Saturday, March 25, 2006

Countertrend Equivalence - An Intermediate-Term Look



My recent Trading Markets article, along with recent blog entries here, found evidence of countertrend equivalence on a 5 day basis and, to a more modest degree, over a 5 hour timeframe. Recall that the idea of countertrend equivalence is that, if the market establishes a strong trend over X period, the next X period will tend to reverse this trend.

I decided to extend the analysis by looking at 5 week periods in SPY. Since March, 2003 (N = 155) we have had 27 strong uptrending periods on my trend measure. Five weeks later, SPY averaged a gain of .67% (16 up, 11 down), weaker than the average five-week gain of 1.36% (103 up, 52 down).

To create a relative match, I looked at the 25 strongest downtrending periods in SPY during that same time. Five weeks later, SPY averaged a gain of 2.36% (18 up, 7 down)--much stronger than normal.

Once again, we see evidence of countertrending, with moves over one period reversed in the next. The effect is especially strong for reversals of downtrends on a five day and five week basis, suggesting that these timeframes might be worth coordinating for intermediate-term trades.

4 comments:

John Wheatcroft said...

This is something I've kind of "known" for awhile - several posts ago I even mentioned that you should "skip a period" and check results that way. But it is always nice to have empirical evidence.

My question now is - why "5"? First how did you arrive at 5 and second why is "5" the magic number? Why not 12 or 13 or 33?

Brett Steenbarger, Ph.D. said...

Hi John,

No reason for the 5--it was the number I started with, so I stuck with it. I didn't try to pick periods to optimize results.

Brett

D TradeIdeas said...

Some have argued a strategy (and we have modeled with success as measured by popularity) that considers 10-day periods but for different reasons: sector rotation. The strategy looks for stocks hitting 10-day highs as a sign of fresh money entering a sector. The opposite is true for 10-day lows. This would seem to counter against the countertrend phenomenon you describe. Is institutional money so fickle as to enter and then leave thus creating the behavior you describe? I just don't know the answer.

Brett Steenbarger, Ph.D. said...

Hi,

I'll check out 10 day horizons later this week and see what I find. It would amaze me to find general evidence of trending, but it is certainly possible that individual sectors might trend better than the S&P. There *is* evidence of upside trending on broad market momentum, as I've modeled with my Demand/Supply index, but the key there is that many stocks have to be making those new highs simulataneously for upside momentum to continue.

Brett