>Let the mount run unguided for a few minutes while watching the graph. Throw out any really big outlier values and take the peak from what is left.
This, ideally, is exactly what PHD2 GA should be doing, right? And it should be what GAs recommendations are based on. Now I haven't looked to see how PHD2 generates it's GA recommendations, but as I mentioned below I believe it's something the PHD2 team is looking at as a result of this investigation.
>Taking the RMS seems like it would guarantee that the mount is going to chase seeing.
To be clear, the general recommendation I have seen is to use some multiplier of the RMS value, not the RMS value as-is.