Mouse-to-crosshair fit
CS2 demos store the raw mouse input alongside the view angles. Fit checks how much of the crosshair's movement that mouse input explains.
Did the crosshair move because the hand moved it?
Mouse inputCrosshair turnIllustration, not real data
What it is
For every kill, Overshoot opens about four seconds of movement and lines up two things: what the mouse did, and how the crosshair turned. A robust straight-line fit of crosshair turn against mouse input gives a number between 0 and 1. Near 1 means the hand explains almost everything. Lower means the crosshair moved in ways the mouse did not.
Nothing is read from the player's PC. The mouse input is already in the demo the server recorded.
Why it matters
Anything that turns the view without the hand breaks the link between mouse and crosshair. Most stats only look at outcomes, such as kills and headshots. Fit looks at how the aim got there.
What ordinary play looks like
Ordinary players do not fit perfectly. Very fast flicks, sensitivity changes, recoil, lifting the mouse and other real-world effects all lower the fit, so a low fit on one kill is common. What matters is how often it happens, how it compares with players using the same weapon, and whether a person watching the moment agrees.
How Overshoot shows it
Every kill gets its own fit. The research fact sheet compares it with ordinary Premier kills using the same weapon class and counts how many kills are unusual against how many would be expected by chance.
What this cannot tell you
- A low fit says the hand did not explain the movement. It does not say why.
- Unusual hardware, software that remaps mouse input and unusual settings can lower the fit for honest players.
- One kill proves little. Many kills, and a person watching the tick, matter.