A Graphical Search problem, denoted $\Pi(X,\gamma)$, where $X$ is the vertex set or edge set of a graph $G$, consists of finding a solution $Y$, where $Y \subseteq X$ and $Y$ satisfies the predicate $\gamma$. Let $\hat{\Pi}$ be the decision problem associated with $\Pi(X,\gamma)$... A sub-solution of $\Pi(X,\gamma)$ is a subset $Y'$ that is a solution of the problem $\Pi(X',\gamma)$, where $X' \subset X$. To $\Pi(X,\gamma)$ we associate the set system $(X, \mathcal{I})$, where $\mathcal{I}$ denotes the set of all the solutions and sub-solutions of $\Pi(X,\gamma)$. The predicate $\gamma$ is an accessible predicate if, given $Y \not = \emptyset$, $Y$ is a solution of $\Pi(X,\gamma)$ implies that there is an element $y \in Y$ such that $Y \setminus y$ is a sub-solution of $\Pi(X,\gamma)$. If $\gamma$ is an accessible property, we then show in Theorem 1 that a decision problem $\hat \Pi$ is in $\mathcal{P}$ if and only if, for all input $X$, $(X, \mathcal{I})$ satisfies Axioms M2', where M2', called the Augmentability Axiom, is an extension of both the Exchange Axiom and the Accessibility Axiom of a greedoid. We also show that a problem $\hat \Pi$ is in $\mathcal{P}$-complete if and only if, for all input $X$, $(X, \mathcal{I})$ satisfies Axioms M2' and M1, where M1 is the Heredity Axiom of a matroid. A problem $\hat \Pi$ is in $\mathcal{NP}$ if and only if , for all input $X$, $(X, \mathcal{I})$ satisfies Axioms M2", where M2" is an extension of the Augmentability Axiom. Finally, the problem $\hat \Pi$ is in $\mathcal{NP}$-complete if and only if, for all input $X$, $(X, \mathcal{I})$ satisfies Axiom M1. Using the fact that Hamiltonicity is an accessible property that satisfies M2", but does not satisfies Axiom M2', in Corollary 1 we get that $\mathcal{P} \not = \mathcal{NP}$. (read more)