Ability scores in general are much too important.
I would argue for moving to a BECMI-like arrangement of bonuses: 3 or less is -3, 4-6 -2, 7-9 -1, 10-11 no mod, 12-14 +1, 15-17 +2, 18 or more +3
Then, eliminate everything that adjusts ability scores after creation. (This may or may not include racial bonuses, but in any case these should be reduced to +/-1 at most. Also, I can readily see an exception for the most powerful wish-level magic as well - thus enhancing the wonder of achieving that level of magic.)
I completely agree with this, although I'd go with the actual numbers from BECMI:
- 3: -3
- 4-5: -2
- 6-8: -1
- 9-12: 0
- 13-15: +1
- 16-17: +2
- 18: +3
What I like about these numbers is that the divisions between bonuses/penalties are a reasonable approximation of how the scores themselves are distributed on the bell curve representing 3d6. The 9 to 12 range covers 48.2% of rolls, so there is no bonus. Each + or - 1 bonus or penalty falls roughly one additional standard deviation away from the mean. It's an elegant system and while it isn't quite as easy to remember where the divisions are as in 3.x, it doesn't take that long to remember how it works. Rolling scores using 4d6, drop the lowest, obviously skews the bell curve to the right and it's not uncommon to see characters rolled using this method with at least a +1 modifier on every score.
This range also keeps modifiers to die rolls small. It's not that adding or subtracting two double-digit numbers is difficult -- after all it's elementary school math -- but it does take more time and attention than when at least one of the numbers is usually single-digit. The smaller modifiers also mean less variance between the top and bottom ends of the spectrum.
I don't mind having rare and powerful magic other than wishes that can permanently increase ability scores, such as the various magical books in AD&D.
That may well not be a popular move. It may even be attacked on grounds of 'realism'. But I'm pretty sure it's the right thing to do for the game, and especially for new players.
I find it realistic enough. The mental and physical challenges which adventurers face on a regular basis are beyond the comprehension of the average person. I don't find it at all unrealistic to assume that the typical adventuring PC is more or less in their prime physical and mental condition. This isn't necessarily to say that the PC has absolutely no room to improve, but that they are at that point where further improvement requires a level of dedication which would leave little room for other pursuits. Running around in heavy armor while carrying a backpack full of equipment and swinging a sword all day will certainly make you functionally strong, but you're not going to have the same physique as a dedicated bodybuilder. And it's arguable that the bodybuilder would have a greater functional strength for anything other than lifting heavy objects.
(For new players, it is vastly preferable to roll stats rather than grapple with the intricacies of point-buy at the outset. Unfortunately, getting the 'right' stats in both 3e and 4e is just too important to leave to chance. That's yet another barrier to new players entering the game, and that's not a good thing.)
Agreed, and in my experience most players like rolling for ability scores.
I don't really care much for point-buy in 3.x and later D&D, as it feeds right into the optimization problem.
I'd like to see ability scores go back to being largely for descriptive and adjudicative purposes, rather than essentially just a way of determining modifiers to die rolls. Keep the modifiers in the +1 to +3 range and scores capped at 18 for humans. Of course it is only with age and years of experience that I realize how ability scores were intended to be used in the first place.