Combat procedures desiderata

I've been using a slightly modified version of Eero's combat structure for about a year, and my brand-new attack procedure for about a week, and I like both. Why? What's so good about them? What's so good about any combat procedure whatsoever?

I'll start by narrowing the field a bit. I'm only interested in combat procedures for a war game that's roughly compatible with classic D&D modules. The procedures need to scale from ordinary humanity to pulp action heroes. (Conceivably it could go higher than that, but I've never played a game that is that high level, so I don't really care.) The procedures need to take as inputs the kinds of things we find in a D&D dungeon key and monster write-up: number, defenses, level, special abilities, place in the dungeon ecosystem. It doesn't have to use all those things, but it can't ask for any information outside of those things.

Ok, that's it for hard requirements. Now let's list the things I want to see in these procedures, in no particular order.

• "Realistic" enough that, if something happens, it fundamentally makes sense.
• Respects positioning and circumstantial modifiers.
  
• Unpredictable enough that you can't reliably call the outcome of the fight without rolling any dice. Otherwise we may as well use static damage numbers.
• Can't be decided by one die roll. So it doesn't boil down to "win initiative and win the fight", even if you've got a sleep spell up your sleeve!
  
• Quick. Can't take a million years to resolve things, obviously.
  
• Easy to explain. I have a rotating cast of players, so I'll have to explain it every session or two.
• Consistent with the rest of my dice tech. This dovetails nicely with the previous two points. If you understand how I resolve tasks, you can understand how I resolve attacks, and vice versa.
  
• Operate at the correct level of abstract, allowing for occasional zooms in. D&D, as I play it, is not a game that asks how exactly you hold your sword, how many times you swing it, or where you're aiming. Maybe fighters can get into that stuff with stunts if they're really interested.
  
• Starts large, scales down small. Big fights happen all the time. We need to model those. If we can zoom in to duels and get good, satisfying results there, too, so much the better.
  
• Weighty in-fiction, requiring commitment. You can't just dart in and stab somebody and then dart out. When you get into a fight, you're stuck in one, and you don't know how long that's going to take.
• Doesn't have the "freeze frame" problem of normal B/X, where you'd have to picture one side actually standing stock-still while the other side acts in order to make sense of things.
• Answers the question, "What happens on a miss?". Misses happen all the time, they're more common than hits.
  Conserves the information generated in the die-rolling process. We get excited when we roll high, even when there's no benefit to rolling high. I want to either take away that excitement (boring) or make rolling high mean something. In other words, I want to...
• Differentiate between slight and strong victories, without adding an extra procedure for critical hits. Respect the thrill, and the realism, of a good hit. But it's clunky to add an extra step, I don't want great hits to be resolved differently, I want the resolution to already tell us how good the hit was.
  
• Use armor and dodge as hit avoidance, not damage resistance. HP (which in my game stands for Hero Points) already operates as damage resistance, in that when you run out you get injured.
• Increase damage with fighter level, but far slower than I increase HP. Combat between high-level characters will go on longer and have more shenanigans (stunts and martial arts techniques, most likely) sprinkled throughout. Think of Inigo versus Westley in The Princess Bride.

A fight between two mid-tier Fighters

That's the list. I might write another post soon explaining how I've tried to check all these boxes. I think I've done a good job!