in Figure 22.5 that have the best results after some 110 trades. They both start to
move downward from there to get closer to the estimated average profit over 200
trades. What if you had decided to double your positions for these trading
sequences after 110 trades? The drawdowns would have been twice as steep as
they now appear in Figure 22.5.
So the question still remains: If the final outcome for one trading sequence
can vary this much with the system still working as it should, how do we decide
when a system’s inherent characteristics change to the better or worse? (A system
changing to the better, or seemingly better, without us knowing is not a good thing
either. The very same reasons that now make us richer than trolls could be what
ruin us in the end, because we haven’t adjusted the position size of the trades to
accommodate for the new circumstances.)
This is a very difficult question to answer, and I feel I really don’t have the
experience to do it. Nonetheless, I would say that if a system performs way out of
what could be expected (say more than two standard deviations away from the esti-
mated mean) over at least two reasonable, long sequences of trades, it’s reasonable
to assume that something has changed so that the estimated statistics can no longer
be trusted. It may be that you did something wrong from the beginning when con-
structing the system, or the assumptions on which the system is based no longer
are valid.
Unfortunately, however, I have a gut feeling that most systems get discarded
way before this happens and while the system is still working as it should, simply
because the trader had no idea that the system could produce the type of drawdown
numbers it eventually did. Hence, he starts trading it way undercapitalized.
Analyzing the systems for consistency could help you get a better feel for what
types of drawdowns your systems are most likely to generate, given the statistical
distribution of the trades. With this information at hand, you should not have to be
forced out of the game, despite a good system but because of lack of capital.280 PART 3 Stops, Filters, and Exits