" In a project with 100 training data points, student A fit a 10 knot cubic spline with training error of 1213, student B fit a 100 knot cubic spline with training error 24, and student C fit a 10 knot cubic spline with training error of 1787. Explain which student will make the best predictions in your judgement. What technique would you use to compare the accuracy of these models? "

The problem that I concerned about is student B's training error is clearly the best one but student B uses 100 knots on 100 datapoints so wouldn't that cause overfitting?