (green) and bottom (red) ranked symbols. It's interesting to note how and where the rankings shift.

Cheers,

Kevin

Posted: Mon Sep 13, 2004 9:06 pm
by Roscoe
The concept of trading trend-following systems in markets that show a tendency to trend makes instinctive sense on the face of it, however the key will be to identify/define that tendency.

The attachment shows the result of testing 63 markets over the same 10 year period with three different contenders for a trend-identifying statistic: ADX, RSI and a standard deviation formula published by Gary Fritz in 1999.

I started by replicating the work of Mark Johnson where he looks at using the ADX value for each market, with a high ADX indicating a strongly-trending market and a low ADX indicating a weakly-trending market and I find that I get very similar results to Mark by averaging the ADX over the test period, however I am not sure that this value is necessarily what I want to see.

As a related variation I tested to see how much time each market spends above a specified ADX threshold, so that we can see (in percentage terms) the time spent above each threshold value. I have used an 18-bar ADX and a 36-bar ADX for this short test. The spreadsheet shows the percentage of time the ADX for each market has spent above 10, 20, 30, 40 and 50.

The same three tests have also been performed using an RSI in place of the ADX.

The last test uses Gary’s idea that:
In a pure Random Walk, the N^2-length sample will have a StdDev of N times the StdDev of the 1-length sample. If a particular price example shows larger-than-expected StdDevs it indicates a trending market; if the StdDevs are smaller than the expected values it indicates an anti-trending (reversing) market.
The test shows the results of the difference between the expected value and the actual value. This indicates:
Trending: from my reading of the notes I believe that the difference indicates trendiness regardless of market direction, similar in interpretation to ADX where > = good, so the higher the positive difference the greater the tendency to trend and the less the tendency to revert or cycle.
Reverting: once again from the notes I take reverting to mean a market that does not show a tendency to trend but instead tends to revert or cycle (which could have other applications), so the greater the negative difference the less tendency to trend and the greater the tendency to revert or cycle.

Based on those two points a sideways market would have a difference of close to zero.

The attached Excel2000 file shows the results of these tests displayed graphically. There is a lot of information on these charts (63 bars on each) but by hovering your mouse over a line you can see which market it relates to. (User note: the data has been arranged using pivot tables and the data sheets are hidden for tidiness – Format | sheets | unhide will allow you to see what is behind the scenes if you wish.) You can cut down on the clutter if you want by using the Excel pivot charts feature that allows you to select only the markets of interest by clicking on the “Marketsâ€

Which Futures Markets Trend Most

Posted: Tue Sep 14, 2004 10:57 am
by AFJ Garner
A related subject hinted at above is that many of even the "worst" markets can be made profitable by trading a much longer term trend following system. There was discussion on the TR List a few weeks ago and some interesting posts made by Mark Johnson and others who tried out a simple two MA crossover system with MAs ranging from 50 to 400 days on a range of 70 markets with favourable results even for markets considered "crappy". But there are of course downsides including the amount of capital reqired, the equity swings etc etc

Markets & Trends

Posted: Wed Sep 15, 2004 7:04 am
by SysTrader
@ AFG Garner

Where can I find the TR List?

@ garryboor

Which system did you use in VeriTrader? Have you also tested your parameters for robustness?


In Trend following we basically dont know wich market(s) will perform well in the coming year. Of course your backtesting will tell you that the interest rate markets are very good for trendfollowing. But adding additional uncorrelated markets will reduce your drawdown, increase your trading frequency and stabilise your portfolio. Even if this market(s) is just breakeven over the long term. On the other side, we all know correlations vary - e.g. Metals will not always absorb a losing streak in interest rates. Diversification in markets, systems and also in timeframes is the key.

Posted: Wed Sep 15, 2004 7:37 am
by AFJ Garner
SysTrader
I am afraid the TR list is only available to those who own the Trading Recipes testing software. Mark Johnson and others tested big portfolios, including some notoriously trendless markets, on very long term trend following systems and were pleasantly surprised by the results. One could just as easily plug these difficult markets into Veritrader.

Posted: Wed Sep 15, 2004 9:27 am
by TC
I have tested extensively with time frame and concur with AFJ that the longer the period you trade the more profitable each market becomes in general terms. MY LTTFS trades weekly bars as I typically hold positions for many months.

As we all know, there is no free lunch, so LTTF naturally has its drawbacks. Chief among these is that wide stops must be used to allow trends to develop. Tighter stops seriously degrade performance.

Since large drawdowns are an inherent characteristic of LTTFS they have to be moderated by countertrending moves in other markets.

Diversification is the key to maintaining an acceptable risk:return ratio with LTTF. It was eithe Bill Sharpe or Harry Markowitz who said that diversification is the closest thing to a free lunch in the markets. This is amply demonstrated in LTTFS.

This brings me to a point I made in another thread; you need to be well-capitalized to trade futures because diversification is an effective and, in my opinion, essential strategy for managing risk and smoothing the equity curve.

Posted: Mon Jan 16, 2006 5:45 am
by rabidric
a bit of a side topic, but relevant to the post further up the thread:

i place trades with a volatility sizing method where big vol-->smaller size.
I analysed my historic trades, and wondered how each particular "tranche" of sizing contributed(i.e. 50-100%avg size, 100-200%avg size and 200%+).

to my surprise the net contribution of all trades at under average size was infact quite negative....so i thought it would be better if i just left these trades alone....

but i found that if i excluded these trades, whilst net profit overall was raised, drawdowns were considerably worsened, as were a bunch of important performance yardstick ratios.

to sum up, this group of trades whilst overall, detrimental to profits, actually had a greater tendency to deliver much needed winners during deeper drawdowns, hence lessening them.

In a way it was a bit like having an uncorrelated product with a net negative profit contribution.

hope that is of interest to some.

Ric

Posted: Tue Feb 14, 2006 4:50 am
by AFJ Garner
Agreed but there are issues of size and granularity. The Dax is a huge contract and if you want to take advantage of profit taking on a retail as opposed to insitutional size portfolio in actual trading, you need to take a look at the smaller contracts. The mini Hang Seng for instance.

Posted: Sat Jan 06, 2007 4:46 pm
by droskill
Came across this post and it's something I've been thinking a lot about.

I also came across a company called Trading System Lab - they are founded by Mike Barna and focused on the area of Genetic Programming. Now, despite what the demos show you, their software is not for sale - but I'm also not that interested in Genetic Programming. What I was interested in, however, were some of his methods for determining whether a market is a trending or non-trending market.

On this page: http://www.tradingsystemlab.com/flashdemo.aspx

Take a look at demo #1 (first one on the page). At the end of the demo (after he brings up a Tradestation screen), he discusses using three indicators for determining trend:

- Discrete Fourier Transform (DFT)
- Serial Correlation - look at tendency of the market to have stable trend by looking at 1 and 2 days back and then at the current day to see if there was price persistence.
- Random Trend - based on a random walk analysis.

Would love to know what people think of this - I know this topic is almost a year old but would love feedback.

Thanks,

/d/