about that graph: I believe it didn't consider the current GS level, only relative change, but I have the impression that the current level makes a difference for the statistical distribution, with an underlying trend towards respectable. not sure about this though.
It didn't -- for that matter of fact I am not sure whether it considered previous weeks, which definitely factors into the distribution. It's obviously not meant to be surgically precise, but as far as being a good guideline is concerned, I think it does an adequate job.
Players who do not play much each week tend to lose focus and their Game Shape will be sloppy. Players who play too much in a week will get tired and their Game Shape will also be low-quality. These changes are gradual, and will build up over a few weeks' time.