76 PART 1 How to Evaluate a System
Understanding Profitability
In this first part of the book, we have learned that, to build good and profitable
trading systems, there are a lot of things to think about. One thing we need to
understand is the huge difference between a good trading system and a profitable
one. Any good trading system can be made into a profitable trading system when
traded on the right market or portfolio of markets, or together with other good
trading systems. The paradox is, however, that not all profitable trading systems
have to be good systems. Most systems will and can be profitable at one time or
another and just because a system is profitable here and now doesn’t mean that it
is a good system.
To understand this, we must understand the difference between the terms
“good” and “profitable” when it comes to system testing and design. A good sys-
tem is a system that works, on average, equally as well on several different mar-
kets. It isn’t always profitable on all of them, but being so on average over time
ensures it is robust and will generate a steady profit over time, when applied to the
right markets. Some markets should not be traded with certain types of systems:
The moves the system aims to catch aren’t profitable enough to make trading
worthwhile on that particular market. A good system that is worthwhile trading on
a market to generate a profit from that market also is a profitable system on that
very market.
To build a good system, we must make sure that the system catches the types
of moves it is intended to catch, no matter where and when those moves took
place, in relation both to time and the current price of the market. To do that, we
must normalize the system’s results in such a way that all back-tested trades get an
equal weighting. This can either be done by always trading one contract only and
measuring the results in percentage terms, or by always trading a fixed dollar
amount so that the number of shares will vary with the market price (higher price,
fewer shares, and vice versa). Only by doing this will a profitable trade in a low-
priced market influence the final parameter setting to the same degree as a prof-
itable trade in a high-priced market.
Also, during this initial research, we should not consider the cost of trading,
such as slippage and commissions. We’re only interested in finding out if the sys-
tem is efficient enough to catch the types of moves on the chart that we want it to
catch. The efficiency is measured in the average profit per trade, which should be
high enough to warrant trading with the costs of trading taken into account. Taking
the costs of trading into account at this initial stage of the testing procedure will
only favor more long-term systems, using few trades that seem to be particularly
profitable on high-priced markets.
However, to do our research correctly, we also must be careful about the data
we use. Many data providers have rounded the data to the nearest tradable fraction
of price after a stock has made a split. After several splits, this error compounds