FILTER PRICE DATA: Moving Averages vs.
Exponential Moving Averages
by Jack K. Hutson
In the process of collecting stock or commodity time series data, such as daily closes, we lose a
significant quantity of information. We are examining a single point (sampling) of price history and
endeavoring to glean some market understanding based only on a single daily price sample. This price
sampling of a trading market frequently distorts the actual price signal. Among the diverse sources of
distortion may be simple record-keeping errors, emotional overflow from other stock issues or
commodities, rumors and political repercussions. Exchange specialists or floor traders also tend to
amplify minor intraday price moves enough to allow them an opportunity to trade, sometimes with a
Modification of this stream of daily price signals is sometimes necessary to remove the noise and reclaim
the original market message or direction. This process is called filtering. Analyzing how much noise may
be separated from the input data, without destroying the tradable movement, is not simple. A properly
optimized simple linear moving average can often perform as well as an extremely complicated filter.
There are cases where the very best results require an elaborate filter but where only slightly inferior
results can be obtained by using comparatively simple filters.
When examining stock price historic data, it is common to look only at a plot of weekly ranges.
Sometimes Friday's closing price is shown as a small tick to the right or simply a cross on the weekly bar
chart. By not plotting daily price ranges, we have applied a simple visual weekly filter to otherwise hectic