Many eyes P(nobody scans) = (1 - v)^N v = share of time one animal is head-up
Detection rate lambda = 1 s^-1 per head-up animal inside range R (wolf crouched: 0.6 R)
Alarm neighbours within 20 m copy the alarm at 3 s^-1 (information cascade)
Flocking separation 4 m, alignment + cohesion within 30 m, flight = 50 % away-from-wolf, 50 % herd heading
Critical speed dD'/dt = -(v - v_c) for v > v_c (D' = finite distance above critical speed)
dD'/dt = +D'max / t_rec at v < v_c
Pursuit lead point = x_t + v_t * min(d / v_w, 1.5 s); catch when d < 1.3 m
The saiga (Saiga tatarica) is a steppe antelope that flees at 60-80 km/h for a few minutes; wolves are faster only over short distances. Both are modelled with the same critical-speed ("D-prime") energy model used for running performance: above the critical speed v_c the body spends a limited reserve D', so a sprint ends when it is empty. A calf-like 15 % of the herd is smaller and 25 % slower, and the wolf picks its target by isolation and distance, which is the standard selective-predation pattern.
Honest simplifications: flat open ground, no terrain, wind or scent; one wolf; detection rate, alarm rate and recovery times are assumed round numbers (not fitted to field data), so the kill ratio shown is a feature of this toy model and not a measured hunting success rate. Sizes and speeds of the animals are real orders of magnitude.
Not run yet.