Mean reversion system

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.
TrendsCatcher
Roundtable Fellow
Roundtable Fellow
Posts: 97
Joined: Thu Jan 10, 2008 3:28 pm
Location: Southern California, USA

Post by TrendsCatcher »

There is no doubt regarding the value of adding more markets to select trades, because you will increase sampling size, N. According to the law of large numbers, as N -> infinity, sample mean -> population mean. In other words, assume that one has a positive expectancy system, increasing the number of eligible trades, by the means of selecting more tradable vehicles, thus potentially increase the number of legitimate trades, will make one more likely to realize the theoretical expectancy.

That's exactly my point, the more markets one trades, the less impact one big trend will have (miss or not), because with a big sample size, plus or minus one (or 10) won't impact the sample size as much as in a much smaller sample: 1 to 1000 is less impact than 1 to 10.

Once one can afford to miss a big trend, one can be more selective by adding in fundamentals, discretionary decisions, etc

So, I assume that if on trades only 1 dozen or so major futures markets, missing a big trend is devastating;
If one trades 100, or even 300 futures market, missing a big trend is not desirable, but not devastating;
If one trades 10,000 stocks (the universe of stocks in the US), missing one, two, or a dozen big trends is no big deal, especially as a trade off, one has also avoided a lot of whipsaws, or sitting with no-movers.
If on trades 100,000 stocks (the universe of stocks in the world? never done it, just a pure guess), plus the 300 world futures markets, plus the kitchen sinks, plus real estate, gold coins, and host of other stuff, missing one, two, or a dozen big trends is imperative to avoid system overload

You know what I mean.
sluggo
Roundtable Knight
Roundtable Knight
Posts: 2987
Joined: Fri Jun 11, 2004 2:50 pm

Post by sluggo »

There's another benefit to adding more and more equity curves (the outputs of running a trading system on more and more markets), when the underlying trading strategies are profitable and the equity curves are not perfectly correlated: adding them together (so the portfolio return becomes the average of a large number of individual asset returns) increases the risk-adjusted return. Boosts the Sharpe Ratio and other measures of gain vs pain.

Harry Markowitz was awarded the Nobel Prize in Economics for his work that explained this phenomenon, in the context of mean and variance of portfolio returns. His original book on the topic is quite readable (link), and just as relevant today as it was when originally released. I've read it and I apply its ideas to my futures trading, every day. I strongly recommend reading the book.

Markowitz called the phenomenon "Variance Reduction" but later writers, especially those targeting a popular audience, just call it "diversification".
TrendsCatcher
Roundtable Fellow
Roundtable Fellow
Posts: 97
Joined: Thu Jan 10, 2008 3:28 pm
Location: Southern California, USA

Post by TrendsCatcher »

Haven't read that book, but fully agree with what you summarized from the book.

So, folks, are all (or, most) major trends driven by positve feedback mechanisms?
alp
Roundtable Knight
Roundtable Knight
Posts: 309
Joined: Mon Aug 27, 2007 8:09 pm

Post by alp »

TrendsCatcher wrote:So, folks, are all (or, most) major trends driven by positve feedback mechanisms?
What is a positive feedback mechanism?
TrendsCatcher
Roundtable Fellow
Roundtable Fellow
Posts: 97
Joined: Thu Jan 10, 2008 3:28 pm
Location: Southern California, USA

Post by TrendsCatcher »

A pretty good defintition is found here:

http://en.wikipedia.org/wiki/Positive_feedback

According to Seykota, positive feedback mechanisms are the driving forces behind trends, while negative feedback mechanisms cause mean reversion. I tend to agree with him. But I am always curious of the "nature" of the trend, because it is the trend that I would like to catch.
alp
Roundtable Knight
Roundtable Knight
Posts: 309
Joined: Mon Aug 27, 2007 8:09 pm

Post by alp »

TrendsCatcher wrote:But I am always curious of the "nature" of the trend, because it is the trend that I would like to catch.
Perhaps you might also consider these Seykota's statements:
Models are mostly useful in understanding specific and simple behavior modes such as the oscillation of a pendulum, the growth of lilies in a pond and the decay of a radioactive isotope.

The Everything Model to which you refer already exists. It is the universe itself, and it generates its own solution real-time. All you have to do is plug in to it to see what's happening. Trend Traders know this and plug into the only model that actually works.

Mental Models and Computer Models can explain modes of behavior, they are pretty much no good for precise prediction:

1. If you go and build a predictive model of the universe, and if your model generates some information that changes the universe in any way, you then have to keep including your own model in itself; it can all become rather complex.

2. The Heisenberg Principle states you can't know both the exact position and the exact momentum of any particle - so there's a limit to your model when you go to initialize it.

3. The double slit experiment indicates that particles are somehow responsive to our level of awareness of them.
Causality is not a property of a person, place or thing. It is a property of the way you think about events that associate with these objects.

Causes and Effects are typically events you happen to notice occurring at different phases of a cycle.

As such, the "cause" event does not have informational linkages with the "effect" event.

In the System Dynamic Model, you register the interconnections of accumulations and information feedback between objects.

Exercise #1: Build the pendulum model, as others are doing in this column. You might notice: (A) structural elements communicate only in the moment of now; (B) behavioral phases do not communicate with each other; (C) you cannot account for the oscillation in terms of what is "causing" it.

You can account for the oscillation by considering the interactions of all the system elements as they evolve simultaneously in the ever evolving moment of now.

Exercise #2: You might consider asking a dozen people for the "cause" of Seattle; you are likely to receive quite different responses, typically including, "I don't know what you are asking."

Please perform exercise #2 and email me the responses.
TrendsCatcher
Roundtable Fellow
Roundtable Fellow
Posts: 97
Joined: Thu Jan 10, 2008 3:28 pm
Location: Southern California, USA

Post by TrendsCatcher »

Very true.

Had I asked Seykota the same question, he would have probably replied with the following phrase:

You might consider taking your feelings about "trying to understand", "cause" to Tribe.

Oh, by the way, he is working on another book.
AFJ Garner
Roundtable Knight
Roundtable Knight
Posts: 2071
Joined: Fri Apr 25, 2003 3:33 pm
Location: London
Contact:

Post by AFJ Garner »

To talk to in person, one to one, ES is very interesting, bright and helpful.

When I met him there was no smoke, no mirrors, no riddle. A straight exchange of views on trading which was very useful.

Personally, I find most of the statements on Ed's FAQ profoundly unhelpful. I accept of course that there are many out there for whom such comments have greater value.
TrendsCatcher
Roundtable Fellow
Roundtable Fellow
Posts: 97
Joined: Thu Jan 10, 2008 3:28 pm
Location: Southern California, USA

Post by TrendsCatcher »

AFJ Garner,

Thanks for the information on Seykota. He is one of my favorite "Market Wizards". I feel I learned a lot just from his Market Wizards interview. I Plan to meet him, maybe the next year, by attending his seminiar. Did you meet him through that route?
alp
Roundtable Knight
Roundtable Knight
Posts: 309
Joined: Mon Aug 27, 2007 8:09 pm

Post by alp »

TrendsCatcher wrote:Very true.

Had I asked Seykota the same question, he would have probably replied with the following phrase:

You might consider taking your feelings about "trying to understand", "cause" to Tribe.

Oh, by the way, he is working on another book.
Yes. He rather follows the Socratic method and the dialogue is focused on feelings.
LeviF
Roundtable Knight
Roundtable Knight
Posts: 1436
Joined: Mon Dec 22, 2003 12:24 pm
Location: Des Moines, IA
Contact:

Post by LeviF »

AFJ Garner wrote:To talk to in person, one to one, ES is very interesting, bright and helpful.

When I met him there was no smoke, no mirrors, no riddle. A straight exchange of views on trading which was very useful.
Did Ed say anything particularly enlightening?
AFJ Garner
Roundtable Knight
Roundtable Knight
Posts: 2071
Joined: Fri Apr 25, 2003 3:33 pm
Location: London
Contact:

Post by AFJ Garner »

levijean wrote: Did Ed say anything particularly enlightening?
Not in the spiritual sense! We talked a lot about his research on mechanical stock trading and his methods of making room for new trades in the "parking lot". That was illuminating and counter-intuitive. Getting rid of the oldest positions to make room for new ones was one idea which looked promising and interesting. Old positions tended to be the profitable ones which eventually simply ran out of steam as earnings growth inevitably plateaued.

At that stage he reckoned it would make a lot of sense to launch a systems traded stock fund. We met at Cambridge in the UK and talked a little of Newton. Ed was holding a seminar down there which I did not attend.

I found him interesting, intelligent and likeable. And almost as offbeat as I am.
stopsareforwimps
Roundtable Knight
Roundtable Knight
Posts: 199
Joined: Sun Oct 10, 2010 1:47 am
Location: Melbourne Australia

Post by stopsareforwimps »

AFJ Garner wrote:...We talked a lot about his research on mechanical stock trading and his methods of making room for new trades in the "parking lot". That was illuminating and counter-intuitive. Getting rid of the oldest positions to make room for new ones was one idea which looked promising and interesting. Old positions tended to be the profitable ones which eventually simply ran out of steam as earnings growth inevitably plateaued...
I am currently grappling with this issue. There is not a lot in the proverbial 200 books about it.

FWIW here are my thoughts and the little I have been able to glean so far aka things I could try.
  • If a trade can't fit with the envisioned size, go in smaller.

    Make room for new trades by trimming the size of current trades, or by getting out of current trades in some sort of priority order.

    Priority could be based on how old the trade is, whether the trade is now over-sized, or the performance of the trade overall or 'lately', or correlations.

    Possibly if room opens up one could reinstate those bumped trades again according to some scheme.

    One could even start trades that were initially rejected if room opens up for them.
The trouble is my "mixer" code is the worst spaghetti of all my code, and I just rewrote it. I don't want to make it too much more complicated. The more parameters you have the more corner cases and the greater the chance of bugs.

One trader I respect has the approach that he has no overall exposure limits. If a trade is triggered he takes it. The risk management is at the level of each trade. This fits with the observation that good trades come in bunches.

Jim Rogers said he would trim other trades to make room, when he was with Soros.

If anyone could point me to any threads, links or books with good coverage of this topic I would appreciate it. Failing that I will code it up and report back. I have a method to avoid over-fitting so the extra parameters don't bother me.
Toosday
Roundtable Fellow
Roundtable Fellow
Posts: 70
Joined: Fri Jun 03, 2011 11:21 am
Location: Austin, TX

Post by Toosday »

I have encountered the same problems with complexity and coding stock systems. The way I look at it is that is what I get paid for, to do the dirty work. It is frustrating and sometimes I take a few weeks away.

The only thing I can say is that based on my work, the work you propose is cumbersome and complex but if it were easy to do in a spreadsheet then the returns would be less.

For me personally I have went through many iterations of data structures and logic frameworks. I have found this to be much more difficult than finding an edge.

Send me a PM if you would like to talk more.
stopsareforwimps
Roundtable Knight
Roundtable Knight
Posts: 199
Joined: Sun Oct 10, 2010 1:47 am
Location: Melbourne Australia

Post by stopsareforwimps »

Toosday wrote:I have encountered the same problems with complexity and coding stock systems. The way I look at it is that is what I get paid for, to do the dirty work. It is frustrating and sometimes I take a few weeks away.
Thanks toosday. I don't seem to be able to PM for some reason.

I think most people take one of two approaches but I would be interested to know for sure.

1. Take all the trades that are signaled, with the size of the individual trades being the primary method of risk control.

2. If the portfolio is too full when the trade is signaled, skip the trade.

As you suggested this issue of "congestion" is a huge issue for stock trading.

For now I am going to code up some other parts of my system and come back to this later.
Anthony Marsland
Roundtable Fellow
Roundtable Fellow
Posts: 83
Joined: Thu Jan 13, 2011 5:33 am
Location: London
Contact:

NAAIM Winner

Post by Anthony Marsland »

This study by Thomas Krawinkel may be of help. He has created an Excel sheet to help study the effects of missing trades and will share it with anyone committed to a thorough investigation of their system.

He can be contacted at TKrawinkel at gmx.de
Attachments
NAAIM-Buying Power.pdf
(609.26 KiB) Downloaded 528 times
Post Reply