A live, stochastic SIR simulation where the transmission rate flickers randomly around a fixed average. Watch the trace, place your bet on the disease's ultimate fate, then let the model run out its clock. Same average transmission, different noise level — different destiny.
Based on:"SIR models with demography, random transmission coefficient and non-autonomous vaccination rate" — arXiv:2510.21371.
The core finding: replace the classic SIR model's fixed transmission constant β with a bounded random process fluctuating around the same mean, and the mathematical threshold that decides extinction vs. permanent endemicity shifts — it depends on the variance of the noise, not just its average. Two epidemics that look identical on paper (same R₀) can have opposite long-run fates purely because one has jitterier transmission than the other. This demo runs a real Euler–Maruyama simulation of that model (births/deaths, vaccination, mean-reverting bounded β(t)) — nothing here is a canned animation.
average infectiousness (higher = more contagious on average)
volatility of β(t) around its mean — the paper's key variable
Transmission-rate trace β(t) the noise you're betting on
Susceptible
Infected
Recovered/Vaccinated
280 dots, colored in exact proportion to S/I/R — not decorative, driven by the live simulation state.
Place your prediction locked until t≥8
Start a new epidemic, watch the β(t) trace for a while, then lock in a prediction before time runs out.
0
Predictions
0
Correct
—
Accuracy
0
Streak
Theoretical picture updates live from sliders
Disease-free S* (equilibrium susceptible share)
—
Deterministic R₀ (zero-noise threshold)
—
Stationary variance of β(t)
—
Noise-adjusted threshold R₀s
—
Critical σ where R₀s crosses 1
—
R₀ > 1 with no noise usually means "endemic forever." But R₀s is ItΓ΄-corrected downward by the variance of β(t) — enough noise can push a supercritical disease below threshold. This closed form is a simplified illustrative guide (the paper's exact theorem has different constants); the sweep below shows what the actual bounded simulation does.
Noise sweep: extinction probability vs. σ the paper's headline result, empirically
Runs many headless trials at your current β₀ and vaccination settings, sweeping σ from 0 to 1.2. Every trial shares the exact same average transmission rate — only the volatility changes. If the paper's mechanism is real, extinction probability should climb from ~0% to ~100% as σ increases, with the dashed line marking the theoretical crossing point.