If you change your story, the probability of your guilt is 1.00.

This is equivalent to stating "no innocent person ever changes their story." Empirically, this is false.

Let me provide a personal example. Several years ago, I was nearby when a co-worker was injured on the job. Two years later, I received a summons - she had ended up suing the company, and as an eyewitness, my testimony was relevant. Lawyers for the two parties asked me questions for a while; I answered to the best of my ability, as honestly as I could. My memory of the event was somewhat fuzzy by then, so I tried to only state things which I was sure I remembered accurately, and expressed my uncertainty when I was uncertain.

Once they finished their questions, they handed me a photocopy of a document in my own handwriting, written on the day of the event. The company had gotten such statements from everyone present on the day of the incident, you see. Not only had I forgotten that I had written such a thing until they showed it to me, my own handwriting directly contradicted the things I'd stated from memory minutes before!

This was before I participated in Less Wrong or was otherwise aware of the breadth of human cognitive failings. I was quite taken aback! My brain had betrayed me - not in old age, in my early twenties! I do not generally consider myself forgetful - if anything, I have an unusually good memory - yet in a fairly short span of time, my brain had managed to confabulate almost every detail of the event.

Since I became aware of it, I've noticed myself mid-confabulation on a regular basis. Despite how much my conscious mind values truth, my underlying hardware doesn't seem to much care.

You may indeed be one of the lucky people with flawless recall. This is not, by a long shot, a universal human trait. Memory is fragile. It is quite easy for many humans to misremember things without knowing anything has gone wrong, especially when under significant amounts of stress.

I know very little of the Amanda Knox case specifically, so this should not be taken as an argument for either side particularly. I am only arguing that your simple rule is not a good rule - it doesn't actually work.