Wednesday, December 20, 2006

SPY and QQQQ: Once It's A Trend, Will It End?

If you're looking at daily data, the shortest "trend" you can identify is a series of two up days or a sequence of two down days. So I decided to look at what happens in the S&P 500 Index (SPY) after two up days and after two down ones.

Since 2004 (N = 745 trading days), we've had 233 occasions in which SPY has been up for two consecutive days. Three days later, SPY has averaged a loss of -.02% (119 up, 114 down)--no bullish edge at all. Conversely, across the remaining occasions in the sample, the average three-day gain in SPY has been .16% (301 up, 211 down). It thus appears that, once we have a two-day bullish "trend", market returns have been subnormal.

When we've had two consecutive down days in SPY (N = 146), SPY has averaged a three-day gain of .25% (89 up, 57 down). Over the remaining occasions in the sample, SPY has been up by .07% over a three-day period (331 up, 268 down). Here we see that, once we have a two-day bearish "trend", near-term returns have been superior.

If we look at the NASDAQ 100 Index (QQQQ), a similar pattern appears. When we've had two consecutive up days (N = 219), QQQQ has been down by an average of -.01% over the next three days (118 up, 101 down). Once again, no evident bullish edge. Over the remainder of the occasions in the sample, QQQQ has been up by an average of .12% over the next three days (284 up, 202 down).

But wait! When QQQQ has been down for two consecutive days (N = 159), the next three days in QQQQ have averaged a gain of .07% (83 up, 76 down). Over the remainder of occasions in the sample, the next three days have averaged a gain of .09% (319 up, 267 down).

Interestingly, we see no evidence of a reversal effect for QQQQ after two down days and a relatively modest effect after two down days. In the S&P 500 Index (SPY), there's a more pronounced tendency toward reversal once a two-day "trend" appears.

Might we be able to classify ETFs and individual stocks based on their simple trending or countertrend patterns? Would such patterns provide useful trading guidance? An interesting post from James Altucher's StockPickr site suggests that similar patterns may indeed guide active trading in the NASDAQ 100 stocks. I'll add a twist to the pattern in my next post.


Anonymous said...

james atucher's system , when i went to the sight doesn't seem to have a money mgt flaw. could be a real flaw in that system i think

Brett Steenbarger, Ph.D. said...

Hi Dan,

Yes, I noticed that also. In a market that picks up volatility, you could get some drawdowns. That having been said, it's amazing to actually test out some of the exit strategies that people bandy about and see how negatively they affect trading performance for a given pattern. Thanks for the observation--


Michele said...

This is very interesting. I have an automatic system that trades ES intraday as follows: it buys if it sees two consecutive bars with higher closes, and sells when it sees a close <= that of the previous bar. (Inverse for going short). This has the expected results. What it's missing is information on current trendiness (and even better, future trendiness). This would help prevent the program from getting whipsawed. I think I'll try computing some statistics on intraday ES trend lengths. It would be interesting to see how that correlates to your numbers for SPY and the Q's. In other words, does trendiness have a fractal component?

Brett Steenbarger, Ph.D. said...

Hi Michele,

So *you're* the one on the other end of my trades! :-)

Seriously, though, your basic trading idea will work in markets that have sufficient volume and volatility, with good participation from the institutions. In slower markets dominated by the locals, you'll get chopped up. (Today's market--12/20/06--was a good example in the AM).

I suspect if you consider a volume filter on your basic patterns, you may avoid some of those whipsaws.

Thanks for sharing that perspective. I find it fascinating to see the different ways people trade--


Anonymous said...

Dr. Steenbarger,

Your post reminded me of a trading system I came across at called Smitlener Streaks.

I ran some tests using SPY only, QQQQ only, and SPY & QQQQ Combined and posted the results on my blog at if you are interested.

Thanks for all of your great work!

Brett Steenbarger, Ph.D. said...

Thanks, Dogwood, for conducting those very informative tests. I'll link your post on my personal site tonight and look forward to following your blog--