The harsh truth is that +5 is equally powerful and desirable at every level. While the cost of a first level spell slot only diminishes as you level up.
Yes, but monster attack bonuses scale with level, and PC AC scales so slowly it effectively
doesn't scale. Sure, you'll gain 1-2 AC when you upgrade armor past starting gear (or invest in Dex for light armor characters), but after that AC is essentially fixed barring magic items. Leastwise, you hit the non-magical cap about level 4, IMX. That's why AC bonuses are so good in 5e. You're scaling something you normally can't. The AC of PCs doesn't scale, but the attack bonus of monsters certainly does. HP and resources like
shield are what scale for PCs, so
shield scales by being cheaper, meaning it represents the PCs getting more durable.
PC AC: Incidental to minimal scaling with level.
Monster Attack [Bonus, Number, and Damage]: Moderate scaling with CR overall.
Now,
shield works against subsequent attacks, so it does help mitigate those. However, what
shield actually says is something like this:
"You have 95% immunity to attacks of CR 1 creatures until the start of your next turn.
You have 85% immunity to attacks of CR 5 creatures until the start of your next turn.
You have 75% immunity to attacks of CR 10 creatures until the start of your next turn.
You have 65% immunity to attacks of CR 15 creatures until the start of your next turn.
You have 55% immunity to attacks of CR 20 creatures until the start of your next turn."
Obviously, you can manipulate that with full plate and actual armor, but in the lion's share of cases, this is what the ability says.
-------------------------
This scaling problem is the same root cause of the problem with the -5/+10. PC attack bonuses scale. PC damage scales (via number of attacks, additional abilities, etc.). Monster ACs, by and large, do not scale. Monster HP scale. So, you end up with:
Attack bonus (scales) vs AC (does not scale)
Damage (scales) vs HP (scales heavily)
The -5/+10 allows you to trade a resource which scales and is opposed by an enemy resourse that does not scale (attack bonus) and instead apply it to an enemy resouce that
does scale. That's the problem.
By the time you reach level 10, you're very likely to have nearly capped your attack bonus. You can easily have +5 at level 1 (2 prof + 3 ability) and by level 10 you'll have +9 or +11 with a bow (4 prof + 5 ability + 2 archery style).
The opportunity cost of taking a feat, so lauded by OP, is completely subsumed by either the Fighter's ASI at 6th or the Variant Human.
Now, yes, for monsters AC scales with CR somewhat -- low CR averages mid/low teen AC and mid/high CR averages mid/high teen AC -- but, almost any given AC can be found at more or less any given CR, with the extremes obviously housing the exceptions. Outside of dragons and humanoids in magical full plate and shield, however, essentially nothing has an AC over 20. A CR 20 Pit Fiend has AC 19. A CR 21 Lich has AC 17. Yet a CR 1/2 Hobgoblin has AC 18, which is just scale + shield (85gp) or chain + shield (60gp). That's starting gear to most PCs. What really scales with high CR creatures? HP. High CR creatures ramp up the HP to very high levels (from less than 10 to well over 200 or 300) and do so to outpace the attack and damage gains that PCs gain from higher levels.
Note that the math here is actually kind of close. If the feats were -5/+7, for example, they would
never be worth it unless you were throwing shuriken. This is because bounded accuracy gives us a fairly narrow window for design of this type of effect.
-------------------------
Note that this is the exact problem that a lot of people have with saving throw DC scaling. It's safe to assume that the non-proficient saving throws you have at level 1 will
never increase. Put a 13 in Wisdom for your Fighter? You'll have a +1 Wisdom save from level 1 to 20. Yet saving throw DCs increase very quickly, starting at DC 12 or 13, by level 10 are virtually guaranteed to be DC 17, and at high level are DC 19.
Imagine if we gave a feat to spellcasters that said, "-5 spell DC or spell attack for +65% damage." That's about as bad as -5/+10 ever gets. Usually it's +90% to +110% damage for the accuracy penalty throughout most of the game. How often would you take that?
-------------------------
In the end, though, while I feel like the -5/+10 ability is a mathmatically sketchy mechanic, I think the core problem is twofold:
1. It encourages players to determine the mathmatically optimal choice during combat for each opponent. Take this formula:
Code:
$AverageDamage * (21 + $AttackBonus - $TargetAC) / 20 < ($AverageDamage + 10) * (16 + $AttackBonus - $TargetAC) / 20
If that's true, then it's mathmatically correct to Power Attack. Yes, you can simplify that down to an alternate form, but even in 5e there are modifiers that come and go and the target's AC is always a bit of a guess. The bottom line is that -5/+10 encourages players to stop playing the game in order to do some algebra. That is not conducive to quick play, and quick play is one of 5e's design goals. That alone makes it a badly designed feat.
2. It's not situational enough. Or, rather, it's applies to too many situations. Most feats offer abilities that are only good some of the time, or offer something that's always good but on a limited basis. -5/+10 is good enough often enough to come up
every combat and often good enough
every round. And that's
before we look at the other abilities of the two feats. And if you can get a high attack bonus such as from Archery style, or get advantage consistently such as a Barbarian, you just always use it every combat all the time. That's silly, and it's supplanting class abilities.
-------------------------
Oh, and for what it's worth, the whole, "Feats are optional rules. If you find them overpowered then just don't play with them," argument: I'd just like to point out that this is essentially a rephrasing of the Oberoni Fallacy. Just because you don't have to use a rule doesn't mean it's not broken. After all, the exact same argument can be used to defend a feat which says, "You gain +2 Strength or Dexterity, up to the maximum of 20. Your melee and ranged attacks deal +10 damage." If WotC published that feat in a book, you'd hear a very large number of people of complaining that it was overpowered, game breaking, or otherwise horrible game design. Being optional doesn't make a rule immune to critique or criticism. As far as rule balancing is concerned, not using an optional rule is functionally the same as banning a not-canonically-optional rule or otherwise employing Rule 0 to modify a rule.
Indeed, the epitome of optional rules -- the 3e prestige class -- is probably the foremost counter example to why optional rules need to be balanced. We all know that 3.x prestige classes varied wildly in power, and created a power creep that the game never recovered fromm. Even 2e suffered from this, with kits like Bladesinger.