I am experimenting with running average and compare... Theoretically, I randomly cycle few parameters until the running average of the input matches the input, resulting in a *bang* that disables random changing of the parameters.

The ratio of match/no-match is around 170000:1. Perhaps this could work? Or am I taking the idiots road around

Lots of ways to do something like this and without seeing your patch difficult to know exactly what your looking for.

Here's a patch where you set a target number and gradually narrow the search until the random number matches the target.

This is an interesting patch, thanks.

I can't show you the patch since it's huge at the moment and needs external files. Basically I'm detecting sinusoids with help of several bandpass filters to clean up the signal and [sigmund~] to detect the frequency of the signal and I want to develop a mechanism that will narrow bandpass filter properties (filter order, passband ripple, width) to get a result as precise as possible.

edit: I'm figuring how to apply your logic to a dynamic target value. And probably the error margin would be set when the program stops narrowing it down.

here's a general method for honing in on a solution-- hope it makes sense --T