The short side and trend following

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.
Eric Winchell
Senior Member
Senior Member
Posts: 38
Joined: Wed May 28, 2003 2:17 am

The short side and trend following

Post by Eric Winchell » Fri Jun 20, 2008 2:56 am

I haven't been super creative, but I've yet to find a traditional trend following or breakout system that has any edge at all on the short side using data back to 1984 for 60+ markets (aggregate portfolio).

The reason seems obvious enough: trend following rules favor a particular type of uptrend that starts calm and ends with volatility and high prices. The nature of price action in downtrends seems to have different characteristics. My test results indicate it's more common for upside momentum to build and accelerate for longer periods than for selling to build and accelerate in the same way. This says nothing about the existence of long downtrends or that markets become oversold, but it does lead me to believe that the path to that oversold low point often takes a different form -- enough that I can't find an edge in shorting downside breakouts.

Comments?

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

Post by sluggo » Fri Jun 20, 2008 10:47 am

The set of {breakout systems you've simulated} combined with the set of {markets you've tested, over the slice of past history you've tested} has not been satisfyingly profitable on short trades.

Naturally this leads to two thoughts. First, what about the systems you haven't tested? Second, will the markets of the future have the same characteristics as the markets of the past?

A lot of people think it is unwise to assume that shorting will turn out to be unprofitable in the future. Some of them have decided that a safe approach is to trade ALL possible markets, and trade both long and short. Don't ever miss an entry signal, because you never know which one is going to be the mammoth +50 R-multiples-of-profit trade of the year. It might even be a short trade.

Something else to think about might be the correlation between (the equity curve of a system's short-only trades) and (the equity curve of the same system's long-only trades). In each market, the short trades and the long trades will be non-overlapping, so we expect the correlation will be very low, for each market. This may lead to a low overall correlation for the entire portfolio, as measured by the equity curve.

If you test it out, you'll probably find that experiment agrees with theory. Short-only trading actually does have a low correlation to long-only trading. Who cares about correlation, you may ask? Modern Portfolio Theory cares. It suggests that combining non-correlated equity curves (also known as "diversification") can result in higher gain-to-pain ratios. Adding a short-only system to a long-only system, can be beneficial, or so says Modern Portfolio Theory. An example is shown below.
Attachments
SHORTZ.png
some recent downtrends
SHORTZ.png (153.94 KiB) Viewed 11138 times
diversification.png
Improving a long-only system by adding short trades
diversification.png (10.42 KiB) Viewed 11136 times

Eric Winchell
Senior Member
Senior Member
Posts: 38
Joined: Wed May 28, 2003 2:17 am

Post by Eric Winchell » Fri Jun 20, 2008 3:13 pm

I'm almost willing to buy the portfolio theory argument but consider that by trading all the margin-to-equity ratio almost doubles and the average total risk profile increases. Do you really feel good about using twice the leverage just to get a slightly higher MAR ratio?

And in some cases, the volatility-adjusted returns don't improve. I'm not comfortable hedging against something without evidence that it ever existed.
Attachments
ScreenHunter_01 Jun. 20 14.03.gif
ScreenHunter_01 Jun. 20 14.03.gif (7.96 KiB) Viewed 11103 times
Last edited by Eric Winchell on Mon Jun 23, 2008 1:30 am, edited 1 time in total.

AFJ Garner
Roundtable Knight
Roundtable Knight
Posts: 2040
Joined: Fri Apr 25, 2003 3:33 pm
Location: London
Contact:

Post by AFJ Garner » Sun Jun 22, 2008 12:39 pm

I do not "like" trading short either; it seems a waste of risk and resources in some ways. It ain't made much money even if it has smoothed the curve. Trading short has been particularly odious this decade. But taking another look at the CRB index, which I posted recently elsewhere on the forum, leads me to suppose that the past 8 years might have been particularly tough and that it won't always be that way.

I suspect you will get good and bad periods over the years to come as you have in the past. I do not suppose it will ever be a money spinner in its totality but I think its usefulness will continue.

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

Post by LeviF » Sun Jun 22, 2008 1:36 pm

I trade forex, so being long/short depends on the arbitrary order you put the pairs together. Long EUR/USD, short USD/EUR, no difference. With futures, isn't the relationship pretty much the same, just replace one of the forex instruments with a commodity? Some food for thought...

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

Post by sluggo » Sun Jun 22, 2008 3:27 pm

The third Turtle educator, the one whose initials do not include the letter "C", says that a piece of advice taught in the Turtle class was
  • If you're not sure whether or not to (perform some action), do half. That way you will only be half-wrong and won't become emotionally destabilized.
Here are a few ideas of ways to apply this suggestion when deciding "To short or not to short"
  • Allow half of the systems in your Suite to trade Long Only, and tell the other half of your systems to trade Both Long And Short.
  • Skip half of the short trades
  • Take all short trades but trade them at half size
  • Take all short trades but scale into short trades in two pieces, don't put on the second piece until the stop on the first piece has moved to breakeven.
  • and so forth

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

Post by LeviF » Sun Jun 22, 2008 5:20 pm

What "three" turtle educators are you referring to here? W.E., R.D., C.F.?

ladadriver
Senior Member
Senior Member
Posts: 25
Joined: Tue Feb 27, 2007 10:08 am

Post by ladadriver » Sun Jun 22, 2008 6:15 pm

sluggo wrote:
  • If you're not sure whether or not to (perform some action), do half. That way you will only be half-wrong and won't become emotionally destabilized.
In this vein, the following thread on non-symmetric long/short bet sizing is interesting.

viewtopic.php?t=1668

Eric Winchell
Senior Member
Senior Member
Posts: 38
Joined: Wed May 28, 2003 2:17 am

Post by Eric Winchell » Mon Jun 23, 2008 1:19 am

> I trade forex, so being long/short depends on the arbitrary order you put the pairs together.

levijean -- I have thought about that and my argument doesn't stand up in the cased of cross rates unless you consider (and you might not) that in most crosses the reserve currency is psychologically important. For example, if the USD is the world's reserve currency it's predominantly either a short or long in the USD. If anything goes to zero it's probably not going to be the reserve. An unrelated quirk happens with bonds where rates are not going below 0.

AFJ Garner -- the test in my previous post is from 1984. I'm going to buy some more data to 1970 and see what that brings to the picture.

ladadriver -- thanks, I knew I wasn't the first to comment about it.

Solong
Roundtable Fellow
Roundtable Fellow
Posts: 87
Joined: Sun Jul 01, 2007 11:24 pm

Post by Solong » Mon Jun 23, 2008 6:09 am

Eric, thanks for posting your conclusions. Don't abandon your search for short-only systems. You are on to something here which sluggo raised: the correlation between portfolios and diversification benefits.

Granted, the Trade Short system in your example was not attractive with a 50% drawdown. It should be feasible for you to find a better short-only strategy which you could combine with your long-only version to good effect. You are somewhat hesitant about the portfolio theory argument but it is well established and if you continue your quest you should eventually find attractive propositions that improve your aggregate portfolio.

The short-only system does not need to have a big edge overall as long as it makes some money when the core TF strategy is in a drawdown. The evaluation criteria for the second complementary strategy are not the same as for the first, single system.

Consider also that the increase in risk through leverage will not be completely additive. By allowing long and short trades you will sometimes carry opposite positions in correlated markets (e.g. same sector or currencies) which will lower your market risk and exposure to unforeseen shocks. Value at Risk, which is harder to calculate, may give a truer picture.

Eric Winchell
Senior Member
Senior Member
Posts: 38
Joined: Wed May 28, 2003 2:17 am

Post by Eric Winchell » Mon Jun 23, 2008 6:54 am

I wouldn't hesitate to trade short in a large account because logically it's less systemic risk and provides some price shock protection.

In a smaller, aggressive account I'm leaning toward trading the higher expectancy until the lower risk is more affordable. The issue is that the higher required margin of including a weaker system can lower the volatility-adjusted performance but it comes at the cost of lowering the potential leverage (and potential returns).

7432
Roundtable Knight
Roundtable Knight
Posts: 199
Joined: Wed Dec 31, 1969 7:00 pm
Location: sandy, ut

Post by 7432 » Mon Jun 23, 2008 12:31 pm

along with the above comments about adjusting the short side of trading, I would add breaking out groups or sectors and testing long/short/trade all.
as great as long only trading in ags, energies and metals has been, I wouldn't want to miss out on the short side of european bonds lately.

if jim rogers is correct that we have years to go in this commodity bull market it might make sense to trade long only ags, energies and metals. but do you really want a long only financials during the same period?

or how about a long only system missing out on 15+ years of the gold bear market?

Solong
Roundtable Fellow
Roundtable Fellow
Posts: 87
Joined: Sun Jul 01, 2007 11:24 pm

Post by Solong » Mon Jun 23, 2008 12:34 pm

Those are important considerations. I did not think of it when I saw your $4.6bn simulation.

Returning to your first post, one could also recognize that some markets have an upward drift, like the stock markets and commodities, whereas others are mean reverting such as interest rates and foreign exchange. It would affect the relative performance of long-only and short-only systems.

Eric Winchell
Senior Member
Senior Member
Posts: 38
Joined: Wed May 28, 2003 2:17 am

Post by Eric Winchell » Mon Jun 23, 2008 7:39 pm

Solong -- arguably all real assets have an upward drift as a result of inflation. The current boom in the Prices of Everything can be partially explained by the lag of prices responding to the expanded money supply of the last 20+ years. A few years ago I remember taking the CRB index and adjusting each component for inflation, and it was astounding how low prices for food, metals, and energy had become. Most have doubled or tripled since then but some haven't moved much at all... sugar, for example.

7432 - funny you mention that. My only "piece of flare" is to trade a short-only bond portfolio. There are great reasons for rates to go higher that I think will catch on in the mainstream. In a nutshell, the unwillingness of foreign countries to hold dollars will bring high rates both directly (unloading of bonds) and indirectly (Fed actions). The U.S. is going to *need* higher rates.

As for missing a 15 year gold bear market -- I'm not worried about "missing out" on short trades as a whole because all my testing indicates they simply don't have the same expectancy as longs. With limited resources I'm choosing to be very picky about what I'm willing to short and when.

I suppose the crux of it is the decision of whether to be a 100% mechanical trend follower or allowing for some discretion. The truth probably is that I'm just not smart enough to model the markets as well as is required to be 100% mechanical. This is why RenTec needs scientists and mathematicians to do what they do. I recently reread a few interviews and books (Way of the Turtle included) and it was a great reminder that most of the famous trend followers do allow for discretion in their trading. It's an unpopular fact that the Turtles were asked to bring something to the table in addition to what they were taught. In his interview Ed Seykota talks a lot about trades he makes that are outside the mechanical rules. I think Ed embraces the ongoing conflict that arises between someone who has found an edge in a system but also has valid instincts. I also hadn't remembered Larry Hite as being the only original Market Wizard who purported to trade a 100% mechanical system.

I have a computer science background and am a fan of frameworks in programming languages. A framework is a common code base that removes a lot of the minor details from projects that have the same foundation. They're basically meant to save time (and potentially represent the evolution of a language). To me trend following systems are a great framework that often have the benefit of an inherent edge. In the absence of any discretionary ideas, the system can be traded on its own. At the same time, if you think bonds are going down for the next 10 years, you can act on that bias by taking only short signals in those markets. In only taking a certain signal direction or using a different amount of leverage the day-to-day actions are still reduced to elegant simplicity in terms of the framework of trend following.

When you look at the performance of well-known systems and you know what the pros can do, mechanical trend following just isn't that impressive. But there is a hidden gain that makes it worthwhile: the potential for emotional detachment. This is what I want to exploit.

jasonz
Full Member
Full Member
Posts: 10
Joined: Sat Apr 14, 2007 10:37 am

Re: The short side and trend following

Post by jasonz » Tue Sep 30, 2008 6:04 am

Eric Winchell wrote:I haven't been super creative, but I've yet to find a traditional trend following or breakout system that has any edge at all on the short side using data back to 1984 for 60+ markets (aggregate portfolio).

Comments?
I have a specific theory on this. Long term trend following is looking to typically capture large breakouts to really get the profitability up, rather than many high frequency small winners.

For the largest breakouts, your position sizing is fixed for the period in which the trade is open. If you go short, the move will generally be inverse-exponentially against you.

To explain, a move of 100% up should be equivalent to a move of 50% down on the log scale. That is, an index of something can go up 300% but it can only go down 99%.

Even at the scale of +25%, the corresponding move down is -20%, indicating 20% less profit for the short side if you believe prices evolve log normally.

Inflation is another point to add to the mix. Inflation (the persistent devaluation of money) is not avoidable. That is, (and maybe someone wants to test this), commodities and stock indices in general trend upward over time due to inflation (we don't generally see a correction to this with deflation).

This means that any long position gains additional profitability bonus due to a small amount of upward trend.

Another point to raise would be that, for many indices, such as stock indices in particular, moves downward are extremely disorderly, and there is little inherent information in the price movement often to predict the snap when they drop like lead balloons...

Ok those are my thoughts, because I have had exactly the same ideas on the short side of trend following systems.

The theory that is most explanatory to me is the log-normal distribution story. Assets are move exponentially upwards, and inverse-exponentially downwards, as price movements day on day are always measured relatively to the present price.

I think there are costs for avoiding the short side however, there do seem to be "portfolio effect" contributions to smoothing net P&L when the short side is added, even when it is not specifically profitable, due to the non-correlation to other trading positions. This is just a loose statement of my intuition here.

freeman
Roundtable Knight
Roundtable Knight
Posts: 114
Joined: Tue Oct 04, 2005 7:12 pm
Location: Switzerland

Post by freeman » Tue Sep 30, 2008 2:33 pm

Have u heard of the the saying (what goes up must come down)Known in the trading world as cycles.
Futures never go much higher than the last high simply because theres more money in the markets,but it still needs to cycle back down.

Fundmentals,no thanks,,,the USD has weakened over the last few yrs because"economic crissis"then y has it strenghten in the last few mths at the crisis worsted point?It just goes to show,,just follow the price.

As in stocks,take yahoo chart as example,it is now the price it was back in 1997,a close today 17.27,,it has been as high as 1360 odd in 2000.
DingDing next,,mmmm,my crstal says google sumwhere,im sure theres another leader soon.

Life is all trends,like the fashion market its in today and gone
tommorow,,jump in then jump short later and its all over.
But its not to say it wont come back in fashion again.
just my own opinion

jasonz
Full Member
Full Member
Posts: 10
Joined: Sat Apr 14, 2007 10:37 am

Post by jasonz » Tue Sep 30, 2008 5:34 pm

freeman wrote:Have u heard of the the saying (what goes up must come down)Known in the trading world as cycles.
Futures never go much higher than the last high simply because theres more money in the markets,but it still needs to cycle back down.

Fundmentals,no thanks,,,the USD has weakened over the last few yrs because"economic crissis"then y has it strenghten in the last few mths at the crisis worsted point?It just goes to show,,just follow the price.

As in stocks,take yahoo chart as example,it is now the price it was back in 1997,a close today 17.27,,it has been as high as 1360 odd in 2000.
DingDing next,,mmmm,my crstal says google sumwhere,im sure theres another leader soon.

Life is all trends,like the fashion market its in today and gone
tommorow,,jump in then jump short later and its all over.
But its not to say it wont come back in fashion again.
just my own opinion
Yes freeman, but the price evolution is such that smaller profits are made going downwards, where as larger profits are made going upwards. A price can easily double on the way upwards, but it can only decrease its relative volatility on the way downwards.

Short side systems have never tested as profitably compared to upside systems.

If we had contracts which paid relative to log-movement then they would be equivalent. However this is not the case.

Think of it this way. Suppose you go long 20 contracts, and your position sizing is relative to the notional size of the position as well as volatility.

You are not equally likely to get a 25% movement upward as a 25% movement downward.

While the mathematical expectation for the future price in 100 days time may be the price today, the mathematical expectation of any trend following system for any price movement which evolves log-normally will be biased toward the upside. This is simply due to the fact that the range of movement is exponentially expansive upwards, and will contract on the short side.

What I am saying is that if the EUR is worth $1.50 today, then its volatility will be relative to the current price. If it's 0.75c then the volatility is relative to the current price again.

Now if you got in the contract at 1.50, and it moved to 0.75, for 0.75c profit, and you are still in the position, the volatility is ever decreasing in "trade size terms"

Whereas a movement upwards which is equivalent to a halving of value is a doubling of value, i.e. $3. And if you are there your volatility is ever greater in notional terms and you are still in the position.

These are the fundamental reasons why longs are more profitable than shorts.

For currencies, as they are strictly relative measures, they can move infinitely (on the **log** scale) in any direction. However, any contraction relative to the notional value of the futures contract is a contraction in profit making opportunity.

If you rebase the contract in the other currency, then you may invert the situation and have increasing growth of profitability per % daily movement away from the entry price, as opposed to decreasing growth of profitability per % daily movement.

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

Post by sluggo » Tue Sep 30, 2008 7:10 pm

jasonz wrote:Suppose you go long 20 contracts, and your position sizing is relative to the notional size of the position as well as volatility.
None of the futures trading systems that are supplied when you buy Blox, size their positions this way. Neither do any of the futures trading systems that I personally am trading with real money.

In Blox's systems and in my systems, the notional size of the position is irrelevant. Notional size is the answer to the question "How far is it from the current price, to Zero?" But I, and other systems traders, don't actually care how far it is to Zero. We only care how far it is to our Exit Point, a distance we call the Estimated Risk.

Does there exist any Eurodollar futures trader, who sizes his positions on the assumption that he might hold onto a long position as the price of Eurodollar futures fell to Zero? (implying a Libor interest rate of 100 percent per annum?) None that I've ever met.

A simple experiment provides further evidence that the distance to Zero is unimportant to futures traders: Generate a continuous contract using "Accumulation Method = Back Adjusted" and run it through a system. Now generate the contract again using "Accumulation Method = Forward Adjusted" and run it through the same system. These two price series have different offsets; their Distance to Zero is different. But they produce exactly the same profit and loss and exactly the same equity curve. Distance to Zero didn't affect the position sizes or the daily returns.

Where notional value DOES matter, is in stock trading. In stocks (unlike futures), the exchange requires you to set aside an amount of money when you enter a trade, which is a function of the Distance to Zero. But in futures, the (tiny) amount of money set aside at trade entry, is not a function of Distance to Zero. So futures traders size positions based on Estimated Risk instead.
Attachments
CSI.png
CSI.png (28.88 KiB) Viewed 10051 times

jasonz
Full Member
Full Member
Posts: 10
Joined: Sat Apr 14, 2007 10:37 am

Post by jasonz » Tue Sep 30, 2008 8:31 pm

Ok, I get what you are saying sluggo. And from what I see on these forums you are a successful and profitable trader, so I am just putting this as food for thought and am not meaning to correct anyone's thinking - rather point out a mathematical fact which I had been fighting with for quite a while, but has now become very clear to me.

If you are selecting your number of units on a volatility basis in terms of dollar risk, then you may see that you implicitly invest on a notional basis even if not explicitly so.

When we look at an exchange rate, suppose a currency started to suffer from significant inflation, and a rate were to traverse toward zero, what you would ultimately see is extra zeroes added to the quotation prices.

You are going to get price movement which is determined ultimately by the "distance from zero" as it approaches zero. Position size is going to therefore be implicitly sized by this relationship.

If your trading system is looking for wins and losses of size ~ 1% of notional, then I generally agree, the log-bias is small enough to be nothing to really be concerned about.

However if the size of the profitable moves is around ~10%, then the corresponding log normal move down is ~9%. Even at 10% move, you are getting -1% absolute or -10% relatively less profitability (i.e. log bias) in the profitability of short side trades.

Also don't forget that you get all the bad things too. Movements against you upwards are log against you when you are short also, so that if your average loss is 10% when going short then it will only be 9% when going long.

We all recognise that short positions are not as profitable as long positions for long term trend following systems.

The explanation is very much to do with the fact that asset price evolution can be assumed to be approximately geometric Brownian motions.

The simpler way to say this is that prices move in proportion to their present values, and they don't move in straight lines without reference to their underlying value. That's not a controversial statement, I think it is a straightforward one.

And what I am saying about log-bias against short positions is a direct inference from this well accepted hypothesis about price action.

What is the inference to model building?

Just to accept that short positions are riskier and less profitable if you are looking to capture large movements, however to exclude them is to exclude a healthy source of non-correlated positions in your portfolio, but one must independently analyse the shorts to ensure they have positive expectation and diversification contribution in their own right, with all these issues against them.

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

Post by sluggo » Tue Sep 30, 2008 11:59 pm

jasonz wrote:If you are selecting your number of units on a volatility basis in terms of dollar risk, then you may see that you implicitly invest on a notional basis even if not explicitly so.

You are going to get price movement which is determined ultimately by the "distance from zero"
Good. A testable hypothesis: Volatility is larger when price (notional value) is larger, thus volatility based position sizing will give very similar results to notional value based position sizing

Have you tested this hypothesis? I ran a quick test, just now. Here is my experiment and my results.

I grabbed the data series for Hogs futures on the CME, and made a scatter plot of Volatility (the 10-bar exponential moving average of True Range) versus Price (the unadjusted Close). If your hypothesis is correct, the dots should slope up and to the right, i.e., larger Price (notional value) will cause larger Volatility.

The linear regression trendline & equation are plotted, along with the coefficient of determination R-squared. R-squared is quite low (0.015) indicating essentially no relationship between Price and Volatility. Numerically, 1.5% of the variation in Volatility is explained by variation in Price. The other 98.5% of variation in Volatility cannot be explained by variation in Price.

This data does not support your hypothesis at all, in fact it argues against your hypothesis: Volatility (position size) is unrelated to Price (notional value).

Could you show some of your testing results please?
Attachments
hogs.xls
Excel spreadsheet with price & volatility. Contains chart.
(659.5 KiB) Downloaded 343 times
scatter.png
Plot of Price versus Volatility for Hogs futures
scatter.png (37.97 KiB) Viewed 10028 times

Post Reply