The source chooses visual octopus count by flooring entered hatchlings divided by one thousand, then clipping between twenty and one hundred sprites. At eighty thousand entered hatchlings it creates eighty sprites. Telemetry later multiplies surviving sprites divided by one hundred by the entered hatchling count, so eighty initial sprites display sixty four thousand rather than eighty thousand. The bars contrast those two modeled totals at point zero zero five pixels per unit. This is a visualization and scaling convention, not a census or a measured survival rate. Real ecological results are not independently validated here. Base shelter is point zero five in lagoon, point four five in meadow, and point eight five in reef. Each artificial structure within thirty five horizontal canvas units adds point four five, with total shelter capped at point nine five. A single lagoon structure therefore changes shelter from point zero five to point five. At six hundred pixels per shelter unit the bars reach thirty and three hundred pixels. The low shelter branch below point two applies a random point four five death chance when a drifting sprite reaches the bottom. These thresholds and probabilities are authored assumptions, not verified biological measurements or a release recommendation. The animation accumulates real frame time multiplied by speed. Once the accumulator reaches one hundred twenty milliseconds it performs one simulation day step and resets the accumulator to zero, rather than processing every missed interval. The illustrative bars compare a one hundred twenty millisecond threshold with a three hundred millisecond delayed frame at one pixel per millisecond. Both can trigger only one step in that frame. Random recruitment, settling and predation make runs stochastic; the paused native step button provides a discrete action. Exported conclusions are fixed prose, so they should not be mistaken for computed validation of a particular parameter experiment.