Approximations of the strict order property

Abstract: We observe that the definition of Shelah's classical $\mathrm{NSOP}<em>{n}$ hierarchy for first-order theories, for integers $n \geq 3$, can be restated so that it extends to the case where $n$ is replaced with any real number $r \geq 3$. Using this observation, we define a potentially larger family of properties $\mathrm{NSOP}</em>{r}$ for real numbers $r \geq 3$. Motivated by the question of whether the integer-valued and real-valued hierarchies are distinct, we translate these hierarchies into the setting of hereditary classes, obtaining a new real-valued quantity of independent combinatorial interest, $\mathfrak{o}(\mathcal{H})$, associated with any hereditary class $\mathcal{H}$. We show that, when $\mathcal{H}$ is defined by a finite family of forbidden weakly embedded substructures, $\mathfrak{o}(\mathcal{H})$ is an integer. While Malliaris implicitly showed that the properties $\mathrm{NSOP}<em>{n}$ are equivalent to closure under helix maps between graphs, both our observation that the properties $\mathrm{NSOP}</em>{n}$ can be restated so that $n$ can be replaced with any real number at least $3$, and our result that $\mathfrak{o}(\mathcal{H})$ is an integer when $\mathcal{H}$ is a hereditary class defined by a finite family of forbidden weakly embedded substructures, are even exhibited by a special class of helix maps, the interval helix maps. These are helix maps that respect the direction of edges, and whose regions are disjoint unions of linearly ordered sets without any edges between them. Toward showing the conjectural claim that $\mathfrak{o}(\mathcal{H})$ is not an integer in general, and therefore that the real-valued $\mathrm{NSOP}<em>{r}$ hierarchy is distinct from the integer-valued $\mathrm{NSOP}</em>{n}$ hierarchy at the level of hereditary classes, we show that the statement that $\mathfrak{o}(\mathcal{H})$ is an integer in general cannot be exhibited by interval helix maps.