I expected each enemy to have a simple rule variation like in 868-HACK - this one moves faster, this one has a ranged attack. Maybe elemental resistances to distinguish their interactions with weapons - this one is immune to fire, this one takes double acid damage. Playing with weapon positioning, maybe some rules could be positional - this one has a shield on one side, this one patrols back and forth along a row. Spent a while trying out with different designs but once I hit on the 2d hp system it turned out just giving them different amounts of blood and mana was exactly enough.
868-HACK's enemies also have subtle movement preferences distinguishing them but it turned out that doesn't make much sense on such a small grid; there are so few different paths that it rarely mattered and the paths are short enough that you didn't see any one enemy moving far enough to get a feeling for it.
Obsessed with the number 4, I thought each enemy would have 4 hit points divided between the two colours, 1/3, 3/1, 2/2, uh maybe a 4/4 boss or like another 2/2 with a different ability or something, until I wised up and noticed 5 total points nicely gives 4 unique enemies and so that stuck.
Enemies are also distinguished by the damage type they do. The obvious choice is to have half of them red and half blue, so I didn't do that. If there's equal numbers dealing each damage type, each hit point bar is equally much under threat, so you'd always choose to lose the one you have most of. But if one of them is more threatened the tipping point is hard to calculate. With 3 red enemies and 1 blue you'll usually prefer to take the blue hit; at some point that's low enough that you'd take red instead but how do you know where that point is? Well with only a maximum of 4 there can't be many different options but hopefully enough other factors come into play to vary it situationally.
The 4/1 and 1/4 enemies have the most variance. Hit them with the right colour and they always die in one hit, but if you get stuck hitting them with the wrong colour that's 4 hits, pretty bad. 3/2 and 2/3 are more regular, there's a preferred colour but if you get it wrong it only takes one more hit - but they always need at least two so in the best case they're harder. I distinguished the weapon colours by making red weapons generally do more damage and blue ones generally have more effects, meaning the 2/3 enemy ends up being easier to kill than 3/2 (though this depends somewhat on your board), so to compensate I made the 3/2 enemy deal the less common type of damage: it's hardest to kill but its damage is the least threatening.
For board construction, I wanted different reasons why you would put choose to put weapons in different places. Think of everything that could affect positioning. Damage types are a good start, you want those sort of evenly distributed so you can usually reach the preferred colour for each enemy. Maybe two weapons work in sequence (this one curses, this one kills if cursed) so they're good near each other. Possible interactions based on position - Blacksmith Tongs affecting adjacent tiles, Whetstone affecting the row. Ranged weapons - putting these in a corner lets them potentially hit 4 additional spaces, on an edge 3 and a central tile only 2 - but paths more often go through the central tiles so they might get to attack those extra spaces more often. In general corner/edge/center tiles turned out to have very different value. Effects when an enemy moves into a tile. These aren't enough.
I came up with the idea enemies entering through the screen edges (rather than spawning anywhere on the map). Makes more sense with such a confined board anyway, and adds more differentiaton to the positional values of corner/edge/center tiles. Then later I realised - the board has four sides, there are four enemy types, what if each side just generated one type? Felt very counterintuitive to me at the time; enemy spawns are the main source of chaos, restricting them to always come in the same places would just be putting you in the same situation over and over again rather than a lovely roguelike engine of endless variation. But I really liked the way it broke the symmetry of board-construction, putting a weapon on the left now meant something different from putting it on the right because it will be near where different enemies come from, eight times the number of practically different boards because reflections/rotations aren't equivalent. Tried it out and I guess it was totally fine. There's plenty of variety from different wall placements and the timings and orderings of enemy spawns, it doesn't feel like it's repeating situations except when the same enemy type appears a few times in a row - and then each repetition is usually worse than the previous because you've taken a hit.
In fact getting the same enemy several times in a row is sometimes downright nasty when they obstruct you from getting the gem to heal and shuffle, especially because they'll all be hitting the same colour. These situations were responsible for a lot of deaths and often felt unfair. So I thought, what if I just eliminate them? Have a rule that the same enemy can't spawn twice in succession. This felt pretty great. Smoothed out all the difficult situations, the game flowed a lot more pleasantly, you could much more reliably survive and level everything up and get a good score and everything is nice and friendly. So I decided not to keep it. All those knots of difficulty aren't nice but they're an interesting challenge to deal with and that is much more the point. I went with a watered down version: it looks at the last five enemies spawned, if three of those are the same then it makes sure the next spawn isn't a fourth. So you can still have three in a row, which is bad, but then you get a break. Four in a row would be unfair because you can't guarantee not taking a hit from each, and four hits of the same kills you.
A method for inventing card ideas: look at the alternate-universe versions of the game rejected during design and push it into one of those. If you were thinking of giving something away for free but it has value then maybe consider charging for it. So we get the Forbidden Scroll to block double spawns, but it takes spaces on the board (one for each gate you want to filter) and has to be leveled up to work.
At first I didn't communicate the enemy spawn locations and it was pretty interesting how players could develop an intuition for it without explicitly noticing it. Nobody ever saw the pattern, they'd see the same enemies appearing in the same places over and over again but still think it was random. Even when I drew the gates shaped like the enemies to make it clearer hardly anyone noticed. But if I then removed that rule it just felt very unsettling, because the pattern had very much been unconsciously recognised. I thought this was quite lovely and I wanted to keep it as a slow realisation even though it meant people would play for ages without knowing about one of the main constraints driving board construction, I only really decided to put it in the manual because the gaps next to the enemies looked weird without text in them and there wasn't much else to say. Maybe I shouldn't have, oh it was so bizarre how nobody noticed, amazing. Until I wrote the manual the enemies didn't have names either, except serpents for Dominic and minotaurs for the slingshot, I kind of liked the idea that you can only infer these names from clues elsewhere, but again the 80s arcade bestiary made demands on me.
Okay I was thinking of this as kind of an "arcade game" in structure, you're fighting off waves of enemies while trying to get high score, like a shmup where the enemies come in fixed patterns, a preset sequence of obstacles to get through. If the enemy difficulty is tied to something the player does then it's possible to manipulate it and that can get really complicated, I saw the game being quite complicated already so I wanted to avoid that. Classical roguelikes increase enemy difficulty with dungeon depth, allowing the possibility of sometimes staying on a lower level to get kills (meaning experience levels and items) before heading to riskier areas - but they generally have a hunger timer or similar so grinding an easy dungeon for hours to level up isn't an effective strategy. This could have been a risk in Imbroglio. In some ways less of one because the most readily accessible source of healing is to collect gems (= advancing the dungeon level) but on the other hand the choice of "do I want to heal to reduce my danger of dying now but also increase the danger of everything else ever" sounds pretty unpleasant. Healing is already a risk because the new level layout might impede you, but that would be even worse. Alt-healing items would probably become pretty dominant. So yeah all in all it seemed appropriate to have a fixed enemy progression.
I started with the enemy spawns divided into fairly explicit "waves", like maybe four would come at once then a break, then two at a time with shorter breaks between, then one boss, that kind of thing. Very shmup. But playing it out, multiple spawns were very dangerous because you can get flanked (at least until you have weapons that can deal with that, but simplicity dictated that those should be higher level effects) and breaks were very boring because there's nothing to do except trudge over to the next gem. So I kept adjusting the patterns little by little, increasing the gaps when a pattern seemed unfair, reducing them when one seemed too easy, ended up with this huge complicated table of spawn times that I don't really understand because it just evolved out of small changes over hundreds of plays. It starts with gaps of 6-9 turns between enemy spawns and gradually drops, with some wrinkles along the way. There are still some bits with the sense of a "wave" followed by a reprieve but they're squishier.
For consistency and simplicity I chose to keep the same wave pattern for all characters. This made tuning it difficult because the different characters have different power levels at different stages of the game, what is a pleasant challenge for one might be trivial for another and an unavoidable death for a third. Probably nobody would have minded if there was a different pattern for each character but it just went against my aesthetics, I wanted to be able to write the complete rules for each of them in a few words, not like "no red weapons (but the enemies come slower at first then speed up more in the middle but taper off again at the end so don't worry)". I didn't like rudely killing people with no way out so I decided to err on the side of it maybe sometimes feeling slow/unchallenging at the start for some characters. I went back and forth on this a lot, solely end-game focused boards were proving very effective so I kept pushing the early spawns a bit faster so there'd be some pressure to include items that peak earlier, and then pulling them back a little because I started getting too many unfair deaths.
I was worried about players thinking the early game was boring if they kept killing enemies with the nearest strong weapon and not using the space they had to level things in a deliberate way. This would be a problem if when they then died due to a lack of leveled-up weapons they didn't see the connection between that and the time they could have spent to focus kills better. But I think I managed to get the numbers to somewhere that there is enough risk early on to make it interesting, and where there's enough motivation to level things up that you'll take some risks for it.
Was worried about the late game not being challenging enough so I went back to the idea of those 4/4 bosses, put them in as minotaurs. They were red for a while but turned blue so as not to stack with the most common other sources of damage. The curse on death was a very late addition, I just thought it would be nice to give curse triggers a small effect in any board.