This isn't realistic at all as no beginner should be able to pick the most difficult lock in the game. But when looking at the newer fallouts, just because you have a lockpicking skill of 75 shouldn't mean that you can pick all hard locks with ease. After all, when distributing points to skills, it signifies that you are learning more about that particular skill. If you are learning more, things should become easier. Perhaps, a system in which at 75, you can attempt to pick hard locks, but they prove to be difficult. As you increase the skill, the hard locks become easier to pick. As you get closer to 100, it becomes fractionally possible to pick a very hard lock, (such as your 5% or higher depending on how close you are to 100) and at 100, very hard locks are still hard to pick, but also somewhat doable.
A very good way ~(Best way IMO), is some form of weighted percentages. The PC having '20' in lockpick skill, would usually have a 1 in 5 chance of opening the lock (per amount of time spent fiddling with it); While the PC with '80' in Lickpick is an expert, and would usually have a 4 out of 5 chance to open it (per amount of time spent on the lock). This does not preclude either from never opening it, or both opening it on the first try, (This
is realistic BTW, just improbable for each). Consider that the locks are in unknown condition; An individual lock may be rusted shut and even the pro can't open it with picks; The novice's lock may have only 2 cylinders in it because it was taken apart and put back together decades (or centuries) ago; (Yeah, its metagaming), but it is possible to open a lock purely on blind luck. (I've had that happen myself ~when I was six years old, trying to open a combination lock on the pool shed in our back yard. I didn't spend all day; I didn't spend 5 minutes. It was locked tight, and I spun the dial a few times and pulled; And I never ~ever got the thing open again after I'd re-locked it).
Weighted percentages tilt the base odds in one way or another. If you have "hard, average, and easy" locks; Those could be weighted -25, +/-0, +25 (respectively), and the guy with 50 in lockpick (1 in 2 chance), would have a harder time opening the hard lock (1 in 4 chance), and an easier time with the easy lock (3 in 4 chance). The novice with 20 in lock pick could never open the hard locks (due to the -25 difficulty adjusting their chances to -5); But with easy locks, their chances are raised to nearly 1 in 2 per attempt.
There is simply no way to implement a minigame that'd be fair for everyone and represent the characters ability accurately. And minigames, in essence, are but busywork - there are ofcourse people who don't mind, but one'd think that for the ones good at them, it's just frustrating having to repeat the same thing over and over again, while it can be frustrating to those too who are not good at the minigame ("My skill is maxed, why do I still struggle with medium locks?").
Barring Weighted percentage... In the recent Fallout's and (presumably Skyrim) the menus are 3d models. A
detailed 3d model of a lock (or various locks :goodjob:) could be used to advantage here. Consider if it were reverted to Oblivion's lockpick minigame, but in 3d. What if the the game checked the PC's skill at lockpicking and incrementally turned the lock (model) from Fallout's frontal view of the cylinder more and more (as skill increased) towards a ? view and making it a cutaway (with faded transparency) of the lock body ~in the way that Oblivion let you see the lock tumblers. If this was done... in practice it could play out that the PC/Player can attempt to pick any lock blind (and possibly succeed), but as the PC's skill in lock picking improves the player is presented with more and more of the lock's interior (reflecting the PC's accurate mental image of what's happening inside the lock when they move the pick).