deterioration of performance over the bull market in the 1990s. Most likely this
would have been counterweighted by a better performance over the last few years
and perhaps also a smoother equity growth over the entire period. This would have
meant that we could have increased the performance even further by daring to risk
a bit more in each trade.
However, to be able to identify and do something about any of these reasons
for system failure, we need to do the research the right way from the very begin-
ning. In Part 3, we learned how to come up with a set of stops and exits that should
work, on average equally as well on all markets.
Although we didn’t do it in this book, analyzing two variables at a time using
a surface chart also can be used for the entries. Here we instead used a set of
entries developed earlier for the Trading Systems lab pages for Active Tradermag-
azine. In Part 2, we took the liberty to tinker around with them a bit, analyzing the
results from the changes using normalized results.
To work with normalized results, we must make sure that the data are cor-
rect, which testing variables are the most important to look at, and what type of
results to look for. Without knowing this, there is no way we can isolate any type
of problem or error, no matter where it occurs during the testing process. These
concepts were covered in Part 1, and thus we have come full circle.
384 PART 4 Money Management