Obviously, the data providers behind the data in Figures 6.2 and 6.3 thought
that they were doing me a favor by not letting me believe that Microsoft traded at
the impossible price of $4.76, so that I wouldn’t go and build myself a strategy that
tried to trade Microsoft at that price. Therefore, in an effort to make things as real-
istic as possible, they adjusted the price to the closest tradable fraction. But this is
all wrong and backwards thinking. Because no matter at what level Microsoft
actually traded several years back, or no matter what the true split-adjusted price
is (or the split-adjusted price that these data providers provide me with), I can’t
turn back time and go trade Microsoft at any of these prices anyway. What inter-
ests me is to see if I can come up with a way to catch the wiggles that the Microsoft
stock created at that time and that it continues to produce today.
All I am interested in, as an analyst, is examining if my system in fact man-
ages to catch the wiggles and squiggles as they actually took place. For this, it
doesn’t matter what the actual price of the stock is pre- and post-splits. What is
important is that the relationship between price points stays the same. Clearly,
they do not in Figures 6.2 and 6.3. Therefore, these data are useless for historical
testing.
We are not interested in “realistic” back testing in dollar terms, because we
can’t trade at those dollar values anyway, no matter how much we want to, because
they represent foregone opportunities. The best we can do is use data in such a way
so that we make our back-tested results as robust and forward-looking as possible.
And the way to do that, as we already know, is to work with percentage moves and
normalized results.
With percentage moves, the distance between point A and point B relates to
point A exactly the same, no matter the price of point A. Look at the two charts of
Microsoft, by Microsoft (Figures 6.1 and 6.5). They are identical, except for the
difference in price. The price differences are because of stock splits, but no matter
the price before and after the split, each price level relates to all the other levels
the same way in both charts. Thus, in the split-adjusted chart, the price might seem
to have been low, but it behaves as it did when it was higher.
That last sentence is very important to understand completely and fully,
because in Part 3, we talk about the various methods for placing stops and exits.
For now, let’s just mention briefly that many system builders think that the lower-
priced the stock, one should work with wider stops, relative to the entry price.
They believe that lower-priced stocks are more volatile than higher-priced stocks,
and therefore they need a little extra leeway. There might be something to that rea-
soning, but for system testing purposes, it is no good, because, as we have just dis-
covered, even though a stock seems to be priced as low as 20 cents, it still can act
as if it were trading at $80, because this was the actual level at which the trading
took place in the first place, with the 20-cent level implemented long after the fact.
Therefore, no matter the price of the market, each stop must relate to its entry
price exactly the same way as all the other stops, no matter which entry price each70 PART 1 How to Evaluate a System