The problem with this is that it interacts in odds ways with Keen and Improved Critical (which I allow to stack). 17-20 becomes 13-20, and then 9-20, which is totally unreasonable.True criticals became very rare, but I added a few feats allowing extra dice for beating an AC by such-and-such amount to simulate 'lesser' criticals. This one looks more balanced, but also more complex, which isn't something needed in DnD... I would reccomend dumping the critical multiplier changes, and maybe simply add one to all critical ranges. (20 becomes 19-20, 19-20 becomes 18-20, 18-20 becomes 17-20..) Since this is reducing luck, that seems reasonable. Yes?
I like the roll to confirm the failure system because it keeps it relative to the PC's skill and also maintains the precendent set. Essentially, although its luck whether the confirm roll is called for, it is the PC's skill that primarily determines whether he ACTUALLY critically fails or succeeds.
I don't understand your first suggestion, aside from that it would be far more complicated and ackward that what I am currently doing. Skill ranks are already taken into account in the modifier on the roll, having them also apply to how the roll is made is convoluted and a major balance issue.In that case, you may wish to roll different dice depending on the base skill in question. e.g:
1-4 ranks: 2d4
5-6 ranks: 2d6
7-8 ranks: 2d8, etc...
For criticals, I always favoured the notion that, if you hit at all, just roll to hit again, at a -4 penalty (modified by 'threat range'.) Repeat as neccesary, doubling damage each time. That'll cut them Hit Points down to size...
For the second, that seems a very overpowered way of doing things that also completely ignores the differences in crit range and modifier in different weapons. Not only are those important aspects of balancing various weapons against each-other, but they also help differentiate the weapons a bit. Think about all the 1d8 and 1d10 weapons that would be made much more mechanically similar under this system. I happen to like assymetry.
I don't like ethier of your suggestions, at all.
Sure, and a certain element of luck still exists under my system. I'm not suggesting that everyone just "takes 10" all the time, and not roll for success resolution at all. I'm merely introducing a rolling mechanic that changes the equation a bit and IMO makes the game more consistant and in general funner.As has been mentioned, a certain amount of luck IS realistic, but strictly speaking 2 dice don't give a bell curve so much as a pyramid.