Very well, the next thieving ability:
I'm going to have to start with a quote from The Bridge on the River Kwai:
"--They say in view of the time element, they don't think a few practice jumps would be worthwhile. --No? --No, they say if you make one jump, you're only got a 50% chance of injury, two jumps, 80%, three jumps you're bound to catch a packet. The consensus of opinion is the most sensible thing for Major Shears to do is to go ahead and jump and hope for the best. --With or without parachute?"
If you are going to train thieves to climb walls, you'd better do it so there's a far less chance of failure. A 15% chance (first level) is plainly ridiculous. Long before the thief lived to become first level, the thief would have fallen to his or her own death.
At the same time, to argue no fail thieving abilities is, all respect to JB who made a comment on this post, is just olympically stupid. Not only because realistically it's not possible, but MORE to the point, where's the freaking drama? JB notes that it takes a lot away from the thief's going up levels ... well duh. How does one improve from a no fail position?
The no fail solution reflects how little imagination a lot of DMs have. There's more to climbing something than simply putting hand over hand.
I had a friend who was passionately into free climbing - no ropes, no equipment, just what you're born with. He considered a mountain like this one an easy afternoon:
Which it is, given that this is the first little lump - called Yamnuska - you come across when you drive west of where I live.
Jan demonstrated his technique by climbing up an ordinary brick wall on the side of an old school we were passing. The mortar allowed for about 7-8 millimeters purchase, but Jan went straight up about two and half stories in about ten seconds. Nothing special.
Now, there wasn't any chance, really, of his falling ... so long as 25 feet was all he was going to climb. Another 25 feet, he admitted, was going to be harder than the first, and he wasn't eager to try a brick wall a hundred feet high.
What I'm saying is that your chance is dependent on how much you climb, not on the mere choice of doing so. Most anyone in school could at least climb five or ten feet of rope. It was another thing to climb to the ceiling, and it would have been quite another to climb three times as high. If you're going to make rules about climbing, those rules have to address the practicality of sustained climbing.
Jan had no trouble climbing something like Yamnuska because the mountain has plenty of places for rest. So long as he doesn't need to sustain his energy for more than short spurts, he could go up and down the sheer face twenty times a season with little or no fear of falling. Any thief with equal chances to rest could say the same - in fact, ought to say the same. If there's a window ledge to rest on every 18 feet, even a first level ought to be able to manage a ten story building.
So my proposition would be that the ability would translate to spurts of dexterity x ft. +3' per level, for anything equivalent to a brick wall. (May seem a bit low compared to Jan, but consider the thief is probably carrying equipment, and isn't wearing modern sneakers). Half the purchase and half the feet climbable; double the purchase, double the feet climbable. And so long as there's a chance to rest, no chance of falling.
If the distance did not allow a rest, then the chance of falling the distance of the second allowable spurt would be the % found in the player's handbook. Thus, the first level thief, faced with a 25' continuous climb, would decide to take the risk once climbing past 18'. If the single continuous climb were greater than 36', then double the chance of falling (the thief would have a 70% chance of success). And so on. The thief would have to make up their mind before making the climb.
If the 1st level thief decided to come down again instead, there ought to be a no fault success for the first 18' ... the second 18', however, that would probably have to be rolled for again.
What's nice about this is that you could establish a flat total for anyone else, still based on their dexterity. Thus, the fighter can climb walls - so long as the distance is not great. I would propose dexterity x ft./2 +1' per level. Thus, a first level fighter with an 18 dexterity could climb 9' up a brick wall before needing a rest.
It's the sort of thing that probably needs playtesting to establish a fixed number.
See comments below for an alteration to the above suggestion.