Portfolio Selection

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.
Sir G
Moderator
Moderator
Posts: 243
Joined: Wed Apr 16, 2003 12:21 am
Location: Salt Lake City, Utah

Portfolio Selection

Post by Sir G » Wed Apr 16, 2003 9:47 pm

Portfolio Selection. This might be the mother of all threads!

Why do it? How does one do it? What are the benefits? What are the downsides?

Forum Mgmnt, If I may be so bold and ask you NOT to respond to this post at this time. I would love to see how we as a group can develop an idea without your guidance and direction.

Lets as a group engage in this topic... and many others, and see if we can impress one of the turtles!

I think we can do it.

My quick thought here is that it's always nice to be in the right place at the right time and I believe that is what portfolio selection is about...

But how?

Karakoram
Roundtable Knight
Roundtable Knight
Posts: 127
Joined: Wed Apr 16, 2003 11:01 pm
Location: Reno, NV
Contact:

Post by Karakoram » Thu Apr 17, 2003 1:29 am

I cannot speak for others, I will give you my opinions and beliefs:

I believe market selection is more of a discretionary process.

I believe in diversification in independent markets. Short term traders have a much harder time doing this. I am longer term.

I believe much of portfolio selection is a process of elimination. The total number of markets I can trade is limited by my available capital. Many markets are too thin (not enough liquidity/volume) to trade, unless I want a lot of slippage. Volatility (in dollars per contract per point of movement) is another criteria. If a market is to "big" for me to trade, I will avoid it. For example, the S&P is a lot bigger than corn, same is true for crude vs. oats.

If I have too much capital, some market are too "small" for me. If I were Soros or similar, then I would not trade Mid-Am ( a/c/e ?) mini-contracts, except for arb type trading.

I attempt to determine which markets move together and which do not. Gold vs. wheat is good diversification. Gold vs. silver is not so diversified, however, if I have the availble capital, I may add more markets, even if some are correlated with others. I will attempt to balance the matching markets.

I believe that the more markets I am trading, the more likely I am to be on board trends, in other words, I am more likely to make money more often.

I believe in diversifying across time zones.

I believe in diversifying strategies and time frames.

When diversifying, I remember to "volatility match" each trade (see original turtle rules).

I also believe the stock market is highly correlated with individual stocks. I also believe diversification is possible, so individual stocks are not correlated with the stock market.

Chuck B
Roundtable Knight
Roundtable Knight
Posts: 469
Joined: Thu Apr 17, 2003 6:34 am

Post by Chuck B » Thu Apr 17, 2003 7:21 am

I really don't think there is "an answer" to this question. It is the most perplexing part about trading to me (like all the rest was simple :wink: ).

Everything is based on assumptions that tend to be grounded in historical data. Things like correlation can and do change such as the recent moves in financials coupled with gold and crude for example. But it even gets deeper than that as you can trade two highly correlated markets with the same system and get substantially different results simply due to the market's volatility about its mean price movement in trends (i.e. Bund and Bobl, US and TY, SP and NQ, Dax and EuroStoxx50, etc...).

What's a trader to do....

Karakoram
Roundtable Knight
Roundtable Knight
Posts: 127
Joined: Wed Apr 16, 2003 11:01 pm
Location: Reno, NV
Contact:

Post by Karakoram » Thu Apr 17, 2003 11:27 am

I agree that correlations change between markets. They sometimes correlate, they sometimes do not. That is why I beleive in trading as many markets as my account will allow, even some markets that are correlated now, as well as markets that are not correlated now (but could be in the future).

You never know when a trend will hit a particular market, so it is my job to catch it. That means I have to trade that market all the time, consistently, and lose money most of the time (keep losses small), while I am waiting for the monster move.

Jimmy
Senior Member
Senior Member
Posts: 39
Joined: Wed Apr 16, 2003 11:31 am

Post by Jimmy » Thu Apr 17, 2003 12:35 pm

How is everyone defining correlation? Do you eyeball returns to determine whether stocks/futures/markets are correlated or are you actually calculating the correlation ranging from -1 to +1 using historical variance and standard deviation of returns?

I'm curious because in classical portfolio management, managers would attempt to diversify across many investment vehicles with low correlation (0 to -1) in an attempt to create a portfolio with higher expected returns without increasing risk, thus effectively pushing up their efficient frontier. How does a small time trader without the same $ resources achieve the same sort of diversification in an effort to catch the big trends?

My initial thought is to keep your position size small enough in as many uncorrelated markets as possible and use money management to add to winning positions as trends develop. This requires continuously monitoring the correlations between your positions and markets traded. Is this too simplistic a view? I guess the question then is what is the optimal mix of units to hold given a portfolio $ amount and I think the turtles handled this with their 4 levels of unit limits.

Sir G
Moderator
Moderator
Posts: 243
Joined: Wed Apr 16, 2003 12:21 am
Location: Salt Lake City, Utah

A better way?

Post by Sir G » Thu Apr 17, 2003 12:45 pm

Good Point Jimmy-

I'm trying to get my mind around the whole correlation issue. Is there a benefit to dynamically calculate the correlations, say once a month, looking at just the past 2 months of data? or some such idea?

Usually it appears that I just see a table that is taken on some set of unknown data period and that is proclaimed to be the correlations.

Is there or should there be a better way?

Sir G

Karakoram
Roundtable Knight
Roundtable Knight
Posts: 127
Joined: Wed Apr 16, 2003 11:01 pm
Location: Reno, NV
Contact:

Post by Karakoram » Thu Apr 17, 2003 1:00 pm

I just have some comments.

"in classical portfolio management, managers would attempt to diversify across many investment vehicles with low correlation (0 to -1)"

-1 means perfectly negatively correlated, or when one goes up, the other goes down. I think you mean -1/2 or more to +1/2 or less (a range of correlation from - to + of the same value)

Also, classical portfolio management makes a HUGE assumption: that the correlation holds constant. I personally feel this is a big mistake to assume that. I believe correlations change over time. It might be possible to track correlations, and somehow incorporate that as a timing rule, but I haven't tested this idea.

I just assume that correlations change over time, so I just attempt to trade as many markets as capital will allow.

If you are a small trader, then find a way to become bigger.

THis may mean doing something else for a while. Start a business, or become a CTA and raise the necessary capital.

Or, you could trade stocks and grow your account to the size you need for futures.

I heard that the Turtles started with around $100K each, but that was a notional account equity. They assumed a total equity of around $1million. I have heard that you need at least $100K to $200K to start trading futures, as a trend following, long term, diversified strategy. I would tend to agree with that.

Jimmy
Senior Member
Senior Member
Posts: 39
Joined: Wed Apr 16, 2003 11:31 am

Post by Jimmy » Thu Apr 17, 2003 2:53 pm

Karakoram,

Thanks for the catch as well as the suggestions relative to starting with a small account. It's been a few years since my quant education, so I appreciate the clarifying comments regarding range of correlation. I also agree with your comments about the incorrect assumptions made by PMs in regards to constant correlation. One aspect of portfolio management that may have some relavence to this topic is the correlation analysis PMs(or rather their quant department) perform to determine the effect on the total portfolio when adding a new security.

From this perspective, does it make sense to calculate the correlation of a potential entry in relation to your current holdings to determine diversification effects? Does anyone implement this in their current trading process? If so, are you doing this manually or can you program this into excel or some other software?

I appreciate all of your thoughts as I'm learning and trying to absorb as much as possible from everyone in the forum.

Cheers,
Jimmy

Karakoram
Roundtable Knight
Roundtable Knight
Posts: 127
Joined: Wed Apr 16, 2003 11:01 pm
Location: Reno, NV
Contact:

Post by Karakoram » Thu Apr 17, 2003 3:24 pm

"From this perspective, does it make sense to calculate the correlation of a potential entry in relation to your current holdings to determine diversification effects?"

If I understand you correctly, I think this is a question I would answer in my backtest. Then I can determine the correlation by simple inspection of the simulated trades.

If you do this, you need software that can do a backtest of a portfolio.

You will find that there are not many packages that can do this (along with money management).

Also, I like to consider the effects of your total longs vs. shorts (see original turtle rules).

Kiwi
Roundtable Knight
Roundtable Knight
Posts: 513
Joined: Wed Apr 16, 2003 1:18 am
Location: Nowhere near

Post by Kiwi » Fri Apr 18, 2003 5:45 am

Philosophically I think that the first challenge is to select a number of historically poorly correlated markets that are likely to be poorly correlated in future. I think the second challenge is to accept that they may not be uncorrelated in future and to have a strategy for dealing with this development. I am a strong believer that you cant know what the market will do next, only how you will respond to it.

Ideally each market selected should provide good returns for each commodity traded but this might not be the case. The approach that I apply is that some markets are not profitable with certain types of systems. So I trade them with systems that should fit them well and look at the combination of all systems + contracts traded to give an output. Examples are trend following for currencies and counter trend and short term systems for indexes but you need to be careful not to curve fit systems to the historical behaviour of the markets.

It helps if there is a rationale to explain why the markets you choose wont be correlated in future - this avoids curve fitting and may alert you to changes in future.

In March we saw energies, metals stocks and currencies become highly correlated. This seems to require a tested response to the situation where correlation is driven by an external event like the lead into a war or other significant news (perhaps a worsening of the SARS situation). The most obvious one is to take profit on some of the positions when the risk becomes apparent.

John

Sir G
Moderator
Moderator
Posts: 243
Joined: Wed Apr 16, 2003 12:21 am
Location: Salt Lake City, Utah

Rotating the question

Post by Sir G » Fri Apr 18, 2003 1:26 pm

Let’s rotate the idea of Portfolio Selection a bit. Is correlation the answer? Is Correlation the beginning point or the ending point of it? Or is it better served in the category of position sizing?

I think I understand the logic of correlations, but does/should it apply to inclusion or exclusion of markets? There are three basic actions to a trading system, Long, Short, Flat. For Portfolio Selection, should we be placing importance in the micro flows of the data or should we be looking at the macro flows of data, not only as they relate to each other, but how they relate to the three basic actions of Long, Short, Flat?

I propose we think of the micro data flows within the concept of risk, while we think of the macro flows of data within the concept of opportunity.

Any thoughts, insight, opinions, beliefs?

Sir G

Kiwi
Roundtable Knight
Roundtable Knight
Posts: 513
Joined: Wed Apr 16, 2003 1:18 am
Location: Nowhere near

Post by Kiwi » Fri Apr 18, 2003 8:07 pm

I dont actually understand the question :twisted:

Sir G
Moderator
Moderator
Posts: 243
Joined: Wed Apr 16, 2003 12:21 am
Location: Salt Lake City, Utah

The Brothers Goldie Hawn

Post by Sir G » Fri Apr 18, 2003 11:53 pm

Sorry about that, I just reread it and see how it wouldn’t make sense.

Is the idea of using Correlation in Portfolio Selection the right area of focus?

Correlation is a comparison of the pulse of the market’s closing prices. It’s just like a heartbeat that is compared throughout each market. I call it a micro flow of the data, as it is simply a very narrow look. My heart beats, as does yours, if at some moment in time our hearts exhibit the same rhythm does that mean we look the same, we love the same person, we think the same, we act the same? To me it just means that we have something in common. How can such a narrow view produced by correlation actually impact & enhance the opportunities of our Longs, Shorts & Flats?

I’m thinking the benefits of correlation might come in handy as a way to reduce risk as prices are marked-to-market. As that is an area that is directly impacted by closing price direction. This can gauge the hotspots of risk that ebb & flow together. I assume this can be tended to by reducing position size in those markets. - We should chat about that in and of itself.

A macro flow of data is a view of things that actually impact & enhance the opportunities of our Longs & Shorts. Does it exist? If it does then I propose that should be the guiding light of Portfolio Selection. It should be some sort of a match maker, one that will direct our longs, shorts & flats into markets that offer each other a greater opportunity.

I think this story explains it better.

The brothers, Goldie Hawn.

They were two brothers, Goldie and Hawn. They both were to inherit $1,000,000 from their rich uncle, who was a crusty old time tape reader. The provision in the Will said the brothers had to trade the money starting in 2 months after his death and at that time they could not have any contact with each other for the next 10 years. Goldie & Hawn moved in together for the next 8 weeks and brainstormed day & night on how to trade. Neither had computer skills, so they had to come up with some logic that made sense, they had little luck but at the 11th hour they came up with some logic. They will buy the market at breakouts of 55 day highs and short at 55 day lows. They were running out of time so they quickly came up with an idea, Exiting winners at breakouts of 20 day lows/ highs, they also quickly arrived at initial stops that capped risk at 2% equity. They allowed for scaling into winning positions. The weeks went by and the final day for them to be together arrived. With Hawn now due to go home and his plane leaving in a few short hours, they drove to the airport, patted each other on the back, shook hands and agreed not to speak to each other for the next 10 years… but with a wink, they agreed they would read about each others successes in the business journals. As they said good bye, a thought quickly crossed Hawn’s mind, he asked Goldie as he pulled away, “which markets do we trade?â€
Last edited by Sir G on Sat Apr 19, 2003 3:40 pm, edited 1 time in total.

Dave S.
Senior Member
Senior Member
Posts: 38
Joined: Tue Apr 15, 2003 4:51 pm

Diversification

Post by Dave S. » Sat Apr 19, 2003 7:05 am

William O'Neill of How to Make Money in Stocks, and Investor's Business Daily fame, has been quoted as saying, "Diversification is a hedge for ignorance."

If we assume diversification among intruments and markets to be necessary, some questions that come to mind are:

1) How do you diversify when you have no positions on?
This could be when just starting out, or after liquidation of previously-held postitions.

2) How do you diversify if you can only trade one market?
Suppose you are a pit trader who stands in one pit all day, or work for a commercial interest that trades only the market pertaining to their business?

It has been industry gospel that diversification is absoluteyl necessary for so long that it seems few question this basic premise. I've read very little that would argue with this diversification idea, yet there are circumstances where this is not possible, and people DO trade profitably.

Can diversification help spread risk? Sure, but maybe we should look at how many different ways we can diversify. Surely just diversifying through intruments and/or markets is limiting.

I'm interested in people's thoughts about expanding the notion of diversification in as many creative ways as possible!

Dave

Robert Nio
Senior Member
Senior Member
Posts: 41
Joined: Thu Apr 17, 2003 11:34 am
Location: Palo Alto, California

Post by Robert Nio » Sat Apr 19, 2003 10:07 am

Just my $0.02:

I see DIVERSIFICATION as just another element in good Money Management. There is a statistical "benefit" to diversify your trades among various markets / stocks.

Diversification can help you smooth the waves of your equity curve, but the effectiveness wears off exponentially, thus the biggest effect is seen in the first few markets.

What measure do you use to diversify? This is an art all by itself and there are as many "opinions" out there as there are traders. At the end, I believe that when you apply common sense you will do just fine. (e.g. don't trade TRAVELOCITY and EXPEDIA and think your portfolio is diversified).

Robert

damian
Roundtable Knight
Roundtable Knight
Posts: 815
Joined: Tue Apr 15, 2003 8:43 pm
Location: dusseldorf

Post by damian » Sat Apr 19, 2003 11:28 am

I see diversification as a means of increasing opportunity.... plus the good and bad that each opportunity may bring.

I have proven that for some systems, increased 'diversification' by adding more markets leads to increased volatility. But it also made you richer*. I recall having seen other people demonstrate the same thing.

The benefits of diversification in an asset management world usually refers to different asset classes. I think the benefits of diversification (as a risk reducer) are reduced when you spread yourself amongst different assets within the same class.

I think diversification within an asset class (ie all futures markets traded using one method = an asset class, the more markets you trade, the more diversified you are) has benefits as a risk reducer to a point, but after that point is reached the marginal benefit is reducing and will become negative. I would like to assess the marginal benefit in net return relative to the decrease in risk of adding one more market. If I pull out an old portfolio theory text (which I obviously didn't open very often!), I will probably find that in some form, this study has already been done for me.

A collection of totally independent systems that do their own things at different times for different reasons is an example diversifying between asset classes in the context of mechanical futures trading and should help to reduce risk. I think that it is perhaps misguided to add more markets to a single system in the hope of reducing risk under the name of diversification.

* but that depends on whether you added the correct markets. Easy to do in hindsight, not so for the future. An apple will only fall on your head if you stand under an apple tree. So stand under many apple trees. which ones? You could try to predict which tree is most likely to drop an apple, but that is crystal ball stuff and I don't use one.

Kiwi
Roundtable Knight
Roundtable Knight
Posts: 513
Joined: Wed Apr 16, 2003 1:18 am
Location: Nowhere near

Post by Kiwi » Sat Apr 19, 2003 7:47 pm

To Sir G: got it, loved the story but do see my response to DaveS for a little support for Goldie.

To Dave S: before I respond be aware that I speak from the perspective of one who entered futures to get diversification that I didnt perceive in stocks.

For me diversification provides returns with psychological comfort. I want returns in the 50% to 100% pa range and Close Trade Drawdowns of less than 30% per annum.

If you look at a long term system (say a trend follower like aberration or gsx) you see that you cant do this when only trading one market because the average returns per year on corn say might be less than the average maximum drawdown (and 1/2 the return on the maximum drawdown over the last 20 years). So if you trade for 50% pa and the return = 1/2 the maxDD then you'd better be ready for a 100% drawdown. :cry:

I can see two obvious ways to overcome this:

1) You do what a floor trader does (or a gold bug could do if good enough) which is that you trade with great frequency but small size. Your frequency is 10-100 times that of a long term trader so you can have wins 1/10 -1/100 the absolute size of a trend follower but 10 or 20 losses in a row doesnt wipe out your capital and you make 50% to 100% per annum.

I trade one system with basic position capitalization of 50,000 units. The bet size is 1 unit (not 1000 units as it would be for long term trend following) and the target win is 50 units (but average =80) but it trades 2-3 times per day so it is able to return 50,000 units+ each year with a max drawdown of 8,000 units in the past 5 years. Despite these figures (which I consider good) I trade it on two markets which have only once had the drawdown occur at nearly the same in that five year period thus smoothing the equity curve .... which brings me to

2) You do what long term system traders do: you bet that not all max drawdowns will occur at the same time. You choose 10-20 markets and you trade 1-3 systems on them. When you look at the statistics you find that you get a Closed Trade MaxDD of 15% and an average return of 50% per annum in the first cut.

Because you are conservative you run Monte Carlo sims on the trade outcomes to determine the likelihood that enough of the market/system combinations will come close to MaxDD at the same time to wipe you out. Most people probably set that number at 1 in 1000 or 10000.

Then you trade. But just like the floor broker who doesnt know if hes going to hit too many bad trades on bonds in a row you dont know if this is the year that the DDs will add up at the wrong time. Sure you can look for system failure but if you assume it too early you are likely to stop just before things come right. So courage (the strength to carry on despite the fear you feel) comes in for either trader (I was inspired that story ).


So my answer to your two questions is:
- 1a. You cant diversify if you are too small and the odds of ruin are likely to be high so you shouldn't trade unless you can find smaller contracts which permit diversification.
- 1b You diversify when starting a new system by taking trades as they come until you reach your limits. In my experience you bet small enough that no one sequence of losses is going to ruin you before you get enough different positions on :)
- 2 Pit traders dont diversify. They trade small relative to their equity but with sufficient frequency to get a good return pa.


To Robert: I agree that effectiveness wears off with number of markets and I seem to recall that between 10 and 20 markets it gets close to zero.

Despite this I think it pays to over diversify to allow for the possibility that historically uncorrelated market/system combinations may correlate in future. I think that it also pays to trade a proportion of your equity with different timeframes because in a period where long trends dont occur, short sharp ones may occur. Similarly a fading strategy is likely to perform well while a trend follower is in drawdown.

The reason that I personally do this that I never know how close to the 1 in 10000 event I am on my long term systems. So the comfort I get gives me the strength to take every trade even when I think that it is going to be a loser (that was also a great story Vince). It helps to make up for my weakness.


Damian: absolutely right about diversification although when you are a Kiwi living on the Gold Coast, Australians dont seem to be thin on the ground :shock: Thanks for the tax hint.

OK ... enough of the forums time from me.

John

Christian Smart
Roundtable Fellow
Roundtable Fellow
Posts: 50
Joined: Fri Apr 18, 2003 8:53 pm
Location: Huntsville, AL
Contact:

Post by Christian Smart » Sun Apr 20, 2003 10:10 pm

One possible way to determine the best possible portfolio for a trend-following system is to choose only the markets that trend well. Has anyone tried this? As far as I can tell, this idea was introduced in a "book" that Bruce Babcock put out years ago called Trendiness in the Futures Markets. He studies 29 different markets, and uses his DMI system (which seems to be a simplified version of Welles Wilder's system) to determine trendiness. Babcock also discuss this system in his Guide to Trading Systems book, where his system has the following rules:

1. Go long on the close when DMI is positive
2. Go short on the close when DMI is negative
3. Positions are reversed when the sign of the DMI changes from positive to negative and vice-versa
4. The system uses no stops

The purpose of this "system" is only to determine the trendiness of a market. He averages results from using every period from 5 to 85 days for the DMI calculation.

Babock produces a ranking list in his book on trendiness that spans the years 1983-1996 using average profit per trade as his criterion. Swiss Franc, the D-Mark, and the Yen make up the top three, while the S&P 500 and the NYFE are at the bottom of the list. (He also produces a lot of other statistics. The book mostly consists of lists of results for various years and for diffrent time periods upon which the DMI is calculated.)

I've tried to use Babcock's book as a starting point upon which to do some trendiness research of my own, but have had problems reproducing his results. I know how to calculate DMI, and I'm using continuous reverse-adjusted contract data from Pinnacle. But I usually wind up with many more trades per year than Babcock does, and lower average profit per trade.

Babcock is not very clear about which system he uses - either his own modified DMI system from his Guide to Trading System book, or Wilder's original system (which has either one or two additional rules, depending on how you read his description of the system).

Has anyone else looked at using a trendiness measure to aid in portfolio selection, and in particular, has anyone successfully reproduced Babcock's results?

Josh M.
Senior Member
Senior Member
Posts: 38
Joined: Mon Apr 21, 2003 2:22 am
Contact:

Post by Josh M. » Mon Apr 21, 2003 12:15 pm

Greetings. It is nice to be here with such talent. I think the tone of this forum is largely correct that diversification for system trading is necessary to be successful over long durations. Good Trading.

SwingTrader
Contributing Member
Contributing Member
Posts: 6
Joined: Mon Apr 21, 2003 3:46 pm
Location: Russia

Re: Diversification

Post by SwingTrader » Mon Apr 21, 2003 3:57 pm

Dave S. wrote: 2) How do you diversify if you can only trade one market?
Easily. In this case you still have an ability to:

1) apply multiple uncorrelated strategies;
2) diversify timeframes;
3) even use multiple position sizing methods for them.

Best regards from Russia.

Post Reply