itable months is more important than a high number of profitable trades, and
sometimes it might be a good idea to take a higher number of losing trades to
increase the number of profitable months. This is especially true when we need to
free up capital from trades that are going nowhere.
Sometimes, when testing the same system on several markets, or the same
market with several systems, the average net profit from all trading sequences (cell
E70) can be negative despite a positive average profit per trade (cell F70), or vice
versa. Because the number of trades can vary within each trading sequence, a trad-
ing sequence with a high number of relatively small losses will weigh down the
average net profit relatively more than the average profit per trade. A trading
sequence with a low number of relatively large winners, on the other hand, will
increase the value of the average profit per trade relatively more than the average
net profit.
However, because we don’t know if a market that produced relatively few or
many trades over the testing period will continue to do so in the future, we cannot
place any significance on the net profit produced by that market. All we can do is
use the average profit per trade, regardless of how many trades it is based on, as
one of many estimates for the average profit per trade for all markets. We use the
average of all averages from all markets as an estimate for the true, but always
unknown, average profit per trade.
Another way to look at it is to assume that each market is producing approx-
imately the same number of trades, no matter how many trades there are behind a
certain average profit per trade for a certain market. In that case, a positive (nega-
tive) average profit per trade will also produce a positive (negative) average net
profit per market. This is also more in line with what we can expect in real-life trad-
ing. One error we deliberately make during this stage of the testing is to equalize
one trading sequence with trading one market over a longer period of time (say 10
years). In real life, one trading sequence is instead equal to trading several markets
(10, for example) over a shorter period of time (say one year). In this case, differ-
ent markets will produce a varying number of trades over different periods, but the
number of trades over all trading sequences (in this case defined by units of time,
rather than specific markets) is more likely to stay approximately the same.
Cell G71 shows the average standard deviation for the profit per trade pro-
duced by each market. The value, 8,337.05, is calculated using the formula
AVERAGE(G$2:G66), where column G refers to the standard deviation of prof-
its produced by each market. It indicates that 68 percent of all trades within one
trading sequence will result in a profit ranging from $7,042.01 (1,295.04
8,337.05) to $9,632.09 (1,295.04 8,337.05), which is further indicated by cells
G72 (F70 G71) and G73 (F70 G71).
Note that the values in cells G71 to G73 are much larger (positive or nega-
tive) that those in cells F71 to F73. This is because the values in cells G71 to G73
give an indication of how much the results per trade can vary within one market,CHAPTER 7 TradeStation Coding 93