Seth brings up something that I've been saying for a long time, but from a different perspective. He talks about being in a local high and having to go through a low to get to a larger high.
As a scientist, we always talk about equations finding the minima - the place where the parameter of the equation are optimized. But, some parameter combinations may exist that are wrong, but optimize on a local minima, meaning that somewhere else there is a better minima, maybe even the global minima.
In regression analysis parlance we used to say that we had to add energy to the equation solving to get us out of the incorrect local minima, over a bunch of local maxima (where the parameter do not want to be), to find the global maxima.
To make it simpler, think of it as one of those ball games where you need to get all the balls into the right holes - you give the game a shake to get the ball out of the wrong holes and into the right ones.
All this to me was basically to tell folks that we might be stuck in a comfortable minima and unless we add our energy to the system, we will never know if it is the optimal minima.
Eh, don't know if that makes sense, so maybe you should go read Seth.
Link: Seth's Blog: Understanding Local Max.
My guess is that you've been wrestling with your Local Max.
