Out-of-Sample Test of Trading Streak Reversals
Posted in Technical Trading
May 13, 2010
The article “If a Stock Keeps Moving Higher, Should You Buy It or Sell It?” cites past research, based on stock market index data from 1989-2003, finding that it is better to be a seller (buyer) after the market has been strong (weak) as indicated by multiple days of higher intraday highs (lower intraday lows). It also reports results from seven million backtested trades from January 1995 through June 2006 that indicate a similar conclusion for individual stocks. The article does not report variability in results. Does the original conclusion for indexes hold up out of sample based on closes (rather than intraday highs)? Using daily closes for the S&P 500 Index and the NASDAQ Composite Index for January 2004 through April 2010, we find that:
Closing levels of an index have fixed measurement intervals using points at which investors/traders in underlying stocks decide whether or not to accept the risk of holding overnight. Intraday highs and lows have variable intervals (for example, consecutive intraday highs could occur at the close on one day and the open on the next). Also, up and down streaks based on intraday highs and lows could overlap when volatility is increasing, thereby confounding trading signals.
Assumptions for testing here are:
- Measure returns one, two and five trading days after streaks.
- Perform return calculations based on index closes coincident with streak tabulations. In other words, determine streak status just before the close so that an associated trade takes place at the same close.
- Ignore trading friction (at least in detail).
The following pair of tables provides the number of instances, the average future returns over the next one, two and five trading days and the standard deviation of these returns for streaks in the S&P 500 Index over the entire sample period. The first (second) table is for upward (downward) streaks. There is one instance of an “8 Down” streak not shown in the second table, because it is the same streak as “7 Down” and not separately tradable.
Results are mixed across streak lengths and return horizons, making selection of specific trading rules problematic. Average returns are mostly small compared to standard deviations and to reasonable trading friction.
The visualization of variabilities below offers another perspective.
The following chart shows the average S&P 500 Index two-day future returns after streaks occurring at least five times during the sample period, with one standard deviation variability ranges. Any tendencies are small compared to variabilities and especially inconsistent for down streaks.
Does behavior of the NASDAQ Composite Index after streaks look similar?
The next pair of tables provides the number of instances, the average future returns over the next one, two and five trading days and the standard deviation of these returns for streaks in the NASDAQ Composite Index over the entire sample period. The first (second) table is for upward (downward) streaks. There is one instance each of “9 Up”, “10 Up”, “11 Up” and “12 Up” streaks not shown in the first table, because all come from the same streak as “8 Up” and are not separately tradable.
Results are again mixed across streak lengths and return horizons, making selection of specific trading rules problematic. Average returns are mostly small compared to standard deviations and to reasonable trading friction.
The visualization below facilitates a consistency check across indexes.
The final chart compares the average two-day future returns after various streaks in the S&P 500 Index and the NASDAQ Composite Index over the entire sample period. Inconsistencies undermine belief in any reliable market reaction to streaks.
In summary, evidence from simple tests does not support a belief that equity markets react reliably after up or down streaks in closing levels.
