“Yunara 31/3/3 (Diamond 3)”, 29 minutes, 60 fps. Analysed 26 Sep 2026.
We tracked the cursor, clicks and Yunara in all 104,519 frames and tested them against how hands move and how scripts behave.
Every test came back human.
Bell-shaped speed, a correction in 22% of movements, and the −⅓ curvature law in every minute (median −0.365).
Each dot is one mouse movement: distance moved (across) against its top speed (up). Human hands scale like this, with lots of scatter.
Share of movements by where the top speed falls, from start (0%) to finish (100%). Hands peak around the middle.
How much the cursor slows on curves, one dot per minute. Real hands land near the purple line; a path that ignores curves sits at the red line.
Moderate evidence: good humanizers fake these shapes.
Movement shape is identical all game (p ≥ 0.67). Speed only drifts, 12% slower by the late game.
Every 10 seconds, compared with this player's average. Grey stretches: dead or in base.
If the same hands play all game, the three colours stay close together.
Why slowing down points to a person: people tire and play more carefully as a game gets tense. A script runs the same code at minute 29 as at minute 1.
No near-duplicate movements. A bot replaying 20 stored motions would cluster at 0.025; this player sits at 0.059.
How close each movement is to its most similar one. A bot replaying stored moves (orange) has near-copies; this player (blue) doesn't.
Supporting only: a generator that makes fresh motion each time would also pass.
Gaps of 183 ms ± 183 ms, never metronomic. Clicks come at the natural end of each hand swing, as the arc slows, not mid-swing.
Calm play only. A script clicking on a timer would stack into one tall spike (orange line).
Every detected right-click, fights included.
It lands on enemies about 1.2 times a second, roughly the attack rate. Orbwalker users don't need to.
Times per second, for each minute. Low minutes: dead, shopping or farming.
It looked like he hit enemies the instant they appeared. Slowed down, two enemies step into view at the top; the cursor finishes the first fight, turns toward them 400 ms later and reaches them at 720 ms. That's normal human reaction, and the red attack cursor only ever lights up on enemies that are already visible.
An aim assist that pulls the cursor would stack it on the centre of each target. Instead it sits anywhere on the body, a median 35 px from the centre, the same in every situation.
Each dot is one moment with the cursor on an enemy, placed relative to that enemy's centre (the cross). An assist that locks onto targets would pile dots on the cross.
One value per landing on a target. A lock-on would put a spike at the left edge; this is spread across the whole body.
Median distance from the target's centre (dot) and the middle half of moments (bar), split four ways. An assist that switched on in fights or on fast flicks would pull one row toward zero.
Share of cursor stops that end in a burst of speed, a mark of being pulled. Higher on targets (13% vs 7%), a gap we haven't explained. A lock-on pulling every landing would push the red bars far higher.
Checked by eye on 20 random moments: about 17 are the right enemy. The rest (a turret, a damage number) are noise, not assist. An assist that redirects the attack without moving the cursor wouldn't show up here at all.
Half the time within 234 px of Yunara, and in lane over 90% of her walking traces back to where the cursor pointed.
Whole game. Brighter means more time there; the star is Yunara.