An Evaluation of Change Point Detection Algorithms

In change point detection, we're trying to ether ify these points in a time series. When it happens abruptly, that's when we talk about a change point. If i get my hands on some real world data sets, that's where it gets interesting. I want to apply these change point detection algarithms to my real world data. What's my notion of ground truth in that situation? That's a great question, and that's typically quite difficult.