Thursday, August 10, 2006

Double Outside Days: What Happens Next?

As a kind reader alerted me, we had a rare situation on Wednesday in which the S&P 500 Index (SPY) made its second consecutive outside day. This means that we had a higher high each day and a lower low. What is equally interesting from my vantage point, and as the Trading Psychology Weblog figures note, is that we had more new 20-day highs each day *and* more 20-day new lows. Clearly there's a tug of war between bulls and bears in this market.

Going back to 1996 (N = 2637 trading days), I could only find 8 examples of back-to-back outside days in SPY. Three days after the double outside days, the market was up 6 times, down 2, for an average gain of 1.33%. That's much stronger than the average three-day gain of .11% for the entire sample. While this is too small a group of occurrences to stand on statistically, it's clear that bulls have tended to win the tug of war in the past when we've had double outside days.

Overall since 1996, we've had 286 outside days in SPY. Interestingly, on the next day, SPY has been down on average -.03% (155 up, 131 down), certainly no edge relative to the overall sample. By three days out, however, SPY is up on average by .23% (171 up, 115 down), which is again stronger than the average three-day change for the entire sample (.11%; 1431 up, 1206 down).

Perhaps the most relevant and interesting finding is that, when the market is down after an outside day (N = 131), the drop tends to be reversed in a majority of instances over the next several days. Four days later, for instance, SPY is up by an average of .66% (81 up, 50 down), which is stronger than the market's average four-day gain of .14% (1441 up, 1196 down).

In short, double outside days are rare, but have tended to lead to bullish outcomes. Outside days have tended to be bullish several days out and particularly if the day after an outside day is down. The market is steeply lower overnight as I write this and we've broken below an intermediate-term trading range, as the Weblog noted last night. Let's see if the breakout pattern overwhelms the outside day pattern.