aim lab: cheater edition

what aimbots look like to a statistician. crank the dial.

drag the sphere to rotate. every dot is one shot direction.

aim parameters
kappa (concentration)8.0
shots sampled200
suspicion meter
0-30 human 30-100 grinding or gifted 100+ no. just no.
human
estimated kappa: --
mean resultant length: --

why you cannot fake it

every shot you fire is a direction: a unit vector from your camera to wherever the crosshair pointed. plot a few hundred of them and they form a cloud on a sphere. the von mises-fisher distribution describes that cloud with a single number, kappa, which measures how tightly the directions concentrate around the mean.

human aim carries motor noise. tremor, correction, overshoot: it all spreads the cloud. casual players sit near kappa 8, and even pros with thousands of hours rarely push past 40. practice narrows the cloud. it never collapses it.

an aimbot snapping to heads produces kappa in the hundreds, a concentration no human hand can generate, and once the fit is done you cannot fake your way back: the estimate falls straight out of the data. this is the actual pipeline: fit vMF, then cluster the outliers with HDBSCAN. more of the math at jeffdoesmath.substack.com.