That got me to thinking about the rigidity of the B/X thieving skills as written, and whether or not it would be a good or desirable thing to make thieves more customizable, and if it would still be B/X if I did.
AD&D 2E has a thief skill system that appealed to me greatly when I first saw it at some point back in the 90s. For those unfamiliar, each skill has a base value, and at each level, including 1st, a thief character receives an allotment of points to be divided amongst the skills however the player desires. If I remember correctly, there were some minor caveats about how many points could be assigned to any one skill per level, something like 35, I think, which does little to slow the maxing-out of a skill, should the player want to specialize to that degree. As with previous editions, thief skills are d%, roll under.
It does give the thief class tremendous flexibility. If you want to be primarily a pickpocket, or a cat burglar, or a tomb robber, you can just dump your points into the relevant skills, and never raise the ones you don't care about at all. It also has a few big drawbacks. Besides the relative ease of mastering any given skill, all those percentile points are a pain to keep track of. I'm pretty smart, and by no means mathematically inept, but I stuck to allocating points in blocks of five to keep the bookkeeping manageable. (If that's the norm, it makes little sense to choose d% over a d20 mechanic, since the probabilities on a d20 are in increments of 5% But I digress.) There's no way at all to tell at a glance if somebody has fudged on points, or made an honest but significant error in calculation, or just forgot to add the points from the last level-up. It's enough of a headache for a player. For a DM, it makes creating thief NPCs an ordeal. Instead of just rolling for hit points, and maybe ability scores, and referring to a table for thief skills as needed during play, you have to go through all the rigamarole of totalling up how many points a thief of the desired level would have, and distribute them without going over or leaving any unassigned.
The 2d6 system mentioned above is a vast improvement, not least because it reduces the figures to a level that's easier to grok on an intuitive level. Dealing in increments of one point is a lot easier than assigning 15 points per level, and I assume the points handed out per level can be counted on one hand. Also, those bonus points aren't added to a base number. They're added to a 2d6 roll. (Did I mention I like bell curves?)
A while back, when I was tinkering with my own heartbreaker, I hammered out a fairly simple system, using the familiar d%. Basically, all skills use the same advancement table, but there are three different rates of advancement: Good, Average, and Poor. All skills are Average by default, but a character could take Good progression in one or more by reducing a corresponding number of others to Poor. I also had consolidated a few closely related skills, for example Move Silently and Hide In Shadows into Stealth, but that isn't strictly relevant. I scaled the tables to stretch over 36 levels, with increments of 5% to start, and diminishing at higher levels. By level 36, a Poor skill maxed out at 80%, an Average one at 100%, and a Good one at 120%. (Higher than 100% offsets situational penalties for particularly difficult tasks.)
Obviously, this doesn't allow for different chances for each individual skill, but when you look at any given line in the B/X thief skill chart, the difference between the best and worst skill is never greater than 10% anyway, if you don't count Climb Walls. A flat Good, Average, or Poor percentage applied to any skill isn't too far out of line, and with a little tweaking, even Climb Walls can be shoehorned into the formula without too greatly diluting its usefulness.
If you want a generic thief, you can just use average for everything. If you want something a little more specific, all you need to do is decide which skills he's good at, and which he's neglected.
Whether something like this is even useful or consistent with the B/X game is an open question in my mind. I'm not at all averse to keeping the default one-size-fits-all thief skills table, and have thief characters distinguish themselves by the skills they choose to use and how they use them rather than their percentile chances of success.
For anyone who cares to see it, though, here's what I came up with for my Good-Average-Poor system: