Change how far and how big the button is, how shaky the hand is and how long the menu is, then watch the replayed cursor and the time per task.
The science
Pointing follows Fitts' law: MT = a + b·log2(D/W + 1) (Shannon form, D = distance, W = target width). Here it is not assumed but emerges from a simple model: the hand makes ballistic submovements towards the target centre; each lands with an error of ε·|distance|·N(0,1); the user corrects until the cursor is inside ±W/2. Each submovement costs τ and the reaction time is 120 ms, so MT = 120 + n·τ.
Because every correction shrinks the remaining distance by about 0.798ε (mean of |N(0,1)|), the mean count is n ≈ ⌈ln(2D/W) / ln(1/(0.798ε))⌉, a logarithm of D/W, which gives the straight Fitts line with slope b = τ / log2(1/(0.798ε)). Choosing from n items costs Hick's law T = b_H·log2(n+1).
Simplifications: one dimension, no overshoot dynamics, Gaussian errors, identical τ for every submovement; the numbers (τ, ε, b_H) are typical illustrative values, not measurements of a person. The closed form and the Monte Carlo are independent (the first ignores lucky early hits), and their agreement within about one submovement is the check shown above. Educational only.