Trading with solid foundations

Foundations

The main purpose of this essay is to establish a sound basis for a profitable trading system, just as we would prepare a strong foundation before building a house. Developing a trading system without a sound foundation is akin to building a house on shifting sand.

It is often said that trend-following trading systems are only profitable when the markets are trending, as most trading systems fail when market conditions are not suitable. It is my contention that these systems are not profitable at all, and that basically they ride solely on the profitability of the underlying trend. As soon as the trend turns, so does the system's profitability - to the point where profits accumulated during trends are returned to the market with interest.

For this reason alone, it is absolutely essential to measure system performance against a solid benchmark to determine whether it is profitable on its own accord.

Benchmark

In the real world of trading/investing, performance is almost always measured against a benchmark - usually to a related composite index.

Most managed funds' performance is measured against a main market index. On average, it has been found that most funds return about 2%pa below their index benchmark - the 2%pa discrepancy being mostly due to management fees.

So now, lets choose our benchmark.
All averages are based on median performance of the universe of backtested securities, using the methodology outlined here.

Universe of stocks:
• 490 ASX All Ords (Australia) securities.

Test Period:
• 9/July/2001 to 3/March/2006 (4.6 years).

Test parameters:
• Buy/Sell on Close of signal day;
• No shorting;
• No profit compounding used;
• All net profit% metrics include 0.2% total brokerage.

ASX All Ords index performance:
• +10.2%pa profit (+47.5% total), 23.3% max drawdown.

Median All Ords security performance:
• +18.9%pa profit, 76.8% max drawdown.

So, why is there a performance difference between the market index and its constituents?
Perhaps we can break down the median security performance into these two components:

• Market trend bias: +10.2%pa
• Survivorship bias: +8.7%pa

Survivorship bias comes about from the tendency of any current list of securities to exclude failed stocks. Indices such as the All Ords, are basically lists of survivor securities which don't include major failed stocks such as Enron (US) and HIH (ASX). This positive bias can easily add 2%pa to any backtest result, resulting in false performance expectations from excessively false profit results.

If we can ignore both market trend and survivorship biases, the +18.9%pa net profit sounds like a good deal until we strike the dreaded interim 76.8% drawdown. Hands up those who would continue trading after finding their capital reduced by 3/4...

So, to finalize our benchmark we need to reduce our exposure to the market and risk-normalize profits to an acceptable level. I chose 30% for our purpose.

Performance metrics

The median All Ords security performance (risk-adjusted to a maximum 30% drawdown) is +7.9%pa net profit. Therefore, +7.9%pa will be our risk-adjusted benchmark when backtesting this universe of 490 ASX securities.

Again, using the backtesting methodology outlined here, let's take a look at a simple price breakout strategy (available in the URSC tool-kit v3.0) and apply it to our universe of stocks.

Price breakout backtest results:

• +1.1%pa net profit, 6.8%pa below benchmark;
• Median trade duration of 16 calendar days (2.3 weeks);
• 37.4% median position size.

If it wasn't for the long-term uptrend and survivorship bias of our universe of stocks, this price breakout strategy would lose an average of 6.8%pa. Not looking good for down-trending years...

Now, applying the same breakout to the security's GICS Sector index, and totally ignoring any price entry/exit conditions for individual securities, yields these interesting results:

• +13.1%pa net profit, 5.2%pa above benchmark;
• Median trade duration of 19 calendar days (2.7 weeks);
• 97.3% median position size.

Suddenly, this simple breakout strategy is a winner, and yet we are not even taking each security's price into consideration for individual entries/exits!

What is happening here?

Market fundamentals are influential

The simple answer to the question above is that the ubiquitous fundamental market forces that drive a whole industry, influence prices to a great degree and are much more powerful than most traders realize.

These common fundamental forces account for an estimated (by my own testing) 30% to 80% of price movement, depending on the trade duration and industry group involved.

Using just price action alone and not taking sector fundamentals into consideration can lead to much frustration in system development, as generally we would be applying T/A techniques to chart noise. Basing trading strategies on price alone, is akin to trying to launch a ship by only looking at the shore waves and ignoring the tide.

This knowledge provides the sector-based trader with a sizable potential advantage over traditional price-based techniques. For example, traders can now afford to apply traditionally risky but potentially rewarding contrarian techniques (such as price divergence) to their sector-based trading, in order to gain a value edge over more conventional price-based strategies.

Choose your industry groups wisely

Not all Industry Group fundamentals are equal in their common fundamental or gravitational pull - some are stronger than others.

Let's take the Energy sector for example.
One of the major fundamental forces driving this whole sector is the price of crude oil. Oil price goes up, and almost all energy stocks usually go up with it. An energy stock would have to be doing something wrong for its price to go down at this time. There is a high positive correlation between the Energy sector's index and its constituent securities.

Now lets take another industry sector, such as Media.
There may be some common fundamental forces at work in this industry sector, but they are unlikely to be strong. There is a low positive correlation between the Media sector's index and its constituent securities.

The URSC tool-kit v3.0 is basically three trading tools in one package. It comes complete with sector-filter explorations that allow the selection of highly correlated industry groups, matched to their constituent securities. This results in the following performance metrics (risk-adjusted to max 30% drawdown):

• +17.9%pa net profit (83% net profit over test period), 10%pa above benchmark;
• Median trade duration of 20 calendar days (2.9 weeks);
• 75.4% median position size.

Putting our sound foundation to the test

So, the URSC tool-kit's sector-filtering strategy returns only 10%pa above median market performance, eh?

Fund managers would kill for this kind of performance. And remember, this is only a base for further development of custom trading systems - I'll leave this task to hard-working individuals.

In the meantime, keep an eye on Polaris, a trading strategy based on the above principles, being traded by a Sydney-based stockbroker with a total portfolio capital of approximately AU$500K

Trade with the strength.

jose '-)

Last updated on 21st August 2006