Assuming no exploding dice, the average results will be:
2, 3, 4, 5, 7, 9, 11, 13, 16, 19, 22
That's one helluva a curve. Without seeing what the rest of the system looks like, it's hard to tell if this would be functional or not.
+1 +1 +1, +2 +2 +2 +2, +3 +3 +3. Working as intended, a little bell curve of avgs. Too steep? This would apply to damage out of a gun in a Call of Cthulhu-style d100 system.
I make the expected scores:
2.5, 3.5, 4.5, 5.5, 7, 9, 11, 13, 16.5, 19.5, 22
The increases are:
1, 1, 1, 1.5, 2, 2, 2, 3.5, 3, 2.5
That means that the biggest jump is from 2d12 to 3d10, so those 3dx rolls are a cut above, though the incremental improvements start dropping off from there. Visitor Q's suggestion of 3d8 fits well if it replaces 3d10. Then the scores become:
2.5, 3.5, 4.5, 5.5, 7, 9, 11,
13.5, 16.5, 19.5, 22
and the increases become:
1, 1, 1, 1.5, 2, 2,
2.5, 3, 3, 2.5
I don't like the drop off of 2.5 towards the end. I'd be tempted to replace the 4d10 with 4d12. That gives:
1d4>1d6>1d8>1d10>2d6>2d8>2d10>2d12>3d8>3d12>4d12
with expected scores:
2.5, 3.5, 4.5, 5.5, 7, 9, 11,
13.5, 16.5, 19.5,
26and increases:
1, 1, 1, 1.5, 2, 2,
2.5, 3, 3,
6.5That last roll is presumably a big increase to represent a truly awesome ability.
But all that's academic. Playtesting is the only way to work out what works.