Curling Dynamics | Physics-Constrained Bayesian Trajectory Analysis
JENSEN-SHEGELSKI 2004BAYESIAN

Why Does a Curling Stone Curl?
A Bayesian Investigation

Compare friction models through physics-constrained inference. Place stones, launch, watch real-time collisions with scoring analysis.

Numerical ODE Integration · Witt-Hansen Elastic Collisions · Monte Carlo Uncertainty
Ready
SHOT RESULT
--
P(SCORING) FROM UNCERTAINTY
Before--
After--
To tee--
Scoring0
Wet
Dry
Pivot
95% CI
House Placement (click to place stones)
Stone Rotation + Forces
Shot Setup
Release Speed v₀2.0 m/s
28m (house)
Rotation w₀1.5 rev/s
Dir
Ice
Model
Scoring Probability
Wet
--
Pivot
--
Dry
--
PosteriorsN=500
phi
-0.19
alpha
0.008
a_T
5.5e-3
lambda_y
1.84
ELPD
Wet
-12.3
Pivot
-18.7
Dry
-45.1
Place stones in house panel, then launch
Translation (Eq. 7)
y(t)=yF[1-(1-t/tF)λ]
Lateral Curl (Eq. 12)
ax=aT-|av||ψ|
Elastic Collision
v'n=un,other (equal mass)

The Mystery

A clockwise stone curls right, opposite to naive friction. 8+ models compete to explain why after 100 years.

Bayesian Approach

Each friction model has physics-informed priors. Posteriors reveal which physics best explains the data.

Key Finding

Dry friction predicts lambda=2.0. Real data: 1.84. The posterior on phi rules out constant friction.

Build Models Like This

Physics-constrained Bayesian modeling for sports analytics.

Register Free
Athlyticz · Jensen & Shegelski 2004 · Witt-Hansen 2018