[step:Apply the ordinary alternation theorem and translate the condition back]We use the Chebyshev Alternation Theorem for Haar Spaces in the following precise form: if $a<b$, $Y$ is an $n$-dimensional Haar subspace of $C([a,b];\mathbb{R})$, and $F\in C([a,b];\mathbb{R})$, then $P\in Y$ is a best ordinary uniform approximant to $F$ from $Y$ if and only if the residual $F-P$ alternates in sign at $n+1$ ordered extremal points. Its hypotheses apply to $Y$ because the theorem statement assumes $a<b$, the previous step showed that $Y$ is an $n$-dimensional Haar subspace of $C([a,b];\mathbb{R})$, and $F\in C([a,b];\mathbb{R})$ by definition.
By this theorem, $P$ is a best ordinary uniform approximant to $F$ from $Y$ if and only if there exist points $x_0,\dots,x_n\in[a,b]$ and a sign $\sigma\in\{-1,1\}$ such that $a\leq x_0<x_1<\cdots<x_n\leq b$ and, with $E_0:=\|F-P\|_\infty$, one has $(F-P)(x_i)=\sigma(-1)^iE_0$ for every $i\in\{0,\dots,n\}$.
Since $F-P=M_w(f-p)$, we have $(F-P)(x_i)=w(x_i)\bigl(f(x_i)-p(x_i)\bigr)$. Since $E_0=\|F-P\|_\infty=\|f-p\|_{\infty,w}=E$, the ordinary alternation condition is exactly $w(x_i)\bigl(f(x_i)-p(x_i)\bigr)=\sigma(-1)^iE$ for every $i\in\{0,\dots,n\}$.
Combining this equivalence with the transfer of best approximation between $X$ and $Y$ proves both directions of the theorem.[/step]