While working on my beat em up AI vid I realized something simple - it helps to split up enemy design into 3 stages (ignore meters/resources/other stats) :
1. Approach - how interesting/challenging approaching the enemies is. How much do they challenge you using their basic movement? Sometimes enemies play a challenging footsies game, sometimes they require manipulation by the player, sometimes they have evasive states, sometimes they block off your approaches via attacks. They play with space.
2. Direct Engagement - how interesting is it to fight enemies when you're in close range? Sometimes enemies have block/dodge/counter states, sometimes their attacks are really dangerous and require varieties of responses, sometimes they have fast micro-level movement sometimes super armor forces you to carefully consider move framedata, sometimes they have hard coded weaknesses. This is where more discrete mechanics come into play.
3. Followthrough - how interesting is the end of your engagement with an enemy? Can you finish them off for situational iframes? Can you use their body as a shield? Can you throw them stunning other enemies? More importantly, how meaningful is this for victory? Does ending the engagement properly help you proactively create an advantageous situation, or does it more or less return you to neutral?
So to make this more concrete, a game like Gun.Smoke (not quite a bmup but a great example nonetheless) mostly just has #1 but it excells at it because both the player's and enemy's movement is fast, free n robust and the attacka are balanced around it. Final Fight and other 2D beat 'em ups have decent #1 and good #3 - the enemies aren't quite as nimble as Gun.Smoke's but there's a whole throw-game layer that creates really fun proactive play. #2 is almost absent. God Hand has crappy #1 but nails #2 and #3. Getting to enemies is trivial but the fighting game mixups, sidesteps, counter hits create an interesting layer of direct engagement, and you are rewarded with a beat 'em up style throw system if you win the direct exchange. Ninja Gaiden 2 has good #1 and #2 but very situational #3.
You can ask questions about your own enemy designs - do they do anything interesting to make approaching them more challenging? What sorta thing could you add? How could you make direct engagement more interesting? And how does winning a single engagement feed back into the rest of combat?
That said, as usual, there is almost no such thing as a pure representation of any of the 3 "phases" and they bleed into each other, so you have to make a judgement call when separating them. The borders being fuzzy is GOOD in fact, means the games won't feel preplanned and rigid.
Games can work no matter which "phase" they focus on, their focus will give them a unique flavor. From a dev perspective, you should probably choose what you want to focus on based on your preferences. I value #1 and #3 and tend to neglect #2 (and think it's over-represented) so I'm going all in on that. That's what I will advocate for and focus on in my games, but bias aside anything can work in the hands of a good dev.
