January 2, 2015

Trading Short-Term or Not? That is the question.

Whatever automated trading methods you might have used in the past, use now, or will use in the future, it has for unique purpose to make you money. It's not important that the code you use is well structured and nice or which software you will use. What's important, however, is the ultimate outcome of the trading strategy. One should understand what it really does and how it behaves under favorable and unfavorable conditions.

I would suspect that anyone's primary objective is not to give their trading capital to other traders, even during the holiday season. Yet, I often see trading strategies designed to do just that.

Should you trade short-term (in the minutes, days to months range) or trade for the long-term (in terms of years and decades), the goal is the same. Whatever time horizon you might want to consider, the overall trading environment should not only be consistent with the trading horizon selected but also within the trading methodology designed to survive in that particular time frame.

The problem is not to solve a payoff matrix equation for one year: A(t) = A(0) + Σ(H.*ΔP), but over a long-term horizon, say in the order of 20 or more years. There is no need to design a short-term trading strategy if it has, for its most probable outcome, to just blow up in the future and bring down your portfolio with it. For instance, an all in buy on the big dip proposition might give great results until you touch that one falling knife with no intention of rebounding, and thereby give you a major haircut, if not totally destroying your single proposition portfolio.

There might be no need for fundamental data in a short-term trading strategy, yet, it could be of help by trading in the same direction as the long-term underlying trend, even if, on a daily basis, it is small. Just as there is little need to pinpoint a perfect entry when your holding time is a decade or two. It might seldom be the best time to enter a trade, meaning that most of the time, your trades will at least initially see some red (say over 90%+). But this is not a sufficient reason not to participate since, over a long-term horizon, you will observe that it was better to participate, even under those adverse conditions than not.

This is illustrated by the following chart:

Fig. 1   The Origin of Long-Term Profits

Long Term SDE

(click to enlarge)

The above chart shows the relative value of each component of an SDE when looked at from a long-term trading interval perspective. When the trading interval is small, most of the price composition is from the random-like component of the equation. At the other end of the spectrum, daily price variations have only a minor impact since the long-term trend is the main reason why prices got there.

Notwithstanding, there is an incentive to shorten the average trading interval for positive trades: this tends to increase the annualized CAGR as shown below: 

Fig. 2   Trade Interval Effect

Shorter Trade Intervals

(click to enlarge)

But there is this dilemma: the more you shorten the trading interval, the more you are faced with randomness, as shown in the first chart. It is not a question of whether a stock price random walks or not. It is: can you predict what's coming in some way? Or another way: can you find trading methods that will behave as if you could predict without using forecasting?

The past is recorded history and has little current value since it can not be traded. The real question is what can you do now that will make you benefit at some time in the future? There is no more important question in this "game". You want ΔP > 0 on most of your future trades. This, in turn, translates to the following objective: Σ(H(S#n).*ΔP) > Σ(H(B&H).*ΔP) > 0. You want any of your trading strategies to outperform the Buy & Hold over the long haul.

Looking at the problem with only a few years of trading is not enough. Do the numbers. I see it more from the perspective of portfolio doubling time.

Let's look at the problem from a normalized viewpoint. In the chart below, 41 price series were normalized to a $50 price. The chart shows a before and after CAGR evaluation. 

Fig. 3   Price Trend Distribution

Price Trend Distribution

(click to enlarge)

All price series converge to a normalized price of $50 at t=0. The past (t<0) is just historical data. Some stocks have degraded down to $50 while others have improved up to $50. You know about everything concerning the stocks' respective paths; they are part of recorded history. You would have wanted to avoid those above $50 since they were decaying with time.

It is just when you look at the future (t>0) that you realize that whatever might happen is unknown. You can't know which stocks will outperform, which order they'll be in, or even which ones will survive. They are not yet part of history. You know, however, that whatever might happen in the future, you will see some stocks prospering quite well while others will simply go bankrupt. It's just that you don't know and have no way of finding out which stock will do what going forward. Nonetheless, one can extrapolate on their general behavior when considering a sufficiently large group.

From Figure 3, due to the curve coloring, one might think that there is a continuation of the trend. But that is far from reality. All the future is unknown. I could not say which of the 41 stocks would be on top or totally fail before the trading interval was completed. And I certainly can't say which will be the best performer in 20 years' time. However, their respective paths, even though erratic, can crisscross in the same general direction for smaller groups. From this general behavior, one can design trading procedures that can take advantage of the price gyrations without having to forecast individual paths.

Fig. 4  Trade Distribution

Trade Distribution


The above chart snapshot illustrates the order placement of a long-term trading method of mine (DEVX V6 and DEVX V6 Revisited, both tested over 25 years of unseen data) where no forecasting is used and yet appears as if some kind of forecasting was done. Shares are sold near tops and bought near bottoms as a result of what I consider administrative procedures. Its trading script is accumulating shares over the long term while at the same time trading over the process and making some money on about every swing of significance. And if you do thousands and thousands of such trades over a long-term interval, you are bound to make money, almost as a byproduct of the methodology used.

When designing a trading strategy, you have to solve all the problems, not just one, but all of them, and there are many. Otherwise, IMHO, you might have to consider your trades just as your contribution to some other player's pocket. If you don't look at the problems your trading strategy might create going forward, should you be surprised if ever your portfolio blows up?

You build on the shoulders of others before you. That is your strength, all the research that has already been done and that is easily available. And whatever trading strategy you design, it must comply with what the past was or improve on other people's designs.

Looking at the long-term picture and trying to have your trading strategy survive and thrive over the entire period will require making some compromises. Because you don't know what is coming, you need to make provisions in your programs to protect yourself. You need to determine the degree of participation, market exposure, and strength of your convictions.

... to be continued ...

Created... January 2, 2015,    © Guy R. Fleury. All rights reserved.