Well, one way to easily replicate point buy's range per stat (if not its distribution limit over all stats) would be 7 + 1d8. You could also do: Start every stat from 12, and if you want to increase one, you can do so by rolling a d4 as a bonus (rerolling on a 4). However, to do that you'll have to decrease a different one by another / the same d4. So you'll still have the same range, but like with point buy there's an element of control and choice to it.
Regarding bigger ranges, one way could be using that same method, only with bigger dice (and possibly other starting points). E.g. you could start from 11 and roll a d8, rerolling an 8 if you're adding it as a bonus. That example would give you values anywhere from 3 to 18, and it's much more swingy than 4d6dl. Of course, if the high variance is an issue, you can experiment with dropping highest or lowest on 2d8.
For example, if you're dropping lowest on bonus rolls and penalty rolls, you'll get characters with high highs and low lows, or if you're doing it the other way around, you'll get characters where each stat is fairly equal, without much variance to speak of.
There isn't much more I can say without knowing how much variance and player choice you want to include.
Point buy is completely deterministic, so I'm not sure what you mean by probabilities... Are you referring to statistics, like which stat is a certain value most of the time?