That depends on allocation of the point buy.
For example, I could, with my 20 point buy allocate ability scores as 18, 13, 10, 10, 10, 10, which is an average of 11.83.
I could also allocate as 14, 14, 14, 14, 10, 10, for an average of 12.67.
The "optimal" allocation in terms of high ability score average is (I think) 14, 13 (x5), for an average of 13.17.
So there's a pretty large spread of averages dependent on point-buy allocation, but each certainly exceeds 3d6 (which yields an average of 10.5) by a large margin.
4d6-drop-the-lowest (which is equivalent to your 24d6 pool) will give average ability scores of 12.24, which is pretty close to the middle allocation. Note, though, that probabilities for at-most and at-least results can vary between statistical ensembles with identical averages.
Now, it's worth noting that, in the point buy method, a 7 (10-3) costs -4 points, while a 13 (10+3) costs 3 - that is, there's a cost asymmetry favoring better scores, which suggests some dice rolls which impose a similar high/low asymmetry, such as 4d4+2 (average of 12, min/max of 6 and 18) or 2d6+6 (average 13, min/max 8 and 18), which give similar averages but constrain the lower-bound and make "heroic" ability scores more likely.
Going old-school: if you're old enough to have read the original AD&D Dungeon Master's Guide, Gary Gygax explained that the "reasoning" behind "3d6" is that it yields a normal distribution of ability scores. Generally, the modern gamer doesn't want this (imagine playing with a 3 or 4 in an ability score!), so it's probably worth shifting curve area "to the right", hence my suggestion of randomization plus a fixed value: it's a compromise between "rolling" and "point buy".