For large $x$ it is common to use the approximation $$\left(1-{1\over x}\right)^{x} \approx e^{-1}$$ leading to $$\left(1-{1\over x}\right)^{y} \approx e^{-y/x}$$ and so $$\left(1-{2\over x}\right)^y\approx e^{-2y/x}\approx \left(1-{1\over x}\right)^{2y}$$
But is there a simple approximation for the difference $$\left(1-{1\over x}\right)^{2y} - \left(1-{2\over x}\right)^y \approx \,?$$
When $x=y$ the difference is about $\frac{e^{-2}}{y}$ but the pattern seems less obvious when they are not equal.
We have
$$\left( 1+ \frac{a}{x}\right)^x = e^a \left( 1 - \frac{a^2}{2x} +O(x^{-2}) \right)$$
Using this, assuming $y/x = a$ (constant) we get
$$\left(1-{1\over x}\right)^{2y} - \left(1-{2\over x}\right)^y \approx e^{-2 a} \frac{a}{x}=e^{-2 y/x} \frac{y}{x^2} $$