I had some thoughts on how to rework some of the thieving abilities for AD&D, without going down the road taken by skill sets and later editions of the game.
The Number One problem for thieving skills would be that they are based on percentages. Percentages are well and good, and have their place, but it is extremely hard to balance play and success as a thief on the basis of pure, unadulterated chance. If the abilities could be reworked to either eliminate or at least reduce the importance of percentages, I think this would be a vast improvement.
Up until recently, however, I haven't given much thought on how to do that. I've been running a few adventures that have been intensely thief oriented in the last year, however (player choice, not mine) and I've been increasingly frustrated at the weak-ass worthlessness of planning activities around these silly percentages.
Count the number of times you can remember where anyone attempting to open a lock in a film or a book stopped, turned to their companions and said, "nope, sorry, I can't open it."
Obviously, if I tried to open a lock with a pick, this would probably be something I'd have to say in short order. But I'm not a thief. I'm not able to cast spells or fall from buildings without taking damage, but in D&D I don't have to throw percentages in order to accomplish those things. Logically, if an illusionist can manage a phantasmal killer without chance of failure, a thief ought to be able to open a freaking lock.
Yet naturally, there's the voice of drama again: "We can't have the party just going anywhere!" It cries. "That would be anarchy! We must make it possible for parties to be stopped by doors, even if they can destroy six stone giants!"
So we are in a dilemma. There is something to be said for thieves having trouble opening locks. Still, I think the drama can be retained if we ignore the chance of success in exchange for time until success. Sure, the thief can open the door ... but can he or she open it right now? Does it take one round, so it can be done quietly and without waking anyone up, or is the thief clumsily taking a dozen rounds to succeed?
A balance that takes into account the thief's level is easy: something like 2d4 - level would be sufficient. Using a small result would mean that by 7th, the thief was well past all that bullshit noobie stuff with regards to opening locks. At first, it could still take a few tricky, inconvenient rounds if opening the door quickly mattered.
Of course, you could easily mess with the players by having 4 classes of lock: 2d4, 2d6, 2d8 and 2d10. A 2d10 lock would be tough and inconvenient even for a master thief, half the time ... it would almost always be a long drawn out struggle for a 1st level. Remember, since actual success is not in question, the main problem would be the thief standing around for 15 rounds while trying to get said lock open. Additional problems could be created by arguing that if the thief is distracted, the lock resets and has to be addressed from scratch; or you could implement the old % roll to see if the thief was able to pick up from where the thief left off, or if the thief lost his or her place. Either way, the drama shifts, so that locks are more an obstacle like a speed bump ... which a good DM could use to build a hell of a lot more tension than Yes it's open or No it isn't.
As a thief, this has to be the most vague, annoying, sometimes useless ability in existence. If there's anything that can be said to be humiliating about being a thief, its when you're 10th fucking level and you still can't simply walk 30' behind a first level guard without his or her noticing you. Sure, a lot of the time, but you're gonna fuck it up 1 chance in 4? I mean, come on! What do you have to do?
We can divide up move silently into basically two features, each defined by what it is used for: 1) thieves use it to get past someone or something; and 2) thieves (and assassins) use it to approach a target. We can call (1) stealth and (2) stalking. Neither should be a percentage die roll.
Primarily, stealth is a defensive ploy intended to avoid conflict, and its success should be dependent upon whether or not it accomplishes that. If I want to move past someone without being heard (or seen, for that matter), then what matters is distance between me and the listener, and whatever material happens to burden me at the moment.
For example - oof! - I am going to carry this - oof! - 113 lb. sofa on my - urg - back while I sneak behind you on your keyboard. My success is going to be - uh - balanced by the fact that I am least several miles distant, and in some cases on another actual continent. This will enable me to move several dozen yards - oh my back - while going completely undetected.
There. I'm done. You did not hear me or see me, and so my success at moving silently, in this case, was 100%. I did say I was doing it, but even despite that, you were completely undisturbed ... and mind you, I'm not even a thief.
On the other hand, if I were to attempt to do the same while almost brushing you with my sleeve, it is unlikely I'd manage. And this is the point. Distance is everything. The better the thief I am, the closer I should be able to pass and the more I should be able to carry while doing so. There is another reason that runners strip down to almost nothing in warzones that has nothing to do with speed - less weight makes you more agile, more silent and less noticeable. You're less likely to snag something on a tree while running along. When's the last time a thief in your world stripped down before attempting to "move silently?"
I haven't quite got an modifier for this, but I'm sure something like 5d4 - level + 1/10 lbs. of weight carried (any weight at all), where the sum equals an optimum distance in hexes (or squares) is a good place to start. All things could be calculated and if the die was less than the distance, the thief succeeds; more than the distance, and the thief fails. Thus, the thief is able to judge the distance against their level and what they take with them, and roll the dice, rather than depending on a ridiculous flat percentage that is supposed to fit every situation.
This would be primarily an offensive use of silent movement, intended to get the thief (or assassin/monk - or even ranger) as close as possible to the enemy before being noticed. It is presumed that at some point the enemy will see or hear you ... but preferably not until you're in sword distance.
Desirably, we'd want this number to occur with the maximum amount of uncertainty while still allowing a reasonable judgement and improvement upon the stalker increasing their level. This is not as difficult as it sounds. If the base die is 3d4 hexes - level of the attacker, a lot of low levels are going to find it difficult to get right next to their enemy. Still, if you reason that the moment of detection is also the potential moment of SURPRISE, it only dictates how far away the thief is before the melee begins. This distance can be balanced by making it +1 hex per level of the highest target ... thus at 1st, there won't be much edge. At 10th, however, its better than average that against a 2nd level target the moment of detection would be zero or less - which means, before the thief attacks, there is no moment of detection. The thief attempts to back stab, the assassin assassinates ... and the melee is resolved.
So, no percentages. Improvement with level, adaptation to abilities as the thief improves ... sounds like good work to me.