Thief skills, however, don't fare as well. First note that there is (usually) no such thing as an "average, untrained" thief: You have to be 1st level to be a thief whereas those untrained combatants above are 0-level humans. But if we look at the average chance for a thief skill to succeed, this is what we find:
- Moldvay Basic: (15%+10%+20%+20%+87%+10%+33%)/7 = 27.9%
- Labyrinth Lord: (17%+14%+23%+23%+87%+13%+33%)/7 = 30.0%
- Swords & Wizardry: (85%+15%+50%+10%+20%+10%)/6 = 31.7%
Again, these are the chances for a trained character to succeed, and they are much lower than the chances for an untrained character to kill another. How can this possibly be right? (Also note that these averages are only "saved" by the fact that "climb wall" is amazingly high.)
Of course we can make up all kinds of explanations. A first one might be that it's simply harder to pick a lock than it is to kill someone. I lack the personal experience regarding both, so I can't really argue with this line of reasoning, but it seems iffy anyway. For one thing locks don't fight back. Also, how come that climbing a wall is so much easier than killing someone if picking a lock is so much harder?
Another popular explanation says that thief skills are "extraordinary" in the same way spells are. So all characters can try to be stealthy, but only thieves can move completely quietly without making any noise whatsoever (silence spell?). Similarly all characters can climb, but only thieves can climb smooth walls (spider climb spell?). Things break down a bit for "open locks" I guess since I don't think too many referees would give an average fighter a chance to open a lock short of bashing it to pieces, but hey, nothing's perfect.
But wait, why is this even a problem? A magic-user starts out with a single spell and has to make it through a few levels before becoming useful, so why shouldn't the same be true for a thief? I cannot argue with that take either, except to say that this attitude makes pretty much all classes nothing more than "bad fighters" for the first few levels. If that works for you, great. Personally I find it lacking.
Clerics are actually pretty decent fighter-substitutes, but thieves and magic-users are not. Having shoddy skills or only a single spell can be frustrating for the player who feels like they are not contributing enough in the early game. And while the power curve of the magic-user seems to justify their initial lack of skill, the power curve for the thief is a lot flatter: they don't become "cool" at level 5, they just finally become usable for their original purpose at around level 7. Not the most satisfying progression.
I feel somewhat vindicated in my "thief skills should be 50% right away because that's how the combat system works too" attitude by the way Delving Deeper handles thief skills: All of them are fixed at 50%. The only "problem" is that there is no improvement as the thief gains levels. So the character is useful right away, but aside from hitting better and lasting longer in combat, there is no class-specific advancement (except for backstab damage which does increase, but that's not really a "thief skill" in my mind).
So what am I doing in my own D&D-variant to address this?
First I want to give thieves useful skills right away, so they must start at around 50%. But unlike Delving Deeper I also want to give them advancement. I would guess that Delving Deeper doesn't allow advancement because it's unclear how far skills should be improved. If they went for the same kind of advancement they use for backstabbing, then a 12th-level thief would have 100% in all thief skills. And since they use "3-6 on d6 for success" as their mechanic, they can neither advance more gradually, something a d20 or d100 mechanic would allow, nor can they account for a persistent chance of failure to keep thieves from being "infallible" at what they do.
But there is a way around all this I believe, and that is simply not giving thieves access to all thief skills from the start. So in my game, a new thief character picks from a list of possible thief skills the four they want to start out with. They also pick them in order, from the most important skill to the least important one. The system assigns chances of success in such a way that their favorite skill will by 50% likely to succeed at 1st level, with the other skills lagging behind in 10% steps. This represents where they chose to focus during their "apprenticeship" as it were. When they gain a level, they can decide to improve skills that are not yet 50% likely to succeed by 5%, or they can pick up a new skill at 10%.
Since I am using Delta's excellent Target 20 mechanic, this works fine for games where characters retire at around level 10 or so (but it has a bit of stretch beyond that as well). But let's look at the details.
The four initial skills are 50%, 40%, 30%, 20% for an average of 35%. This is not too different from the traditional averages, but note that this thief is good at what they want to be good at, not just good at climbing walls like all the other thieves in the world. Advancing 9 levels allows for a total of 9*5% = 45% of improvement. So if the character never picks up another skill, the final skill profile would be 50%, 50%, 50%, 35%. Since level is added to all checks, this character would have three skills that fail only on a roll of 1. Fair enough, such single-minded dedication is (a) unlikely and (b) should indeed lead to formidable skill.
The more probable case in which the character picks up three more skills on the way would lead to a skill profile of 50%, 40%, 30%, 20%, 10%, 10%, 10% before advancement, or 50%, 50%, 40%, 25%, 20%, 20%, 15% after advancement, so we'd get two "perfect" skills and the lowest skill would have an overall 65% chance of success at level 10. Seems reasonable. And a welcome side-effect is that not all thieves are the same, one of the very few things AD&D 2nd edition actually got somewhat right.
I think I'll call this the "Thieves Evolved" mechanic.
Yes, it's a little more fiddly for the referee than plain Target 20, but it's a lot more enjoyable for the player. And if there's really a need to keep the stat block for NPC thieves down, a table of "typical progressions" should solve that well enough.