$100,000, which is just above the lower one-standard-deviation boundary, there is
nothing wrong with the system. It just so happened that you caught a sequence of
trades with a lower average profit per trade than could have been expected from
the historical estimate, based on 20 such sequences.
In fact, more than every third trading sequence will produce a profit between
the average expected profit and its lower one-standard-deviation boundary, and
about every fifth trading sequence will produce a profit that is lower still. And
every so often (perhaps every 50th sequence or so) there will even be a loss. Also,
the fewer the trades used to calculate the average profit per trade, the more likely
it is that that average is negative. Figure 22.5 shows that still plenty of trading
sequences are left in negative territory after 50 trades or so, and that it takes some
120 trades before the last sequence turns positive. The longer the sequence, how-
ever, the closer each sequence will be to the estimated average profit.
Similarly, a system that produces better than expected results (as indicated by
the estimated average profit per trade) still works as it should: you just happened
to luck out with a couple of extraordinarily good trades over a period. The most
dangerous thing you could do at this point is to increase your positions in the belief
that the system has become more robust and reliable. Look at the two sequences
CHAPTER 22 Variables 279
FIGURE 22.5
Loss of close to $75,000.