Exit Strategies (Part 1)

Introduction

For most non-discretionary traders, trade entries are always predicated by a strict set of rules. When these rules are
met, the trade is deemed "triggered" and the position is opened. Conversely, when the stars are not aligned, the
systematic trader simply sits on his hands and does nothing. No trigger, no trade.

Interestingly, while strict with their entry rules, even seasoned traders tend to exhibit considerable more flexibility
when it comes to their exit rules, and complacency with sticking to them. The reason is simple: when entering a
position the trader's P&L is de-facto neutral, and so is his emotional investment in the trade. But once the trade is in
play, both rising unrealized profits or rising unrealized losses will tend to play on the trader's emotions and incite
him to "manage" the trade, all too often to the detriment of the system's long-term profit expectancy.

In this study we will use a simple mean-reversion system and look at several exit strategies and analyse their
respective merits. The objective here is to illustrate how quantitative analysis can be used to objectively determine
the best exit rules for a specific system and, with that knowledge, motivate traders to stick to both their entry and
their exit rules.

Analysis

The instrument used in this study is the INX (S&P500 Index), from Jan 1st 1991 to Dec 31st 2015. Since the INX is
not a tradable financial instrument, actual positions would have been taken either with futures (ES) or with ETFs
(SPY or IVV).

System rules:

 Entry: go long on the lowest close in 10 days

System data:

 Instrument: INX (S&P500 Index), from Jan 1st 1991 to Dec 31st 2015 (25 years)
 Initial capital: US$ 100,000
 Profits not reinvested
 No allowance for commissions or slippage

Fixed Target and Fixed Stop

In this first example we will place a symmetrical bracket order to dictate both the profit target and the stop loss. So
the trade will be exited either at an X% profit or at an X% loss.

Table 1

1
Copyright 2016
www.TheMechanicalTrader.com
As we can see in Table 1, the system would have been profitable throughout the 1%-10% bracket order range. We
notice that, as the size of the bracket increases, so does the system's overall net profit, average trade profit and
profit factor. The system's win rate also increases, albeit to a lesser degree.

There are however some serious issues with this system. Firstly, by setting fixed targets for both profits and losses
the system caps each trade's maximum profit potential, meaning that it doesn't allow profits to run. This results in
both small average trade profit numbers and small profit factor numbers. Moreover, achieving an even modest profit
factor of at least 1.5 would have required a profit-target and stop-loss figure of 7% or higher. This is a large risk to
take for any one trade, and back-to-back losses would have compounded to a huge maximum drawdown of up to
82%, as seen in the red column above. This would have had a catastrophic impact in 2002 and later in 2008, as
shown in the equity curve below (using a 7% target/stop).

Time-Based Target and Time-Based Stop

In this second example we will let a specific number of days dictate both the profit target and the stop loss of our
mean-reversion system. So the trade will be exited, at a profit or a loss, after exactly X days.

As we can see in Table 2, this simple time-based exit strategy has some interesting benefits. First of all it is
extremely easy to implement: the trade is entered into and a conditional sell order is placed to trigger after a set
number of days. That's it. Secondly, swing-trade systems tend to have a "sweet-spot": a window of days where the
system's edge is at its peak. Before this sweet spot, there are still potential profits left in the trade; after it, the edge
that triggered the trade has essentially disappeared and it is probably time to exit the position. This concept is
discussed in some detail in the "Swing Trade Cycle Analysis" paper published on our website. In this example, the
sweet-spot appears to be somewhere between day 7 and day 10. This is the compromise place where net profit, win
rate and profit factor numbers are at relative highs, and maximum drawdown numbers are at relative lows.

It is quite surprising to see just how effective a simple mean-reversion system like this can be when coupled with a
time-based exit. There are, however, some major caveats. Specifically, a lot can happen in X days. The system could
have caught the beginning of a multi-day market crash and the time-based rule might just exit the trade at the very
bottom of the market. And this market bottom, also a 10 day low, would trigger a new trade which could also
continue running south, further compounding losses. This issue is seen in the high drawdown numbers of the full 5
to 15 day exit rule range. The use of a 10 day exit rule, for example, would have resulted in a maximum drawdown
of 34.5%. Not nearly as severe as in our previous example, but nevertheless enough to put a large dent in a trader's
account and possibly an even larger dent in his confidence in the viability of the system.

2
Copyright 2016
www.TheMechanicalTrader.com
Table 2

The problem of large drawdowns is particularly noticeable in the equity curve shown below, which uses a 10 day exit
rule. The steady upward-sloping chart indicates that the system's edge was strong and persistent throughout the 25
year period under study. The choppy nature of the equity curve, however, would have tested the resolve of most
traders.

Floating Target and Fixed Stop

In this third example we will use a floating target coupled with a fixed stop. A "floating" or "dynamic" target is an
exit rule that changes as a function of time and price. Floating targets are typical of mean-reversion systems and
come in several flavours. In this example the stop loss will be set to range from 1% to 10%.

As we can see in Table 3 the results are fairly good throughout the 1%-10% stop range. The floating target and the
absence of a timed exit did a good job at allowing trades to resolve themselves profitably. From a 4% stop and
upwards, win rates are above 70%, much higher than in the previous two example systems.

3
Copyright 2016
www.TheMechanicalTrader.com
The highest net profit figure would theoretically have been achieved using a 1% stop, but at the expense of a low
win rate and an even lower average profit per trade figure. This is a problem as real-world commission costs and
slippage risks might well have rendered this system unprofitable. A better stop value would probably have been 5%
or 6%, which would have yielded higher average profit per trade, win rate and profit factor.

The major downside of this system is, once again, potential maximum drawdown: almost 30% using a 5% hard
stop. As in our first example, fixed stops mean that high unrealized losses often become realized, sometimes
preventing the trade to otherwise turn profitable, particularly in volatile environments. Stops hit in succession would
have resulted in some steep losses, as seen in the choppy nature of the equity curve below.

Table 3

Equity curve for this system, using a 5% stop-loss exit rule:

Floating Target and Time-Based Stop

In this fourth example we will use the same floating target used in the previous study but this time coupled with a
time-based stop. So if the dynamic target is not hit within X days, the trade will automatically be exited.

Table 4 shows the results of this system. As we can see, all key performance figures (net profit, win rate and profit
factor) increase in a fairly linear manner as the number of days increase. These results also help us pin-point the
new system's sweet-spot, which in this case happens to be day 7. This is where all performance values reach their
peak and beyond which holding the position - on average - becomes a 50/50 proposition. These findings are very

4
Copyright 2016
www.TheMechanicalTrader.com
much in line with swing-trading theory which sees the bulk of mean-reversion edges expressing themselves within 1
to 3 weeks, rarely beyond.

Table 4

Below is the equity curve for this system, using a 7 day exit rule. We notice a strong, persistent upside sloping chart
with only some chop during the bear years of 2001-2003. The system would have incurred its maximum drawdown
of 23.3% during the market collapse of 2008, but would have quickly recovered.

Floating Target and Stop

In this fifth and final study we will use the same floating target as above to dictate both our target and our stop. So
essentially the system has no stop, and simply allows time and price-action to exit the trade.

Table 5

5
Copyright 2016
www.TheMechanicalTrader.com
As we can see, the results are identical to those shown in the last row of Table 4. The system allows the mean-
reversion cycle to take its course to completion, well beyond its sweet-spot. Results are overall rather good, but the
system would have underperformed somewhat during strongly bearish and volatile times, particularly during 2000-
2003.

Below is the equity curve:

Summary

While our exact findings are specific to the test system analyzed in this paper, they are undoubtedly relevant to the
majority of long-side swing-trading systems:

 Bracket exits are often favoured by short-term investors because they appear to provide a balanced
risk/reward proposition. But fixed profit targets and stop losses are problematic because they do not
successfully manage downside risk. Moreover, since they intrinsically limit the profit potential of the trade,
they negatively impact the overall long-term profit expectancy of the system.
 Simple time-based exit strategies can be surprisingly effective and are very easy to put in place. They key
here is to understand the system's natural "cycle" and to select a timed exit that captures the bulk of the
system's profit potential. All swing-trading systems have their respective "sweet-spots", so it is essential to
identify the one for each system. Note however that time-based exit strategies sometimes have the
unfortunate tendency to exit positions at market bottoms, so a number of back-to-back losing positions
could result in large cumulative losses.
 Using floating targets is probably the best way to capture the maximum profit potential of a swing-trading
system. Using hard stops, however, is likely to reduce the effectiveness of floating targets due to the
likelihood of consecutive losing trades and the resulting risk of high drawdown.
 The use of floating targets and stops is an elegant and simple way of achieving strong and consistent
system performance, while limiting downside risk. Adding a time-based exit has the additional benefit of
exploiting the system's "natural" cycle. Using quantitative analysis can help determine the system's sweet-
spot, and hence the culmination of its trading edge.

6
Copyright 2016
www.TheMechanicalTrader.com