Finding reliable trading strategies
by Steve Kille
Developing a trading strategy over historical data has become common practice among traders in
recent years. The proliferation of micro-computers in our industry has spurred the trend to the point
where nearly every system sold or used has been developed with a computer using past data.
The focus of most research projects has been to develop trading strategies which produce the highest
level of profit over historical data. This is understandable since accumulated wealth is the ultimate goal
of every trader. Unfortunately, most technicians are simply searching for a set of rules which produce the
highest level of profits over historical data and then using these same rules to make trading decisions.
The real focus of any research should not be the level of profits but the degree of reliability the system is
likely to maintain during actual trading.
For good reasons, optimization has become the technique most widely used in the development of new
trading strategies. It is very powerful and extremely flexible. You could use optimization to find the
best-length moving average or the most appropriate stop level. For some purposes, optimization is the
only viable procedure. For instance, it is the best technique when trying to determine the average
half-cycle of a market for performing Relative Strength Index (RSI) studies. How do you determine the
most suitable number of days to use in an RSI calculation? You would first test a 4-day RSI, then a
5-day, 6-day, and so on to find the length which produces the most profitable results.