Confused about hyperparameter selection for elastic net regularization using glmnet

I am following the glmnet tutorial here and confused about the statement:

We see that lasso (alpha=1) does about the best here. We also see that the range of lambdas used differs with alpha.

which is based on the figure:

I am confused how they find out alpha=1 is the best choice given the fourth subplot: they seem to have similar trend of MSE even though the lambda values are smaller when alpha is 1.

• I wouldn't interpret "about the best" as meaning "clearly the best". The fourth subplot has points that are much too thick to see what is happening to the left of -2 (as you realize), but it does look like the red dots are a little lower than the grey ones in the -5 to -3 range. Perhaps if you run the examples you can see what the MSE actually is for the various models. – jbowman Jan 9 at 3:13