testing, how much the price might fluctuate for a specific market, or how many
markets you’re testing the system on. But—and this is the important part—instead
of measuring the results in dollars won or lost, the outcome of each trade should
be measured in percentage terms in relation to the entry price of the trade.
The second way is to vary the number of shares or contracts traded in such a
way that the number of shares increases as the price of the stock or the futures con-
tract decreases. For example, if a stock is priced first at $100 and then $50, you
should test the system with twice as many shares when the stock is priced at $50
as compared to when it was priced at $100. You should do this no matter if that’s
what you would have done in real life or not. Using this technique, the results can
still be measured in dollars.
To illustrate why all this is important, let’s look at two examples, both of
them in the form of questions. (Hint: before you answer the questions, go back to
the beginning of this part and re-read the story about the two cubicle buddies and
“Tom’s unfortunate stock purchase.”)
Say that stock ABC currently is trading at $80, and a trading system that con-
sistently buys and sells 100 stocks per trade shows a historical, back-tested profit
of $250,000 over 500 trades, for an average profit per trade of $500 and an aver-
age profit per share of $5. These results are to be compared to those of stock XYZ,
with a back-tested profit of $125,000, also over 500 trades, for an average profit
per trade of $250 and an average profit per share of $2.50. Over the entire testing
period, the price of XYZ has more or less constantly been half that of stock ABC.
Now, everything else aside, do the above numbers indicate that stock ABC is a bet-
ter stock to trade with this system than stock XYZ?
(This reasoning can also be translated over to a stock split. For example,
say that stock QRS is trading at $90, and a trading system that consistently buys
and sells 100 stocks per trade shows a historical, hypothetically back-tested
profit of $150,000. Tomorrow, after the stock has been split 3:1 and the stock is
trading at $30, the historical, hypothetically back-tested profit has decreased to
$50,000. Does this mean that the system suddenly is three times as bad as the
day before?)
No, it does not: A stock that is priced at twice the value of another stock also
can be expected to have twice as large price swings, and therefore twice as high a
profit per share traded. If you look at the above numbers, you will see that a prof-
it per share of $2.50 relates to a stock price of $40 in the same way that a profit
per share of $5 relates to a stock price of $80 (2.5 / 40 5 / 80 0.0625
6.25%). That is, in percentage terms, the profit per share is the same for both
stocks. (As for the stock-split example, it is easy to see that after the split, we will
need to trade the stock in lots of 300 shares per trade to make the new results com-
parable to the old presplit results.)
To make the same net profit trading stock XYZ as trading stock ABC, all you
need to do is to trade twice as many XYZ stocks in each trade as you would stock8 PART 1 How to Evaluate a System