into a new trade that is still open on the day the second trade started in the origi-
nal system. Note that the second trade in the latter version of the system is possi-
ble because a signal that day went unnoticed by the original system because it was
in a trade already. The second trade in the latter version of the system then is
stopped out after 10 days, which would have been while the second trade in the
original system still remained open.
From this simple example, it is already easy to see that although the original
entry and exit rules are the same for the two system versions, the stop loss and the
profit target added to the second version make it a completely different system,
with a completely different set of trades. Therefore, when experimenting with a
system, you always have to do the research from scratch for every little change you
make. You still can work according to the general plan outlined by Figures 17.1 to
17.5, but you have to remember that for every little change you do, you might end
up with a completely new set of trades, with their own combined characteristics,
forming the characteristics of the system.
This also hints at another important consideration: A real-life track record is
worth nothing in this case, since you cannot do this research on a set of historical
trades that you might have generated in real life. This is because a lot (if not all)
of those trades would not have happened in the way suggested by the changes you
make during your research. Instead, you only can do this with hypothetically gen-
erated make-believe trades, because that’s the only way you can track and make the
most out of all the changes you make to the system and arrive at the robust solu-
tion you’re looking for.
To add extra reliability and robustness, you could do your testing according
to the principles outlined in Chapter 15, where we looked at a method for testing
the signals and the individual trades rather than the system itself. There, each trade
was traded at random and completely separated from the other trades generated
during the same test run. This can be done without renouncing any of the princi-
ples we’ve learned about in this chapter. We will try to use all that we have learned
so far in this chapter when we look for the optimal stop and exit levels for all the
systems described earlier. First, however, we need to learn how to examine the
data, which we discuss in Chapter 19. But first, let’s talk about why and how to
exit a trade in the first place.CHAPTER 17 Distribution of Trades 199