Herzog, while I'm quoting you for context, most of my discussion is directed at the OP rather than you.
If I understand correctly, you want to change a +4 bonus into 4 +1 assignable bonusses.
So, instead of a +4 ability bonus (let's say, strength) giving a +4 to all related skills, attack and damage, you need to spread that around, giving a +1 seperately to (specific?) skills, attack and damage.
As a result, someone with a +1 ability bonus to strength can only have a +1 on skills, attack OR damage.
To me, a bonus or penalty of +1/-1 is not overly granular. Even a 4 point difference changes the probability of success by only 20%. The smaller the bonuses, the greater the impact of the dice rolled. Now, if you change the basis for rolling, you can make a 1 point difference more relevant. For example, if you need to roll 11+ on d20 to succeed, a +1 bonus or penalty changes that success chance to 55% or 45%, respectively. If, instead, you roll 3d6, and need an 11+ to succeed (50% likely), a +1 bonus makes you 62.5% likely to succeed, and a -1 penalty drops you to 37.5%.
This is pretty easy to change by changing the bonus. If you want natural aptitude (high stat score) to be relatively insignificant, stick with +1 for each attribute assigned. If you want it to be more important, make the bonus higher - +2 or +4 for each attribute assigned ;eaves the same customization, but more importance to that natural aptitude for the areas to which it is applied.
In this scenario, assigning penalties (as discussed before) becomes a problem. Where a character will assign a bonus in a field where (s)he will profit most, the penalty will be assigned in a field the character will (almost) never use.
I find it odd to say that a major problem with feats is that non-combat feats are rarely selected, then provide players with a choice of combat or non-combat attribute bonuses. Why would they pick the non-combat bonuses for attributes if they don't pick them for feats?
Practically, campaign style determines the utility of many feats. If my game focuses largely on combat, and everything else is just the fluff to lead the PC's from one combat to the next, noncombat feats seem pretty unexciting. If my game is less about combat and more about intrigue, investigation and interaction, then those skills become a lot more relevant to success or failure, and feats to enhance them will be selected more. Feats are a limited resource. No player, IMO, wants to look at his character sheet and think "I forgot I had that feat - I haven't used in in two levels since I selected it."
I forget where I read the suggestion of assessing your GM's game style with the example that, if your GM likes to use a lot of Giant-type opponents, get Power Attack, since Giants have poor AC and high hp for their challenge rating. If your GM really likes creatures with relatively high AC's and low hp, maybe you're better off with Greater Weapon Focus, since a bonus to hit is more relevant than a bonus to damage in that game.
To counter this, change the penalty system into a bonus system with a lower starting point.
Since the maximum penalty from ability should be -4, (assuming 3 as the lowest possible stat) all ability-bonus influenced character traits (skills, attack, damage) start with a -4 penalty.
A score of 3 will give you 0 assignable bonusses, a score of 4 will give you 1, 6 will give you 2, etc.
(if you assume 1 as the lowest possible stat, start at a -5 penalty and 0 assignable bonusses at a ability score of 1)
Then, allow more than one bonus to be assigned to skill(s), attack, damage, etc. Allowing people to counter those negative offsets somewhat.
Completely tangential, but I've often thought it would be interesting to set 0-1 as +0, 2-3 as +1, etc. Everyone gets an extra +5 to hit, to damage, etc. Of course, everyone also gets an extra +5 DEX bonus to AC, a +5 hp/die bonus, etc. Some extra work would be needed to ensure all tasks increase in difficulty by 5 points, but it's across the board. But this means players are always choosing how large a bonus characters get, not avoiding a penalty. While the math is unchanged, the psychology is very different for some players.
The second thing you should consider is grouping the traits you can assign those bonusses to into somewhat equal trait-groups.
If someone can assign a bonus to either basketweaving or damage, most people will ignore the basketweaving. However, if ALL skills of a specific ability are affected by the bonus, assigning that +1 to skills instead of damage becomes more interesting. You'll probably need to do some tinkering, since not all ability stats have the same amount of skills.
For melee combatants, STR attributes will always go first to To Hit and/or Damage. For wizards, not so much. Again the theory that this reduces character similarity doesn't ring true. If I'm trying to min/max a warrior, why would I not (remembering OP comment that all stats feed to saves):
- take 16 STR and put bonuses to melee attack and damage, and saves
- take 16 DEX with bonuses to ranged attack, AC and saves
- take 14 CON with bonus to hp and saves
- take 12 INT, WIS and CHA, with all bonuses to saves
What has this character lost in combat? Now, maybe he has to sacrifice one or more saves because that stat array is too high. If he has to tank one stat, then it will probably be INT, WIS or CHA and, having lost the save bonus, drop it to 8 (or less) and put the negative(s) on skills (or other elements not affecting combat)?
And all the fighters still look the same, just different from fighters under the current model. If he doesn't care about skills, only combat, no rejuggling of the combat/noncombat choices will encourage him to take skill bonuses over combat bonuses.
