Given a subset L of the 2d
closed orthants in d-dimensional Euclidean space, is there a convex set K which
intersects those closed orthants in L, while missing those not in L? A strong
combinatorial condition on L, which is necessary for the existence of such a
convex set, is exhibited. This condition is studied and its close connections
with the theory of oriented matroids are examined. The sets L satisfying
this condition—the “lopsided” sets—have a rich combinatorial structure
which can be exploited in the study of convex sets and systems of linear
inequalities.