It's super hard to see how many steps/changes any task/feature will require.
So it's just fundamentally impossible to give accurate deadlines.
I try to drill this into all non-programmers I work with.
They hate how wild my ranges are.

If you understand the change(s) required, and you understand the platform you are changing, and can think of the affects these changes will incur, then no, it's quite easy to throw an semi reliable estimate out there.

But when you don't understand the software or the changes required, then your best guess will be wrong every time.