Caught in a Mosh

The physics of moving together.
Explore the physics ↘

Build a crowd. Find the flow.

A free, interactive mosh pit simulator. Change the crowd, adjust the forces, and explore how individual choices become collective motion.
Concert floor20 × 10 m
STAGE
person
A top-down interactive particle simulation. Use the person selector below as an accessible alternative.Paused
Counterclockwise preference
Mean speed1.02 m/s
Circulation CCW − / CW ++0.05
Contact pairs0
Kinetic energy6.6 kJ
Click to inspect · ← / → to move selectionWhite: selected net force

Select a person to inspect their forces.Seed 12 · 160 people

One person. Many forces.

Click a particle. Follow the push and pull.

right = +x · down = +y

Same scale for every arrow.
Net force is white.

Person #7383 kg

Speed
1.31 m/s
Velocity (x, y)
-1.17, 0.59 m/s
Preferred heading
153°
Personal gap
0.35 m
Nearby people
5

Heading: 0° right, 90° down.
Dashed halo: half the desired gap.

Forcex (N)y (N)Size (N)
Self-propulsion-35.517.939.7
Neighbor alignment199.4-63.7209.3
Personal space / avoidance0.00.00.0
Circle-route preference340.13.0340.1
Body contact0.00.00.0
Damping31.2-15.734.9
Boundary response0.00.00.0
Resultant force535.2-58.6538.4

Acceleration = net force ÷ mass. Values are model units expressed in N; they are not measurements of real people.

Why does a pit move?

Simple rules. Collective behavior.

↗

People supply the motion

Each person tries to reach a preferred velocity. Muscles turn chemical energy into movement; feet push against the floor, which pushes back. Music motivates movement—it is not a mechanical force in this model.

●●

Contact passes a push along

Overlapping bodies act like soft, damped springs. A pair feels equal and opposite contact forces. Walls also push back, exchanging momentum with the venue.

◌

Steering is a choice

Personal space and looking a little ahead encourage avoidance. Alignment nudges a person toward nearby velocities. Circle mode adds a gentle preference to run around the center; no person follows a fixed track.

↘

Energy in. Energy out.

Drive and steering can add or remove kinetic energy. Damping and inelastic contacts dissipate it. Switch self-propulsion off: all active steering stops, and the crowd settles. People are an active system, so crowd momentum and kinetic energy are not conserved on their own.

Under the hood: equations & limits
m a = Fdrive + Falign + Favoid + Froute + Fcontact + Fdamp + Fwall

Drive: m(v₀e − v)/τ, with τ = 1.5 − 0.85I seconds. Intensity I also increases random heading changes. Damping: −0.32mv. At zero preferred speed, avoidance may still reposition people; turn propulsion off to remove all active behavior. Alignment matches the mean velocity within 1.8 m. Avoidance uses the desired edge-to-edge gap plus a 0.45 s look-ahead.

Contacts use a 30,000 N/m spring and 450 N·s/m normal dashpot, clamped to repel only. Agents have masses 55–100 kg and radii 0.22–0.255 m. These are illustrative choices, not fitted measurements.

Circle mode steers headings toward a selected tangent and adds a soft radial preference near 3.15 m from the center. It deliberately supplies a route and turning direction. Collective spacing and disruptions emerge from local forces; this does not demonstrate spontaneous direction selection.

The solver uses fixed 1/120 s steps and semi-implicit Euler integration. Boundary impulses prevent escape; an 8 m/s numerical speed ceiling prevents blow-ups. Long frame gaps are discarded instead of jumping the physics. Reset reproduces the same seed; New crowd changes it.

Circulation is the mean signed tangential fraction of velocity: −1 is counterclockwise, +1 clockwise. It is a coherence score, not angular velocity. Density counts particle centers per 1 m² cell. Contacts counts touching pairs at the current physics step.

No fear, fatigue, falling, limbs, consent, obstacles, exits, music response, or real human decision-making. No parameter has been calibrated for injury or crowd-safety prediction.

From the video

The simulator was inspired by Dr. NEMO’s exploration of mosh pits. At 1:44–2:44, people circulate around an opening. The narrator reports 129 counterclockwise examples. That sample is not proof that clockwise pits never occur.

At 20:29–24:10, he explores alignment and directional bias, then explicitly treats the explanation as a hypothesis.

In this model

Soft discs, masses, steering strengths, the venue, and circle-route preference are our modeling choices. This is an original educational sandbox, not a reproduction of the video's experiment.

Related research: Silverberg et al. (2013) studied gas-like mosh motion and vortex-like circle pits using a simplified flocking model.

i

An educational model

Explore patterns.
Not a validated crowd-safety prediction tool.