I Released One Zombie in Manhattan. 4 Months Later...

I released one zombie in lower Manhattan and simulated what happens to all 329 million people in the lower 48, using the zombie model from a real physics paper and the 2020 census. Then I ran 4,000 more outbreaks to see where it's most dangerous, and where you'd last longest. The answer: the cities fall first, the places between cities are the riskiest after a month, and the last humans hold out on the empty plains of eastern Montana, northern Wyoming and eastern Nevada. Chapters 0:00 One zombie, 329 million people 0:22 The rules, from a physics paper 1:19 The census grid 1:40 Checking my version against the paper 2:49 Patient zero 3:27 One zombie in Manhattan 4:46 One in a million 5:22 Where you'd survive 6:05 The most dangerous places 7:02 Fast zombies, hiding, fighting back 8:07 The lesson 8:38 Check my work How it was done • The model: A. A. Alemi, M. Bierbaum, C. R. Myers and J. P. Sethna, "You Can Run, You Can Hide: The Epidemiology and Statistical Mechanics of Zombies", Physical Review E 92, 052801 (2015), arXiv:1503.01104. Humans, bitten, zombies and destroyed zombies; bite rate 0.0036 per hour per pair, fight-back/bite ratio 0.8 (estimated by the authors from Night of the Living Dead and Shaun of the Dead), about 30 minutes to turn, zombies walking at 1 foot per second • The people: 2020 Census, every block group in the lower 48 and DC (US Census Bureau centers of population and cartographic boundary files), 329,260,615 people placed at random inside their block group and counted on 3 km squares. The busiest square, in lower Manhattan, holds 299,457 people • Inside each square, everyone can meet everyone; only zombies move between squares. Zombies don't cross water • Checked against the paper: 200 people with one zombie at ratio 0.6, zombies win 40% (39.96% here, 200,000 runs); one person per square, tipping point 0.4373 (0.4367 here); single-person outbreaks still going after a week, 21% (19.3% here, 4,000 runs); highest one-month risk in northeastern Pennsylvania and near Bakersfield (same here) • Code (Python + Numba): https://github.com/anishortsnow-bit/b... Results • Patient zero is stopped in about 4 of 5 outbreaks. Each zombie bites 1.25 people on average before it is destroyed • From one person in lower Manhattan: Allentown after 7.6 days, Philadelphia 8.7, Boston 19, Washington 20, Chicago 78, St. Louis 100. After four months, 56% of Americans bitten and nothing west of Iowa and Missouri reached • One in a million infected at the start: half the country gone in about 9 days, 86% after a month, more than 99% after four months (averages) • Riskiest after one month (counties with 50,000+ people): Lehigh PA, Northampton PA, Hunterdon NJ, Carbon PA, Warren NJ. In California: Tulare, Kings, Kern, Fresno. The differences between the top counties are small (Lehigh 20.2% overrun on average, Manhattan 19.0%) • Zombies 10x faster: half the country in 2.9 days instead of 8.7. Everyone meeting half as many people: no help (7.6 days). Humans as good at fighting back as zombies are at biting: the outbreak dies This is the paper's model, with numbers taken from movies. It is not a forecast. The movie scenes are original illustrations. #zombies #simulation #halloween