Monday, July 17, 2006

When Failed Patterns Are Good Market Indicators

Many of the measures I follow in the Trading Psychology Weblog each day are reflecting broad selling in the past several days. We recently saw stocks with significant downside momentum outnumber those with significant upside momentum by more than 10:1. Moreover, we've seen more sellers hitting bids than buyers lifting offers in the broad market for three consecutive days. During that time, declining stocks have outnumbered advancers by more than 2:1.

Does that reflect a degree of indiscriminate selling that might be associated with a market bounce once more rational valuations kick in?

I went back to 1990 (N = 4165 trading days) and, to my surprise, only found 41 occasions in which we've had a more lopsided ratio of decliners to advancers over a three-day period. That tells me that the recent selling is relatively extreme.

I looked at all occasions since 1990 in which the number of declining stocks over three days outnumbered advancing issues by more than 2:1 (N = 140). Two days later, the S&P 500 Index was up on average by .49% (93 up, 47 down). That is much stronger than the average two-day gain of .07% (2225 up, 1940 down) for the entire sample.

Interestingly, this pattern has held up during recent market history as well. Since 2004, we have had 27 occasions in which three-day declining stocks outnumbered advancers by more than 2:1. The S&P 500 Index was up two days later by an average of .52% (20 up, 7 down).

It does appear that such broad and concentrated selling yields favorable returns in the short run. That doesn't mean, however, that the pattern is not without its risks. As we were making lows in August, 1990, the pattern yielded two-day losses in excess of 3% before the market righted itself. The pattern also got ugly as we made an important bottom in March, 2004, giving us two-day losses of 3%. In late August, 1998, the pattern would have provided us with a bonecrushing loss of 8% in two days, and in September, 2001 (right after 9/11), we would have had a loss of over 5% in that time.

Note that the times the pattern failed big were relative washout periods in the market that ultimately provided superior buying opportunities. Ironically, when the pattern has failed in the short run, it has provided some of the best long-term returns. When markets go from a selling extreme to an even *greater* selling extreme, that's when you get bottoms of long-term interest.


D TradeIdeas said...


Great article. I reported on the same sense of overreaction but got my cue from Monday's WSJ article in the M&I section. Here's what we wrote

Even if this is indeed a market overreaction to the selling if we are not headed to some serious bear cycles - the likes of which Martin Pring presents here.

Brett Steenbarger, Ph.D. said...

Thanks for the perspective and the links, David. I'm presently working on the (thorny) problem of quantifying overreactions and their subsequent corrections. It's a fascinating area, but not one that's easy to systematize.


Barnabus said...

I thought Wednesday was a 10:1 upside volume vs. downside volume day on the NYSE? Not down over up.

Brett Steenbarger, Ph.D. said...

Hi Barnabus,

I show 7/12, 7/13, and 7/14/06 as consecutive down days in the ES with declines significantly over advances in that 3-day period.


vic said...

Brett: Do you watch 1 tick count on Market Delta to track large trades? If so, it seems to go by so fast that its difficult for me to take a position.

Brett Steenbarger, Ph.D. said...

Hi Vic,

You raise a good point. To truly scalp the market based on trade by trade information, you need a professional order entry system and probably one that is automated or semi-automated. I do track large trades, but do not act on each tick up and down to scalp the market. Mainly the data act as a warning signal if size enters the market against my position.