EVENING STAR PATTERNS
The
evening star is a three bar reversal pattern that may signal the end of a
stock’s uptrend. The first candlestick in the pattern has a long white body,
indicating the current uptrend is continuing. The second candlestick gaps up
and then forms a small body. It is as if the enthusiasm of the initial opening
gap up cannot be maintained, and the distance between the open and close is not
very large. The third candle is a black candle, indicating a reversal of
sentiment; the black candle must close at least halfway down the range of the
first day’s candlestick.
Figure 7.1 shows an evening star pattern in
DIA. The stock was in an uptrend and showed a long white body on 05/09/06. The
following day DIA gapped up and formed a small body, which was followed by a
large black body on 05/11/06 to complete the pattern. After the formation of
the evening star pattern, DIA dropped about five points during the next few
days.
FIGURE 7.1: EVENING STAR PATTERN IN DIA
Courtesy of AIQ
Figure 7.2 shows another evening star
pattern that occurred in MDG on 02/23/07. MDG had been in an uptrend and formed
a large white body on 02/21/07, which was followed by a gap up the following
day and the formation of a small black body. On the third and final day of the
pattern, MDG formed a large black body which marked an end to the recent
uptrend, and the stock dropped nearly 11% during the next four days.
FIGURE 7.2: EVENING STAR PATTERN IN MDG ON
2/23/07
The
definition of the evening star leaves some room for interpretation, and
different traders may not classify some patterns as evening stars. For example,
the definition of the third day of the pattern requires the candle to close at
least halfway down the range of the first day. If the third day gapped down
well below the low of the first day, it would still meet this definition, yet
some traders would not look at the resulting pattern and call it an evening
star.
Some
of the literature on evening stars requires a gap between the first and second
days of the pattern and also a gap between the second and third days of the
pattern. As we’ve seen with the other patterns, variations among the
definitions and some uncertainty of the adjectives sometimes make it difficult
for new traders to understand exactly what to look for. Backtesting not only
gives us an idea of what one might expect in trading the pattern, it also gives
us a tool for exploring how different variations in the parameters of the
pattern’s definition affect the results. Using this technique allows us to test
different definitions and select the one(s) that seem to work best.
INITIAL TEST ON THE
STANDARD ELEMENTS
In
order to backtest the evening star pattern, I started with the definition below
and used a four-day holding period. The initial test period was 01/03/06 to
05/01/07. After a pattern formed that met the definition below, a short
position was entered at the open the following day and held for four days
before being closed.
The
stock must be in an uptrend.
- Day one requires:
- A white body.
- A body larger than
the previous day’s body.
Day
two requires:
- Either color body.
- A body less than 60%
of the previous day’s body.
- A gap up, defined as
the opening being above the previous day’s close.
Day
three requires:
- A black body.
- A close below the
midpoint of the range of day one.
The
initial test results were a little surprising, as shown in Figure 7.3. Taking all of the evening star trades
during the test period showed an annualized ROI with an 18% loss, and less than
46% of the trades were winners. Not exactly an inspiring start; but trading is
about knowing what is, not about hoping what might be. Whether or not a trading
pattern or a filter works, we learn something by knowing what it does, and in
the trading business knowledge can make you money and keep you out of trouble.
FIGURE 7.3: EVENING STAR TEST RESULTS DURING
01/03/06 TO 05/01/07
Testing
the initial evening star pattern in the longer time frame of 01/02/04 to
05/01/07 yielded about the same results, so it is possible that the poor return
is due to the pattern itself and not just a function of one specific time
frame. We can use backtesting techniques to investigate the affects of changing
parameters of each day of the three day evening star pattern.