Page 1 of 1
What is the shortest period of M.A. to be used?
Posted: Thu Jun 06, 2013 8:48 am
Selecting the shortest periods for entry, do anyone have any suggestions on what periods to be used for moving average in general?
Thanks in advance for any suggestions
Posted: Thu Jun 06, 2013 1:45 pm
Lots of people run backtest simulations and make 2D or 3D plots to help them choose system parameters.
On a 2D plot they'd put (moving average #days) on the horizontal axis, and (Kestner K-Ratio) on the vertical axis. Or whatever "measure of goodness" they prefer instead of K-Ratio.
Some people pick the very highest peak on the plot.
Others look for a zone of acceptable results, and then choose the center of the zone, feeling that this increases their robustness when the unseen future operates differently than the development data from past history. After all, the center of the zone of acceptability has, by definition, the farthest distance to the cliff(s) of death.
On a 3D (or higher dimensional) plot, they could calculate the center-of-mass of the zone of acceptability. Or, for each candidate parameter value (i.e. for each point within the zone), calculate the distance to the cliff(s) of death and remember the closest distance. Then choose the candidate whose closest cliff-of-death, is farthest away.