Let $p$ and $q$ be real numbers such that $p^2+q>q^2+p$. Then the minimum value of $p^2+q$ is what?
My argument is as follows:
Since $p,q$ are real numbers, it should not matter if we take $p$ as $x$ and $q$ as $y$. \begin{align*} p^2+q &> q^2+p \\ x^2-y^2 &> x-y \\ (x-y)(x+y-1)&>0\end{align*} So, we have to maximize $x^2+y$ subject to the conditions that: $$x-y>0,\quad x+y-1>0$$ or $$x-y<0,\quad x+y-1<0$$ From the graph of thse, it's clear that the minimum value occurs at $(1/2,1/2)$ in region $1$ (labeled on the graph). So, $p^2+q = (1/2)^2+(1/2) = 3/4$.
However this is not the correct answer. Can anyone suggest where I'm making a mistake?

Given $p\in \mathbb{R}$, we find the set of $q\in\mathbb{R}$ that satisfy $p^2+q>q^2-p$. Rewrite the inequality:
$$q^2-q-(p^2+p)<0$$
The roots of $q^2-q-(p^2+p)=0$ are
$$q=\frac{1\pm\sqrt{1+4\left(p^2+p\right)}}{2}$$
Notice that $4p^2+4p+1={(2p+1)}^2$, and hence the roots are $p$ and $p+1$. Thus, $q \in (p,p+1)$.
For any fixed $p$, the task of minimizing $p^2+q$ therefore reduces to minimizing $q\in(p,p+1)$. We see the infimum is $p^2+p$, but there is no actual minimum.
Finally, letting $p$ (and thus $q$) vary throughout $\mathbb{R}$, the minimum of $p^2+p$ can be easily found to be $p=-\frac12$. Any methods for min-maxing a quadratic equation will do.