Query to the System Design Community

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

Post by LeviF » Fri Feb 27, 2009 8:22 am

I have not tried this because I trade forex, but I wonder what a max-liquidity filter would do? If the most liquid instruments are more efficient and less trendy, it could keep you out of some bad trades...

Paul King
Roundtable Knight
Roundtable Knight
Posts: 207
Joined: Mon Feb 23, 2004 9:13 am
Location: Vermont, USA
Contact:

Post by Paul King » Fri Feb 27, 2009 8:43 am

"If the most liquid instruments are more efficient and less trendy"

I would test this assumption - liquidity, efficiency, and trend are not directly related as far as cause-effect goes (in my experience).

More liquid instruments have smaller spread, less noise, and therefore generally lower implementation costs and more orderly continuous pricing; nothing to do with "efficiency" or "trend".

More volatile (relative to price) instruments are more "efficient" since they can implement the same amount of risk with smaller size. Again, not directly to do with trend.

Certain instruments that are based on the "real world", rather than purely currency or financial related tend to be more "trendy". Again, not directly related to liquidity or efficiency.

sluggo
Roundtable Knight
Roundtable Knight
Posts: 2986
Joined: Fri Jun 11, 2004 2:50 pm

Post by sluggo » Fri Feb 27, 2009 9:49 am

Here is one way a systems researcher might explore the question of which markets are, in general, the friendliest to trendfollowing systems.

Choose ten or twenty very different trend-following mechanical trading systems. (not just the same single system with twenty different parameter settings). Select some systems that are purely based on indicators (like "RSI Trend Catcher", like MACD, like n-Moving-Average-Crossover, Parabolic SAR), select some that are based on breakouts, select some that are based on volatility envelopes (like Bollinger Bands, Keltner Channels), select some that are based on momentum oscillators (CMO, PGO, Blox's ADX system, etc).

Run each system on the same large (at least 50 futures markets, but preferably, 150 futures markets) portfolio, over a decent sized span of recent history. Like 1/1/1999 to 1/1/2009.

For each system test, capture the reward/risk performance of each market in the portfolio. I happen to like Blox's "Percent Profit Factor" found in the Instrument Performance Summary, but anything else that provides a normalized gain/pain PER MARKET would do. Then rank the markets by reward/risk: #1 is the highest reward/risk (greatest Percent Profit Factor), #150 is the lowest reward/risk.

Finally, for each market, calculate the average ranking of that market across all the systems tested. Add up all 20 rankings for that market on all 20 systems, and divide the sum by 20. You've now got the overall average goodness (or badness) of each market, for variety of trend following systems. In other words, its friendliness to trendfollowing in general.

For those who actually do perform the experiment, I predict you will find some of the results quite surprising. It may induce you to change the composition of your portfolio, or to change the way you think about composing portfolios. And for those who don't, think of the time and money and effort you conserved.

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

Post by alp » Fri Feb 27, 2009 10:33 pm

adamant wrote:Not sure why you appear to be asking me this question.
Well it was not directed to you particularly. But I am here to learn from you.

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

Post by alp » Fri Feb 27, 2009 11:12 pm

sluggo wrote:For those who actually do perform the experiment, I predict you will find some of the results quite surprising. It may induce you to change the composition of your portfolio, or to change the way you think about composing portfolios. And for those who don't, think of the time and money and effort you conserved.
The way you state it, it sounds like the last option (conserve time, money and effort) is a good choice. :) I mean, the most suitable markets for trend following are those which are trending! I sounds pretty obvious but it's not psychologically appealing.

sluggo, I have been browsing some CTA's performance profiles, looking at their equity curves and trying to extract some information from the disclosure papers. It's rather difficult to separate the wheat (useful information) from the chaff (marketing gimmicks or intentional disinformation).

Whatever it seems to me most professionals look to diversify across several markets and parameters looking to smooth as much as possible the equity curve. The common theme of emphasis on robustness across several markets and parameter sensitivity make me wonder if they are actually diversifying systems. I can't think of many options here.

Instead, it looks to me they generally are working with a plethora of markets, staggered parameter sets, different rule sets and filters, looking for diversification and high probability signals. A good example is Clarke Capital: http://www.iasg.com/group/clarke-capita ... obal-basic If the compound RoR is net of fees, that's good performance for what appears to be a fully systematic approach across 42 markets with a minimum account of 50K. What do you think?
Last edited by alp on Fri Feb 27, 2009 11:22 pm, edited 2 times in total.

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

Post by alp » Fri Feb 27, 2009 11:20 pm

Paul King wrote:More liquid instruments have smaller spread, less noise, and therefore generally lower implementation costs and more orderly continuous pricing; nothing to do with "efficiency" or "trend".

More volatile (relative to price) instruments are more "efficient" since they can implement the same amount of risk with smaller size. Again, not directly to do with trend.
Canadian Dollar, Corn and SP are some examples of "liquid" candidates to test some premises about liquidity, efficiency or trendiness.

By the way, I am not sure what you both mean by "efficiency".

rhc
Roundtable Knight
Roundtable Knight
Posts: 464
Joined: Tue Nov 18, 2008 8:46 pm
Location: Oz

Post by rhc » Sat Feb 28, 2009 5:30 am

From Sluggo;
And for those who don't, think of the time and money and effort you conserved.
Classic!!

Insights, discoveries & success, more often than not, tend to accrue to those who are willing to put in the 'hard yards'

Or as Thomas Edison might say;
“Success is 10 percent inspiration and 90 percent perspirationâ€

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

Post by alp » Sat Feb 28, 2009 4:36 pm

rhc wrote:Insights, discoveries & success, more often than not, tend to accrue to those who are willing to put in the 'hard yards'
100% agreed. Some food for thought from a complementary study: viewtopic.php?t=3326 Also, the previous "trendiness" study could be restricted to the actual investable universe of the trading system and does not address the issue of "dynamic portoflio selection" which could be some kind of ideal filter: viewtopic.php?t=2576#20738

nzbryant
Senior Member
Senior Member
Posts: 29
Joined: Mon Dec 13, 2004 10:43 pm

Post by nzbryant » Sun Mar 01, 2009 9:25 pm

sluggo wrote:Here is one way a systems researcher might explore the question of which markets are, in general, the friendliest to trendfollowing systems.

Choose ten or twenty very different trend-following mechanical trading systems. (not just the same single system with twenty different parameter settings). Select some systems that are purely based on indicators (like "RSI Trend Catcher", like MACD, like n-Moving-Average-Crossover, Parabolic SAR), select some that are based on breakouts, select some that are based on volatility envelopes (like Bollinger Bands, Keltner Channels), select some that are based on momentum oscillators (CMO, PGO, Blox's ADX system, etc).

Run each system on the same large (at least 50 futures markets, but preferably, 150 futures markets) portfolio, over a decent sized span of recent history. Like 1/1/1999 to 1/1/2009.

For each system test, capture the reward/risk performance of each market in the portfolio. I happen to like Blox's "Percent Profit Factor" found in the Instrument Performance Summary, but anything else that provides a normalized gain/pain PER MARKET would do. Then rank the markets by reward/risk: #1 is the highest reward/risk (greatest Percent Profit Factor), #150 is the lowest reward/risk.

Finally, for each market, calculate the average ranking of that market across all the systems tested. Add up all 20 rankings for that market on all 20 systems, and divide the sum by 20. You've now got the overall average goodness (or badness) of each market, for variety of trend following systems. In other words, its friendliness to trendfollowing in general.

For those who actually do perform the experiment, I predict you will find some of the results quite surprising. It may induce you to change the composition of your portfolio, or to change the way you think about composing portfolios. And for those who don't, think of the time and money and effort you conserved.
Ditto this. I do the same kind of analysis - some may call it anally retentive. I call it "being a professional and maximising ones contribution to trading."

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

Post by alp » Mon Mar 02, 2009 8:34 pm

Another complementary thread: viewtopic.php?t=5892

Post Reply