Dear Justin:

As I understand this, you will have, at a minimum, these columns:

Stock identity (ticker?)
Date
Quoted Price

Do you have high, low, and closing prices, or just one price from which to
work for each stock for each day?

Now we need to discuss the "given" values.  Perhaps you have a target date
and a number of days over which to run this.  Then, there must be a
threshhold, say a minimum increase or decrease of so many dollars.

Or is there a target date?  Do you want to consider ALL the periods covered
by the data?

Now, realize this,  If you have 5000 stocks and 200 trading days in a year
for 5 years, that represents 5 million datapoints.  If you want to find the
high and low over every 10 day period (2 weeks) you must look at 50,000 data
points for each day, or 50,000,000 values for the 5 years.  This is going to
take a LOT of time to produce.

If you want the analysis for 1, 2, 3, and 4 week intervals the whole thing
will be 5 times as bad.

With very judicious choices of databases, database construction, and
computer equipment, I believe this could be done in less than an hour.
Without considerable care, it could be a lot worse.

It is possible, however, to store the historic analysis.  This assumes the
daily data is accurate and won't need corrections.  The change in each stock
for a set of time periods can be stored, and this "partially cooked" data
can be searched for differing price change amounts, and much more quickly.
The invariability of historical data makes this a good candidate for
breaking the rules about storing derived values.

Because of the potential challenge of making hundreds of millions of
calculations, I'm going to recommend for this project some careful analysis
before you proceed.

Tom Ellison