(For purposes of this topic, we are assuming that the RPG uses dice, rather than collision tests, to determine whether the attacks hits.)

Suppose that a level 1 fighter (or equivalent) is attacking an enemy that the developers expect a level 1 fighter to be able to defeat. (Examples of such enemies vary, but include things like generic goblins/kobolds, rats, slimes (if you're talking Dragon Quest), and similar.) When that fighter attacks:
* How likely should the attack be to hit?
* How much of the target's health should the attack take off?

Similar questions could be asked if it's the enemy making the attack.

That really depends on a lot of factors, e.g. is the fighter using a weapon he/she's proficient in, what damage range that weapon has, is the target wearing armour, and how does armour work in the game (reduces damage vs reduces chance to be hit), so imo no single answer can be given.

* EIther the fighter hasn't had a chance to build up proficiency, or the game doesn't have such a mechanic.
* The damage range hasn't been decided at this point, but the weapon being used is either the starting weapon or one that's bought in the starting town with starting funds.
* Armor's behavior hasn't been decided yet, but again assume that the character is wearing starting armor or armor that's bought in the starting town with starting funds.
.

Default 70%, with factors varying overall chance; I'd prefer to deal hits with low damage rolls than just whiffing.

To be a good game, the attacker should be able to defeat the opponent while losing health proportionate to the importance of the enemy. The player should never lose.

By lose, I don't mean lose the combat. I mean that they shouldn't have to repeatedly retry the attack until the dice roll in their favor. I mean that the story should go on and lessons be learned.

In general, fodder should incidentally lose no more than an 1/8 or so player's life. Important enemies do about 1/3. Main enemies should drop shot 1.5x life over combat, allowing for the consumables' representation in strategic play to shine. Ample time and warning need to be given to the consumables.

Again, I think death is a bad gameplay mechanic for most games, especially in the beginning. So finding a way to teach the game and show why it is special and fun should be the primary objective.

Interestingly, I think I feel this way about growth systems; the player's character should improve in some way early in the game. It's one reason I don't like the approach games like Baldur's Gate take, where level-ups are infrequent and you're stuck at 1st level for too long. It's also one thing I like about the SaGa approach; your stats improve frequently as the game progresses, instead of being given to you in big bursts. (With that said, SaGa games do have a learning curve, so it helps the player to learn the system.) Or, of course, healing spells and abilities. There are some games where this works well. While not an RPG, Celeste handles death well; while in Chapter 1, just learning the mechanics, you will die a lot, start the room over (and rooms are shot), and learn with each attempt. I also think death works well in troll platformers, provided checkpoints aren't too rare (and the game isn't too cruel about sending the player back to a previous checkpoint).

I prefer accuracy being a damage scalar rather than a chance to hit. Especially in more tactical games where each turn is precious. I can deal with misses in games where there are a lot of attacks and missing one isn't going to mean a significant loss of momentum or even the battle (such as in most jRPGs). Xcom-type games, and the indie game sRPG Regalia are both totally ruined to me by their accuracy/miss mechanics.

Yes, in tabletop RPGs, I have used "glancing blow" mechanics of sorts ("miss -> minimum damage" or "miss-> strength only, no other modifiers")

Depends on the type of game. If the game is pseudo hardcore, with high enemy damage and high player damage (a la Dark Souls and the like) then no more than 60% or so. If it's more like a typical JRPG which favors weaker attacks in quicker succession to get lots of numbers and those "Dragon Ball Z type fights, then 90% or so.

I'm thinking that going through 3 walls after being punched would be more suitable for an action or puzzle game than for an RPG.

Also, in this case perhaps the only way to get past a spot, or to get into a secret room, might be to get hit by such an attack.

Possibly. D&D has stats for material types and walls so you can punch through them. I was thinking more Herosystem with someone that has a huge strength (40+) where the person may very well be unconscious before hitting the wall...

Or the Atomic fist who by a fluke chance did so much damage that he became infamous even though he wasn't trying...

I think a possible solution would be to have different types of hits, ie glancing hits, regular hits, and then crits, or maybe even more types of hits.

I love how the combat in Wizardry: Proving Grounds of the Mad Overlord feels and scales, so I'll choose that as my perfect example of an RPG combat system.

Let's see if I can work through the system...

Apparently the formula for hit chance is something like:
({CharacterLevel/3) +2} + MonsterAC + (3*Victim) -1) * 5%

So let's try to simulate (roughly) a battle between an average level 1 Fighter and a Bubbly Slime.
Something like:
((1/3) +2) + 12 + (3*1) -1 *5%

translating to about a 75% chance to hit the enemy.

This would be further modified if we had a good strength score (giving us a maximum +15% chance to hit at STR of 18).

If the hit connects, lets say we hit with a typical amount of 1d6 damage vs the slime's 1d3+1 HP.
So that is something like a 50/50 chance we kill it in one blow if our blow connects. (a roll of 4-6 will kill it assuming it has the full 4 HP total)

So a fighter has a 75% chance of hitting the slime, and a 50% chance of killing it in one hit. Interesting.

It would be interesting to watch how these mechanics scale throughout the progression of the game.
I don't know what specifically makes it feel so good, but combat in Wizardry 1 always felt snappy and satisfying to me. I think if you want to make a solid RPG experience, you can't really go wrong with copying Wizardry 1.

And just to add to that, there would be a 37.5% chance of killing it in one hit on any given hit.

There's one factor in both Wizardry and early Final Fantasy combat that's worth noting, and that's the fact that fighters (and in W5 and Final Fantasy, eventually other classes as well) get multiple swings per attack. As a result, characters will miss noticeably often at lower levels, but at higher levels, you're unlikely to miss entirely. You might do less damage due to having some swings miss, but you're at least going to do *some* damage. (This is even more prominent in FF2-FF4 where characters (whether friend of foe) get only a certain number of chances to dodge, so if the attacker has enough hits, it will be impossible fore the target to dodge enough times to avoid the attack entirely. Also worth noting that spells use similar mechanics, though in FF2 you're guaranteed some damage; this guarantee does not hold true in FF3 or FF4.)

People who've played Final Fantasy 4 may not even remember attacks missing, since the main character starts at level 10, and if you have a level 1 character at any point, that character won't stay level 1 for long (the second instance this happens, the character will likely jump to around level 7 or so after one battle).

(Note that this discussion doesn't apply to the FF4 DS remake, or the versions based off it, which changed the accuracy mechanic.)

Personally, I'm more into to-hit mechanics like Dragon Quest and Final Fantasy 5, which actually take level and stats out of the equation:
* In Dragon Quest 1, enemies have a small chance of dodging attacks (1/64 or so is typical I believe, with one particularly shadowy enemy getting up to 1/4, and that particular enemy is annoying to fight), and the player has no chance of dodging. (I personally would prefer it to be possible for the player to dodge, but that didn't happen untile DQ2 with its cloak of evasion.) (Note that Metal Slimes aren't good at evasion; rather, they have defense that's so high that, unless you get a critiical hit, you will deal 1 damage 50% of the time and 0 damage otherwise (but note that this monster has only 3 HP).)
* In Final Fantasy 5, attacks tend to hit by default. Some weapons (notably bows and axes, but not daggers or most swords) have a chance of hitting. When the attack hits, some enemies have a chance of evasion; a party member can evade if they're equipped with a shield, and there are a few specific pieces of equipment (something like 4) that provide extra evade chances. Furthermore, some weapons (daggers and bows, for example) pierce evasion, meaning that the target has only half their evade chances. Of note, in FF5, at the start attacks are not going to miss at all; misses don't come until later.

Thing is, when I hear "knockback", I think action game, not RPG. For example, in Castlevania 1, when you get hit, you get knocked back and up a little bit. This can be fatal (if you get knocked back into a pit, or into the water in Stage 2), or it can be helpful (if done right, you can damage boost onto the wall in Stage 2, bypassing the water area entirely).

I tend to prefer RPGs that abstract away positioning, at which point knockback no longer makes sense as a game mechanic.

Edit: Knockback also makes sense in SRPGs. In Disgaes, for example, the first fist skill is useful for knocking gatekeepers in the item world off the dimensional gates, allowing you to proceed to the next floor without having to kill it,