Friday, October 19, 2007

Five-Day Price Relationships in the Stock Market

In a recent post, we took a look at simple 20-day price relationships and what they told us about future price changes. In this post, we'll move to a five-day basis and see if we can learn anything from similar price relationships.

As before, I went back to 1990 (N = 4482) and investigated five-day new highs and new lows in the S&P 500 cash index ($SPX).

When we have made a five-day high in the S&P 500 Index (N = 1381), the next five days in $SPX have averaged a gain of .04% (740 up, 641 down). When we've made neither a five-day new high nor a five-day new low (N = 2081), the next five days in $SPX have averaged a gain of .14% (1166 up, 915 down). When we've made a five-day low in $SPX (N = 1020), the next five days have averaged a gain of .48% (611 up, 409 down).

Once again, as with the 20-day data, we see subnormal returns following five-day highs and above average returns following five-day lows.

Interestingly, when the S&P 500 Index makes a five-day high *and* it closes above its 50-day moving average (N = 1145), the next five days in $SPX average a gain of .02% (610 up, 535 down). When the S&P 500 Index makes a five-day high and closes below its 50-day average (N = 236), the next five days in $SPX average a gain of .12% (130 up, 106 down).

It thus appears that returns are lowest when we make a five-day closing high in a market that is already extended to the upside.

When the S&P 500 Index makes a five-day low *and* it closes above its 50-day moving average (N = 466), the next five days in $SPX average a gain of .37% (284 up, 182 down). When the S&P 500 Index makes a five-day low and closes below its 50-day moving average (N = 554), the next five days in $SPX average a gain of .58% (327 up, 227 down).

As with the 20-day data, we see superior returns whenever a five-day low is made, with the best returns coming following five-day lows in markets that have been extended to the downside--a clear reversal effect.

These five-day findings largely confirm observations made by Larry Connors and Connor Sen in their book "How Markets Really Work". The test of any market indicator that is not purely price-based is whether or not it adds value to an analysis of price relationships alone. That will be the topic of follow-up posts on this theme.

RELATED POST:

The Trend You See Is Not What You Get
.

4 comments:

Paolo Pezzutti said...

Brett,
I find this quite interesting. It is true that so far contrarian techniques applied to the up side have worked particularly well. Probably this repressents the basic strategy of strong long players in the marketplace that whenever the market prints new lows accumulate long positions.
That is ok for the past. How can we monitor whether cycles are changing, whether these big players cease to look at the market as continuously going up.
How long it takes before we realize that this basic strtegy is not applied any more. How do you select the data and how back in time you test the market's behavior to identify a change in investors' approach and the end of a market inefficiency

Paolo Pezzutti

Anatrader said...

Brett

With the 20th anniversary of the 1987 crash, would you pay heed to the Hindenberg Omen ?

Every NYSE crash since 1985 has been preceded by a Hindenberg Omen.

Brett Steenbarger, Ph.D. said...

Hi Paolo,

I generally look over a several year horizon to identify trading regimes. Those regimes are most likely to shift when we also see shifts in related markets (currencies, rates, etc). It helps to look at the most recent instances of a pattern to detect any degradation of the value of the signals.

Brett

Brett Steenbarger, Ph.D. said...

Hi AnaTrader,

Yes, it makes sense that a crash would be preceded by a surplus of new lows over new highs. The question is how often such a surplus *doesn't* lead to a crash.

Brett