Basketball Free Throws: Muscle Memory vs. Calculus

When Steph Curry shoots, his brain isn’t solving $$y = x \tan(\theta) - \frac{g x^2}{2 v^2 \cos^2(\theta)}$$. He is using Muscle Memory (Feedforward Control). His brain knows: “If I contract my tricep this much, the ball goes that far.” He has done it 10,000 times. His neural network is trained.

Robots Have No Muscles

A robot cannot “feel” the shot. It has to calculate it every single time.

1. Vision: Camera sees the hoop. “Distance = 15.2 feet.”
2. Physics Engine:
   • Gravity is constant ($$-9.8 m/s^2$$).
   • Launch Angle is fixed (45 degrees).
   • Solve for Velocity ($$v$$).
   • “Required Exit Velocity: 8.4 meters/second.”
3. Motor Control:
   • “Spin flywheel to 3450 RPM.”
   • PID Loop holds it at 3450.
4. Fire.

Reliability vs. Adaptability

• Human: Adaptable. If the ball is slightly heavier, the human adjusts instantly.
• Robot: Precise but Brittle. If the ball is slightly wet (heavier) or slightly soft (less bounce), the calculation fails. The robot misses. This is why robotics teams spend months tuning their “Feedforward” constants to account for battery voltage drop and friction. We are trying to hard-code Muscle Memory into silicon.