Mouse Movement in BAS: Speed vs Human-Like Behavior

A surprising number of bots get caught not by their fingerprint but by how they move. BAS gives you real control over the cursor — it moves along an actual trajectory rather than teleporting onto elements — and three parameters decide whether that motion looks like a machine or a hand.

BAS moves the mouse, it doesn’t teleport

When BAS clicks an element, it can physically move the cursor across the screen to get there. That path isn’t fixed — it’s shaped by three settings:

• Speed — how fast the cursor travels along its path.
• Gravity (attraction) — how strongly the cursor is pulled toward the target. High gravity makes a beeline; low gravity lets the path drift.
• Deviation — how far the path wanders off the straight line. High deviation produces curves and overshoot; low deviation keeps it nearly direct.

Together they define a spectrum from “instant straight line” to “meandering human hand.”

The performance end of the dial

If a site doesn’t analyse behaviour and you just want throughput, you tune for maximum speed. The book’s reference values capture this end:

• Speed: 10000
• Gravity: 6000
• Deviation: 5000

High speed and strong gravity send the cursor straight at the target with minimal wandering, as fast as possible. Across thousands of iterations that saved motion adds up — when nobody’s watching how you move, there’s no reason to move slowly.

The human-like end of the dial

The opposite tuning matters when a site is watching. Many platforms fingerprint behaviour, not just the browser: they record cursor dynamics and flag anything mechanical. The tells are obvious once you know them:

• A cursor that teleports straight onto a button without travelling there.
• A path that’s a perfectly straight line at constant velocity.
• Clicks that always land on the exact pixel centre of an element.

Real human motion is none of that. A hand moves in curves, accelerates and decelerates, overshoots the target slightly and corrects, and lands at a slightly different spot each time. To emulate that in BAS you go the other way on the dial: lower the speed, ease the gravity, and raise the deviation so the path bends and varies run to run.

Picking a setting

The choice is the same trade-off that runs through all of BAS — raw efficiency versus looking human:

• Optimise for speed when the target doesn’t do behavioral analysis and you want maximum throughput. Straight, fast, done.
• Optimise for human-likeness on anti-bot and behavior-tracking sites, where a natural cursor is part of not getting flagged. Slower, curved, varied.

Mouse motion is one more layer of the same identity problem as fingerprints and proxies: every signal should tell the same story. A flawless fingerprint — even one with a real canvas — paired with a cursor that snaps to pixel-perfect targets in straight lines is still a contradiction — and contradictions are exactly what detectors look for. Match the movement to the target, and the behaviour stops giving you away.