Defenses protect prey, while offenses arm predators. Some defenses and offenses are constitutive (e.g. tortoise shells), while others are phenotypically plastic and not always expressed (e.g. neckteeth in water fleas). All of them are costly and only adaptive at certain prey densities. Here, I analyse such density-dependent effects, applying a functional response model to categorize defenses and offenses and qualitatively predict at which prey densities each category should evolve (if it is constitutive) or be expressed (if it is phenotypically plastic). The categories refer to the step of the predation cycle that a defense or offense affects: (1) search, (2) encounter, (3) detection, (4) attack, or (5) meal. For example, prey warning signals such as red coloration prevent predator attacks and are hence step 4 defenses, while sharp predator eyes enhance detection and are step 3 offenses. My theoretical analyses predict that step 1 defenses, which prevent predators from searching for their next meal (e.g. toxic substances), evolve or are expressed at intermediate prey densities. Other defenses, however, should be most beneficial at low prey densities. Regarding predators, step 1 offenses (e.g. immunity against prey toxins) are predicted to evolve or be expressed at high prey densities, other offenses at intermediate densities. I provide evidence from the literature that supports these predictions.