Objective research and reviews to aid investing decisions
Reader David Zaitzeff inquired about the predictive powers of bullish and bearish engulfing candlesticks, which he defines as:
Bullish: A down day followed by an up day, with the latter having a higher intraday high and lower intraday low and closing in the top quarter of the daily range.
Bearish: An up day followed by a down day, with the latter having a higher intraday high and lower intraday low and closing in the bottom quarter of the daily range.
We test these signals on the S&P 500 index using average daily returns for each of the 20 trading days immediately after a signal. Using daily high, low and closing levels for the index from January 1962 (the earliest available with intraday data) through June 2007, we find that:
There are 171 bullish engulfing candlesticks and 171 bearish engulfing candlesticks during January 1962 through June 2007. We look first at the bullish set.
The following chart shows the average daily returns for the 20 trading days after the 171 bullish engulfing candlesticks within the entire sample, with one standard deviation error bars. The average daily return for all 11,450 trading days in the sample is 0.03%. While there are a few days out of the 20 with higher than normal returns (Days 1, 2, 8 and 20), there are also days that are lower than normal. There are no strong streaks of abnormal performance. The average of the average returns for the 20 days after bullish engulfing candlesticks is a 0.05%, slightly above normal. As usual with daily returns, the standard deviations are much larger than any abnormalities.
The very large standard deviation for Day 19 is an artifact of the 10/19/87 stock market crash.
To check stability of abnormalities, we look at a subsample of the bullish signals.

The next chart compares the average daily returns for the 20 trading days after the 171 bullish engulfing candlesticks in the entire sample with those after the 91 bullish signals occurring since the beginning of 1990. This chart has no error bars and uses a finer vertical scale than the preceding chart. There are noticeable differences, suggesting noisy data, as does the lack of streaks. From a trading perspective, the diminished Day 1-2 peak may be most notable.

The next chart shows the average daily returns for the 20 trading days after the 171 bearish engulfing candlesticks within the entire sample, with one standard deviation error bars. The average daily return for all 11,450 trading days in the sample, as noted above, is 0.03%. While the first post-signal trading day has lower than normal returns, there are other days that are higher than normal. There are no strong streaks of abnormal performance. The average of the average returns for the 20 days after bullish engulfing candlesticks is a normal 0.03%. The standard deviations are again much larger than any abnormalities.
To check stability of abnormalities, we look at a subsample of the bearish signals.

The final chart compares the average daily returns for the 20 trading days after the 171 bullish engulfing candlesticks in the entire sample with those after the 93 bearish signals occurring since the beginning of 1990. This chart has no error bars and uses a finer vertical scale than the preceding chart. There are noticeable differences, suggesting noisy data, as does the lack of streaks. From a trading perspective, the diminished Day 1 drop is most notable.

In all of the above charts, the abnormalities in daily returns are small compared to standard deviations and small compared to trading costs. They present no tradable streaks. Moreover, the averages of the average daily returns during the 20 days after both bullish and bearish signals are close to normal. And, any tendency of returns to react the day after a signal has weakened in recent years.
In summary, bullish and bearish engulfing candlesticks for the S&P 500 index have no practical value for trading.
For related research, see Blog Synthesis: Some Trading Indicators. See especially our blog entry of 4/18/07 for a different analysis of candlestick signals.