Volume Based Indicator Analysis Part 4

This will be my final post on volume-based stock market indicators, at least for a little while.

If you recall, in my previous post I formed a weighted average of the previous 15 days of (daily volume), and used that to predict relative price performance over the next week.  The results were encouraging, but not great.  In this post I’m going look at some variations on the original.  I’ll put the original graph of results here for easy reference:

The first variation is to use an equal weighting of the most recent 6 days, rather than a decreasing weighting of the previous 15 days.

This really isn’t much different.  There’s a little more weight on positions 1 and 2 than previously, but position 0 had fewer occurrences.  It’s interesting that a drastic change like going from decreasing weight with age to constant weight has so little effect on results.

Next, we’ll go back to scaled weights, but put the maximum weight on the day prior to the most recent, rather than the most recent day.

What stands out in this graph isn’t so much the performance of ranks 0-2, but rather the performance of ranks 26-27.  They’ve been suppressed, leaving rank 28 with many occurrences all by itself.  Still, there isn’t much here to say this would make a good indicator.

The next graph is a difference between the 3 day average and the 10 day average.  I.e., each is a constant-weight average of the most recent N days, and we subtract them.

Now, that is an improvement!  Rank 0 occurrences improved from the 350-375 range to over 450.  Rank 28 occurrences improved from about 325 to over 400.

Those are pretty strong signals.  It seems that volume suddenly increasing or decreasing over a couple of  days is correlated with stock price moving up over the next week.

The correlation is probably strong enough that you could trade on it, but it’s also possible to use this as one metric of a more involved indicator, which is one of the things I’m working on.