Let $w \in \mathbb{R}^d$ have unit norm and $x_1, ..., x_n \in \mathbb{R}^d$ be $n$ randomly sampled vectors from the uniform distribution over the $d$-dimensional unit sphere. Can one obtain a lower bound of
$$\max_{i} |w \cdot x_i|$$
as a function of $n$ and $d$ with high probability?
It seems like given symmetry, we can WLOG $w = e_1$ and just work with the first coordinate. However, I'm not sure how to lower bound the max absolute value of $n$ draws from distribution $\frac{Z_1}{\sqrt{Z_1^2+...+Z_d^2}}$ for iid $Z_i \sim N(0, 1)$.
Thanks for your help.