But the idea is that when you use a time based estimation you do add a buffer so that you can be sure that you finish the task in the estimated time.
In order to prevent the "buffering", we says we do not estimate the time but the complexity of a task.
That's the theory. In practice: You have a story which is e.g. 3 points complex and you can not really say how long it will take you to finish it. It could be a simple task like painting the wall. The complexity does not depend on the size of the wall so it does not matter if I have to paint a 5 meter wall or a 20 meter wall it still will be 3 points?
Now how is this suppose to help anybody???