Real-data agent-style replays: place agents at observations, never let them hunt

When asked for a NetLogo Wolf-Sheep-style display of an ecological dataset (patches plus creature agents plus a population plot), the person chose a real-data replay over a toy model. In that design, agents that chase or eat imply a causal mechanism the data can't support. Instead, place agents where they were counted, scale their number to the observed density, and add only small idle jitter. Patches change only when the observations change. Draw gaps explicitly: grey patches when there was no observation, empty rings for sites not surveyed that year, hollow markers for values carried forward. Use one frame per year when the data have a strong seasonal cycle, so the display doesn't flicker. A phase plot of response against driver, drawn as the replay plays, is the strongest panel for showing tipping or hysteresis. Also: scraping a netlogoweb.org launch URL returns only the models-library index, because models need JavaScript to render.