Algorithm Design, Testing, Optimization, and Implementation

This is what I've come up with so far:

Double top criteria:

A trend has been established according to preset criteria maximum of several Fibonacci spaced simple moving averages

A peak has been established by a minimum decline of at least 21% of the determined uptrend with a maximum decline of 34% of the upward trend to its peak at that point.  I start with these natural "Fibonacci" numbers that can be broadly optimized over time and across a wide range of securities.

A resumption of the trend follows to a minimum distance of 21% below the first peak basis initial decline from the first peak with a maximum of 89% above the first peak basis initial decline

A breakdown from the 2 ^ndpeak at least 3.1416x faster than the move from the low of the first qualified down move drives the price to one tick below the previous low at which a stop order triggers a short sale or put purchase for the stock at hand

An automatic stop loss is placed at time of entry 89% above lowest of the two peaks basis second decline

Exit order is placed at time of entry 100% below the lowest point of initial or second decline basis 3.1416x faster decline vs rise from lowest point between peaks to second peak

Exit order is modified at a Fibonacci reverse logarithmic multiple of 1.618034x the minimum 3.1416x figure for a 100% exit projection, eg a speed multiple of 6.2832 would project to a 161.8034% of initial decline below its lowest point and decelerate with increased multiples

Logarithmic smoothing would be the domain of someone who could surely derive an off-the-shelf formula for a reasonably smooth gradient of such a diminishing multiple of trade determining decline

With smaller percentages of initial trend before the double top remaining down to 66% (100-34), profit objective would be lowered; with larger percentages up to the maximum of 79% (100-21), profit objective would be raised (TBD)

Wider overall tops increase profit objective (TBD)