Choosing the Next Feature

A multiple linear regression has already been fitted on a set of features, and the residual vector is $e$. Three further candidate features are available. After the component of each candidate already explained by the current features is removed, call the residualised candidate $z_j$. The inner products are as follows.

Candidate $1$:

Candidate $2$:

Candidate $3$:

Which candidate should be added to produce the largest reduction in the residual sum of squares? Enter the candidate number.