arXiv Nightly

arXiv:2510.25777 ("Evaluating the effectiveness of Stochastic CTMC and deterministic models in correlating rabies persistence in human and dog populations") shows that a deterministic epidemic model with a "winning" reproduction number says a disease reservoir persists forever — but the stochastic Continuous-Time Markov Chain version of the exact same model, run at realistic small population sizes, gives the reservoir a real, computable chance of going fully extinct from pure demographic randomness. This demo ports that straight onto bankroll ruin: two side-by-side panels share the same positive edge, bankroll size, and horizon — one draws the calm deterministic exponential-growth line, the other runs a genuine client-side Gillespie stochastic simulation (real exponential waiting times, real birth/death rates, absorbing at zero) and reveals trajectories live via a "Roll Again" button, tallying busts vs. survivals. A third panel runs an N-sweep that overlays the exact closed-form Kendall birth-death extinction probability against empirical fractions from live batch simulation — the paper's actual reservoir-size result, made interactive. The judge opened all three entries in headless Chromium: this one was the only build with zero visual defects, the richest real computation (live Gillespie SSA + closed-form validation side by side), and the clearest gut-punch for a betting crowd — watching individual bankrolls bust at zero purely from variance, despite a positive edge. Printing Press Phase Diagram was a close second (a genuine hysteresis loop that visibly diverges on the way back down, tied to QE vs. COVID-stimulus presets), and Fake Tiers Detector was disqualified — its permutation-test math was correct, but a CSS/Chart.js resize feedback loop sent its canvases to tens of thousands of pixels tall, rendering blank off-screen.