What Do Normal Stocks Normally Do?

Revision 1.1 ~  2010-05-14 ~ images updated, added new last line to Preface

Disclaimer: I am not, in any way, a statistician nor a "quant". I cannot pretend to be an expert in the way that stocks move or markets function. I do have a background in programming and and visual design. I do like to identify patterns.

Preface

Using Quantum4D I am able to overlay large numbers of stock charts over one another. This is a feature rarely supported by stock market charting software. While investigating the application of these charts and the various capabilities of the software, I seem to be uncovering patterns in markets that I have never seen before. I would very much appreciate your help in identifying the underlying causes of these patterns.

This paper is the first in a series of papers that seek to highlight interesting patterns in stock markets.

This paper examines the price movements of the common stock of 25 major corporations listed on United States stock exchanges.

I look at ticker adjusted closing prices over periods of time. A ticker adjusted closing price is the closing price for the requested day, week, or month, adjusted for all applicable splits and dividend distributions. See http://help.yahoo.com/l/us/yahoo/finance/quotes/quote-12.html.

In order to examine movements, I normalized the stocks. I did this by examining not the values but the daily percentage changes over three different periods of time ranging from a day, to a week and to a month. Also I was able to accentuate movement by normalizing data to the same size box and using a logarithmic scale.

Finally, I am looking for patterns here - not hot stocks (yet), so the display of dates, actual percents, symbol names etc has all been turned off. Note also that I can reproduce similar patterns using wildly varying types of stocks and sectors.

The daily chart compares the percent change in each days price with the previous day's price. The highest peak here is just over 5%.

The weekly chart compares the percent change each days price with the previous week's price. The highest peak here is just over 25%.

The monthly chart compares the percent change in each days price with the previous month's price. The highest peak here is just over 50%.

Perceived Similarities

The daily, weekly and monthly charts all appear to have a similar look and feel. This is a highly subjective judgment. This paper is an attempt justify this assertion.

Some of the sub-patterns I perceive include:

The changes are "pointy". Within any group of outliers, there are a lot of acute angles. If a stock diverges from the average at one interval, at the next interval it will regress to the mean.There are almost no instances of a normal stock being an outlier more that a day or so.

The changes are "collective". If the market is down, it means all the stocks are down. If the market is up, all the stocks are up. Some more than others, but all move in the same direction. 

The changes are "zig-zaggy". If the bunch of stocks are spiking up at one period, the next period they will spike down. Occasionally they stay in a jumble in the middle, but they are more often away from the median.

The background is just as "spiky" as the foreground. If you squint your eye and focus on the black background, it feels just ass spiky as the foreground. In other words, the valleys are quite spiky too. There are not a lot of stocks sitting down in the valleys. The valleys are created by stocks crossing over to the other side.

The overall pattern appears to me to be some kind of regression towards the mean. The stocks diverge from the average and then seem inevitably to revert to the average - or at least pass through it on the way to the other side. Another term that could be applied is "feedback" or perhaps even George Soros' "reflexivity". Or as Wikipedia puts it: "A system prone to hunting (oscillating) is the stock market."

What is remarkable is that one sees the same pattern recurring on a daily weekly and monthly basis. As we zoom in, we see the same amount of detail. There is a feeling of fractals or Mandelbrot Sets. As defined in Wikipedia: A fractal is 'a rough or fragmented geometric shape  that can be split into parts, each of which is (at least approximately) a reduced-size copy of the whole,' a property called self-similarity."

Similarities to Other Data

What other data could I compare this too? Maybe lots of other data also behave this way.

How do these charts relate to overbought/oversold? The terms generally, but certainly not always, refer to individual stocks. Occasionally, pundits will declare that "the market is oversold", but a quick overview indicates that the pundits are typically referring to individual equities.

How do these charts relate to Wilder's Relative Strength Index (RSI)? Again the RSI is usually applied to individual stocks and not to groups of stocks.

Predictions

Please remember that all the charts are based on individual days of data. The day chart compares the current day with the previous day. The weekly chart compares the current day with the close of seven days previously. The monthly chart compares the current day with the the close and the end of the trading day thirty days previously.

Thus a stock that was a huge outlier of say, ten percent or more is likely to be an outlier in a day, week or month.

Outliers are outliers for only short amounts of time. Outliers start near the middle of the pack. They become outliers for a day, then they come back to the middle of the fold.

Outliers do not buck the trend. If the market is down, the outliers are further down, If the market is up, the outliers are up.

Conclusions

This is not the end. It's the beginning.

Using Quantum4D I can now overlay and display in a visual manner the activities of dozens even hundreds of stocks. What patterns can we uncover that we've never been able to see using traditional charting software?

I would very much appreciate your advice as to references of other studies of this kind. Who else is doing this kind of research and what are they finding?

I would like to know what questions you think I should be asking. How can we use these data sets in order to uncover some predictably trade-able opportunities?