Volume Based Indicator Analysis Part 3

This post continues my analysis of volume-based stock market indicators.

If you recall, in my previous posts I formed a weighted average of (daily volume * percent price change from previous day).  In this post I’m going to analyze a similar indicator, but instead use a weighted average of just (daily volume).  I.e., the percent price change from the previous day will not be factored into this indicator.

The methodology followed is also the same as that used before, which I’ll describe briefly here.

We want to know if the daily trading volume of a stock gives any indication of its price movement over the next five trading days.  For each of 29 stocks, I calculate the weighted average of the previous 15 days of trading volume.  The most recent trading day gets the most weight, trading volume from 15 days prior gets the least weight.  This weighted average is then divided by a simple average of the previous 15 days of trading volume.

The division by a simple average of the trading volume normalizes the indicator across stocks, some of which may have naturally high or low trading volumes.  The use of a weighted average allows us to determine if volume has been increasing or decreasing over the previous 15 trading days.

The 29 stocks are then ranked, each day, based on the volume indicator calculated above.  A rank of 0 means it had the highest indicator value, a rank of 28 means it had the lowest indicator value.  Finally, a count is kept for each rank as to how many days a stock with that rank had the best performance over the next five trading days.

By looking at a graph of these rank counts, we should be able to tell if this volume-based indicator has any probabilistic predictive value for stock price change over the next five trading days.  I.e., if there is no predictive value, each rank would be expected to have approximately the same number of best performance occurrences over the next five days (with a little randomness thrown in).  If there is some predictive value, one or more ranks should have unusually high (or low) counts.

A graph of this indicator is shown below:

That looks pretty random.  There does, however, seems to be a correlation between having a rank of 0 (increasing trading volume) and relative price performance over the next five trading days.  As before, with the indicator based on (daily volume * percent price change from previous day), the correlation isn’t strong, but there is one.

I do want to emphasize that what’s being tracked here is ‘relative’ price performance over the next five trading days, not absolute price performance.  That means if all of the stocks went down over the next five trading days, the best relative performance is the stock with the smallest price decrease.  But if one or more of the stocks went up, the best relative performance is the stock with the greatest price increase.

We’ll look at variations of this indicator in future posts to see if we can do better, but this is encouraging for a first attempt.