As it turned out, this was a very interesting piece of research where very few
of the ten different versions of the five systems—to which we just applied the
stops and exits—did better with the intermediate-term filters. Looking through the
results, there are several possible reasons for this.
Reason number one is that none of the filters is an actual trend filter, but
rather a relative strength filter, which means that we will filter away plenty of good
trading opportunities in good trends, only because the stock in question wasn’t
among the very strongest ones at that particular point in time, despite the fact that
it too was in a good trend. This actually means that we will filter away more good
trades than bad trades. Because during times when the trend goes completely
against the anticipated direction of the trade, the trend filter will still allow a trade
to take place, as long as the stock in question is among the strongest ones, no mat-
ter the actual direction of the trend.
Reason number two is that, while each filter might do a good job isolating
the intermediate-term relative strength between the stocks in question, it seems it
isn’t necessarily a good idea to trade the strongest of these stocks in the short term.
In fact, reversing the logic for a few of the filters before trading with a few of the
short-term systems reveals that it actually could be a better idea to trade the weak-
est stock in a group of stocks, as long as it trends in the right direction. The logic
is as follows: If the trend for all stocks is up, those that seem the most likely to
explode to the upside in the short term are those that have lagged behind in the
intermediate term.
Reason number three is that, the more rules we stack on top of each other
within one single strategy, the fewer the markets that will be able to produce a
profit, simply because the added rules mean curve fitting to more specific market
conditions, which fewer and fewer markets will be able to meet. During the test-
ing, this can show up in a higher average profit factor for the filtered version of
the system because it is doing extremely well on a few markets, which even can
boost the average profit factor for all markets—especially if the winning markets
also trade relatively infrequently, so that the gross profits and losses are relatively
small. At the same time, though, the average profit and number of profitable mar-
kets will decrease, because the added rules make an increasing number of markets
unprofitable.
For example, if one market produces a gross profit of $20 and a gross loss of
$10, its profit factor will be 2 (20 / 10), and its total profit and average profit per
trade will both be positive. At the same time, another market produces a gross
profit of $980 and gross loss of $1,020, which results in a profit factor of 0.96
(980 / 1020), and its total profit and average profit will be negative. Looking at the
two markets combined, we see that the average profit factor will be 1.48 [(2
0.96) / 2], suggesting that both markets will trade profitably together. However,
that this will not be the case is revealed by the combined total and average profits,
which will be negative.CHAPTER 21 Systems as Filters 259