Thank you so much for your answers, @gbernstein! Your answers are very clear and I think I can start implementing the calibration assessment in python now.
Especially thank you for switching the notation to the regression problem. It allows me to understand better and conveniently translate to my problem, which is actually exponential fitting (Y = A + B*exp(-t/tau)) with limited data i.e. from 3 to 10 data points.