Modelling the Sun for a dual-axis solar tracker

My family has been thinking about rooftop solar in Tokyo, but a fixed panel only faces the Sun squarely for a moment each day. How much does following the Sun really buy you?

I modelled the Sun over Tokyo minute by minute in Python, compared four panel strategies, and mapped the motion onto a Keyestudio KS0530 Arduino kit.

Dual-axis tracking collected 27.7% more than a fixed tilted panel, but the kit's electronics use about five times the energy tracking gains: it models the motion, not the return.