Below is the light curve for RT Aur. what’s going on with this blue model?
Indeed, drawing of smooth relevant approximated light curve from scattered points is not trivial task! High degree polynomial regression is not the best solution.
Sharp tooth on the top of your curve is typical for W Vir pulsators, but not for classical cepheides. So “tooth” well as “waves” are artifacts of approximation, on my modest POV.
Well, I plan to learn How to use VStar CHOICE course this year for understanding of all VStar features 
Especially I hope to understand optimal approximation methods for correct color indexes calculation.
Thanks for your question and follow-up comments, and apologies for the delay.
In the first instance I would suggest reading the Polynomial Fit section of the VStar user manual, from page 83. Page 90 and 91 show the metrics you can use to determine goodness of fit as well.
I agree that it’s possible to overfit a polynomial with a degree that is too high so that the model is following the noise as much as the signal.
There is a lot more that we can talk about in relation to modelling. It’s a big topic.
Happy to have further conversation.
David
Yes, David, it’s very big topic. Maybe, a lot of regressions are used in Mathcad or universal statistics software, but it is not the shortest and easy way. In the field of light curves leadership belongs to Odessa astronomy school, which develops different special methods of time series analysis:
Time Series Analysis of Amateur Observations: Various Methods and Some Results
Time-Series Analysis of Irregularly Spaced Data. Comparative Analysis of Complementary Methods
MAVKA: Investigation of stellar brightness extrema approximation stability for various methods
Statistically Optimal Phenomenological Modelling of Variable Stars
Regretfully, VPhot or Peranso uses just polynomial regression, but not splines, running parabolae, running sine or other more suitable methods for light curves global smoothing and their extremums local fitting.
The next release will add more modelling and period analysis plugins.
David
Excellent news, David!
Optimal global smoothing of phase curves will give the opportunity of correct colour indexes calculation.