An issue with what you're saying is that you could pick $x$ orthogonal to (or nearly-orthogonal to) $y$, and then the value could be very large (depending on $A$). Since the orthogonal complement of $y$ has codimension $1$, in a sense this is a "very big issue."
What is true is that if you consider the quotient $$ \frac{x^t A x }{x^t x}$$ for a symmetric matrix $A$, its value is bounded above by the largest and below by the smallest eigenvalue of $A$. This is like restricting your function to a complementary subspace of the orthogonal complement of $y$.
The lower bound will not hold if $A$ is not symmetric. For instance, if $A$ is a matrix that rotates your space by $90$ degrees in some direction, $x^t A x$ may be zero, but $A$ is nonsingular so does not have a zero eigenvalue. An upper bound will always hold simply based on the operator norm of $A$, which is bounded e.g. by the largest absolute value of an element of $A$.