I'm not sure how much adventuring in the city would help the class. Sure, they have a great climb walls percentage, but if they tried to use their pickpocketing skills they'd probably get into trouble. (Or maybe that's the point)
That's the one thing I liked about the opposed roll rules introduced with 3e. Instead of a 21% chance to pick pockets at 1st level, their chance of success would be balanced out by whose pockets they're trying to pick.
But of course, the new opposed roll and skill check rules created issues of their own.
The other item that hampers the thief is how so many DMs supply the idea that if you have a 21% chance of success, that means you have a 79% chance of a negative outcome.
Picking Pockets: Failure allows additional attempts. The victim might
notice and allow the thief to operate anyway in order to track him or her
back to the place he or she uses as a headquarters. Up to two attempts at
picking a pocket can be made during a round.
Getting a 23% just means you didn't pick the pocket. It doesn't mean the target noticed you attempting to pick their pocket. Maybe the DM gives some small chance of being noticed.
The same misconception is usually applied to the other low thief percentiles. failing to remove a trap DOES NOT MEAN YOU SET OFF THE TRAP. Opening a trapped item is how you set off a trap. But if you found a trap you couldn't remove there are many ways to open a known trapped item with little risk. Failing a hide in shadows check doesn't mean you're seen. Failing a move silently check doesn't mean you're heard. By that logic, since other classes have 0% in all of those scores, sneaking up on someone is impossible - but that's not the case.
As is plainly stated in PLAYERS HANDBOOK, this is NEVER possible under direct (or even indirect) observation.
So it is not possible to go from visible to essentially-invisible while your visible self is observed. But this is to achieve a state that is near-invisible
even if observed. That does not mean people automatically observe those in dark corners. Only that if they're suspicious and look to see, they can see. But really, do you go peering at every dark corner to see what might be there? That would quickly become counter-productive.
The thief has, at worst, the best non-preternatural chance to hide you'd extend any other non-thief character who tried not to be seen. And at best something they could never hope to emulate. But it is not a 15% chance to not be seen. It is a 15% chance to defeat active suspicious looking.
People don't like to play thieves because DMs are afraid of being easy DMs and tend to being DMs of NO, far beyond instruction. The proper way to DM a thief is to ask yourself: if there were no thief in the party and some fighter asked to do this - would that have a reasonable chance of success? If yes, the thief might be able to go far beyond that to success even in unreasonable conditions for a fighter.
As for town thieves - what benefits do you give the thief for belonging to a guild? If that isn't a long list of urban bennies, in directed info-gathering, free hot tips, rope-jumping, corrupt watch, compliant merchants, smuggling favors, etc., then yeah, the thief is being shorted again.
Backstab TSR tightened too much. It should have simply stayed as expressed in the PHB - all that is necessary for the damage multiplier was position and an unaware target, and you get an extra, additional +2 to hit if they are surprised - which could be more easily accomplished through use of such skills as HiS or MS. Requiring surprise went too far. (And note that neither the PHB or the DMG state you must roll a MS or HiS necessarily - they were often done to make achieving surprise easier, that is all)