I've got this for a while and, as far as I know, this is a novel approximation of $\tan(x)$:
$$f(x) = \frac{x(\pi^2-4x^2(1-\frac{8}{\pi^2}))}{\pi^2-4x^2} \approx \tan(x)~\quad\text{with}\quad{-\frac{\pi}{2} \lt x \lt \frac{\pi}{2}}$$
It has a very interesting property, which is the one I'd like to prove, that the limit at $x=\pm\frac{\pi}{2}$, is finite, a rather interesting property for an approximation, at least to my knowledge, as they generally tend to have, at most, a ratio that approaches $1$, and the difference tends to diverge to infinity, like Stirling's. Thus allowing the creation of a simple correction term to minimise the difference even more.
$$\lim_{x\to\frac\pi 2}(\tan(x) - f(x))$$
According to WolframAlpha, this equates to $\frac{5}{\pi}-\frac{\pi}{2}\approx 0.0207531$, which seems about right, but it doesn't show how it got there. I'm just a hobbyist, so my abilities mostly stop at L'Hôpital's...