Saturday, September 23, 2006

Tracking Significant Market Days


When analyzing market patterns, are all data points equally meaningful? I've begun looking at market indicators in a different way, eliminating all neutral values and instead focusing only on extremes.

Above we have a chart of the S&P 500 Index ($SPX) from January, 2005 to September 21, 2006. The red line is the closing daily price of the index. The blue line is a 20-day moving average of the net number of "significant" market days in SPX. A significant day is defined as one that is one standard deviation or more greater than or less than the median price change over the past 65 trading sessions. The net significant number of days is simply the number of large up days over the past 20 sessions minus the number of large down days.

As markets top out--and often when they bottom--they produce fewer significant days. The net number of significant days very frequently tops out ahead of short-term market price highs. A sharp shift from net negative to net positive significant days is common during the transition from market bottoms to new short-term bull phases.

Note how our net number of significant days topped out well ahead of the recent market.

Clearly, it will take further investigation to establish the value of tracking significant days. The idea behind the measure is applicable to any stock, commodity, or index--and indeed could be applied to such indicators as advances/declines, TICK, and TRIN. The basic idea is that, just as the largest traders determine market order flow and short-term market movement, the largest market days establish the overall market trend. I will report further on these significance indicators in the Weblog.

2 comments:

yinTrader said...

Hi Brett

Your chart reminds me of the Buy/Sell Confidence Index from Sentimentrader.

Their buy/sell sentiment is represented by say Smart Money 42%, Dumb Money 50% for the past few days for S&P.

Do you see any correlation of these two charts?

Brett Steenbarger, Ph.D. said...

Hi,

That's an interesting question, and not one I've considered. I'd have to track that correlation over a while to see if there's something to it. Thanks--

Brett