The first thing I would like to talk about is the importance of establishing a range of values of sufficient length.  If you pick points rather than ranges, the window of any given trade can be too small.  The program might not make any trades at all!

This point leads to the use of indicators as the gearing function.  Using an indicator, let us use cci as an example, would prove problematic as the values change too rapidly.  The quick solution to this would be to set the period to some large number, or to apply it to a larger time frame.

So, to establish a range of sufficient length and to have a slow moving gearing function.

As to specific examples let us start with what Snow Leopard suggested initially, price deviation from an ma.  First there are the issues raised above.  Also, let us say that price has deviated 2 pts above or below the ma.  This could be a part of any number of price movements.  It could be the peak of a small move, the beginning of a huge leap, or the growth of price could be slowing from a recent surge as the ma starts to catch up.

But this could be a problem not just with price deviation, but also for indicators.  CCI can bounce off the 100 line or shoot up past 250 but it still has to pass by 50, right?  Therefore it might be good to instead, define the recent history.

One way to do this would be to use fractals or zig zags and calculate frequency or magnitude of trends as the gearing function.

Other options might be to use yet more ma's as the gearing function.  You could look at price deviation between two or more ma's as Snow Leopard suggested.  Alternatively, there could be a defining ma.  The ma's used in the ma cross would be based on whatever ma is declared the defining ma.  This defining ma could be one that has a positive second derivative or has a slope greater than some value proportionate to its period.  Just suggestions.

In all this there has been no mention of timeframe.  These are just general ideas in need of work.  Also any criticism is by no means a dismissal of the concept, it is important I think to raise questions of the validity of any method to prove its validity as well as to improve it.  

I entertained the concept of stma's before, but I decided to work on other things for awhile.  I am willing, however, to return to this idea if someone is interested in working with me.

-atacii