62 CHAPTER ◆ 5 Types of Trading Systems
such as a maximum VaR or a limitation on currency positions, on the portfolio risk to limit
exposure. Trigger systems very often make use of high leverage.
A trigger signal is based on descriptive statistics of the inputs. To analyze and control
a trigger trade, we apply statistical process control to:● Total returns
● Winners versus losers
● Turnover per month
● Average trade length
● Number of trades over a time period.5.2. Filter Systems
Filter systems select positions based upon a screen, or filter. In such a system, an opening
position may initially be taken based upon a trigger, say the purchase of a bond with cer-
tain duration and credit rating and in a specific industry sector. However, unlike a trigger
trade, a filter trade is dependent on other instruments, positions, or market conditions to
set the filter parameters. So, for example, the filter might be that a bond must yield in the
top 25th percentile of the sector or the universe for a given credit grade. Even though the
parameters for the trade selection appear to be trigger trades, what is different is that in
this case one bond or 30 bonds may be purchased depending on the number that meet the
criteria.
The second step in a filter trade is portfolio management, from basic risk restric-
tions to attribution analysis. A filter system holds the instruments in portfolio until they
no longer meet the initial selection criteria or until they pass some other filter. (There
could be one filter to enter the position and another to exit the position.) Statistical proc-
ess control for a filter trade applies to the percentage of stocks that pass or meet filter
requirements.
A filter is really a hypothesis test between two mean returns, the mean of the sample
selected and the mean of the sample not selected. The normal outputs of a signal trade to
be analyzed and controlled using statistical process control are● Returns, excess returns versus the universe, excess returns versus the benchmark.
● Volatility of the portfolio for the entire period and rolling periods versus the uni-
verse and benchmark.
● Risk return measures such as Sortino and Sharpe ratios.
● Cumulative returns and annualized effective returns.
● Drawdowns.
● Correlation to the benchmark.5.3. Signal Strength Systems
A signal strength, or multifactor, system ranks instruments based on factors or indica-
tors, such as P/E, P/S, or relative strength. Ranking stratifies the instruments into quar-
tiles, deciles, or percentiles, and may be calculated relative to the investable universe or