Is the MPA curve fitted to this power data? Meaning previous data was not used to define the parameters of the MPA curve (his fitness signature). That would be analogous to training a machine learning algorithm, obtaining good model fit, but not validating the model on separate data.
The model doesn’t work like you would expect that it needs to be trained. There are only 3 parameters. Sometimes they can explain an athlete’s MPA, even with just one ride. You might make some assumptions like in this case where we estimated his PP to be around 1700 but otherwise the numbers fall within reason of what we expect from him and they tell the story of the race.
