Define $k$ (number of faces on the fair die) and T (number of trials).
Roll it twice and record the outcomes as $x,y \in {1,2,\dots,k}$.
Let $C$ count the number of times that $GCD(x,y)=1$ after $T$ trials.
The probability of being coprime (i.e., $GCD =1)$ is
$$ \frac{C}{T} $$
However, the Probability that two random numbers are coprime is $\frac{6}{\pi^2}$
So,
$$\frac{6}{\pi^2} = \frac{C}{T}$$
Thus, an estimate of $\pi$ is
$$\pi \approx \sqrt{\frac{6T}{C}}$$
Wrote a C program to estimate $\pi$ using $10,000,000$ for the faces and trials to improve accuracy. (Approx pi value: $3.141503$)
Is this a valid method to calculate happy $\pi$ day?