The game's up for download now over at Itch.io: https://mcneja.itch.io/sword-horde
I marked it as an Incomplete in the 7DRL challenge because it is very short and doesn't feel much like a Roguelike. The closest thing to it is probably Unexplored, an action Roguelike with a similar top-down view. Unexplored has some nice dungeon generation techniques. I felt like its combat was underwhelming, though.
Sword Horde was an attempt to work on group movement and formations. I didn't get very far on that because I needed a core combat mechanic, which is something I'd been putting off for a while. It was good to spend some focused time on developing the core though. With some tuning I feel like this could be a workable base for a bigger game.
I had talked with a friend about Dynasty Warriors since that is the main franchise I can think of that handles a hero versus an army. I have played a little bit of Dynasty Warriors over the years but probably not more than a couple of hours. Consequently I was working mainly from my own imagination and from previous combat prototypes that I've made.
Some of the ideas that were on the list and didn't make it in:
Enemy types. The obvious things: a boss who's bigger, slower, hits harder, has more hit points. A guy with a shield on one side. Projectile launching guys.
Formations. I was thinking the guys would be in groups of three or four, with a marked leader. Any group with a leader would attack in a pattern. If you kill the leader the guys attack in a less coordinated fashion until they can hook up with another leader. Whether patterned or chaotic attacks would be harder to handle is unclear; ideally the former though.
Combos. This is the core of Dynasty Warriors and I did not get this in at all. It would have been relatively easy to do a combo meter and have some area-of-effect attacks that you can discharge when the meter fills to certain points.
One of the things that has bothered me about these kinds of games is that the player has free run of the whole battlefield. There's very little feeling of enemies denying access to spaces. My initial experiments were aimed at trying to produce a well-defined territorial boundary that would move as you fought the battle. In the end I scrapped it and let the allies run around on the battlefield with you.
I suppose one of the problems with having a well-defined front is that it reduces the dimensionality of the encounter space. In a 2D-surface-based game you have a line where the action is occurring. With ranged weapons that might be a little less of a hard line, but it's still a smaller space for action, and at any given time you're not using the whole space. I think the idea of a battle front is still useful, though, to try to guide the player toward the action.
Figuring out how to make armies look like they're fighting but still keep the player as the driver of the action is an interesting challenge. In the little bit of Dynasty Warriors I've played it felt like they were lazy on this front. Guys are mostly just standing around when I come upon them.
Combat targeting is a hard problem to solve. I have some ideas of improvements that could be made but it's hard to know what will work until you try it. The characters kind of flip themselves 180 degrees if they think it will help them swing at a better target.
Combat flow is hard, too. Sword Horde makes use of Gabriel graph structure to decide who is eligible to do a running-charge attack at the player. Only people who have an unobstructed sphere of space between them and the player are allowed to attack. Once they have started charging they keep charging, which can mean they are trying to find their way through crowds to get to you. But the initial decision is based on clearance.
The biggest problem, I feel, is that it's hard to predictably avoid getting hit accidentally while swinging at someone. I did a lot of tuning—adjusting elasticity and friction, mass and moments, sword lengths, attack timings, attack steering directions, stun durations—to try to come up with something that would reliably separate the characters after a hit. It's not really there yet. It seems like maybe it's close. All of the afore-mentioned parameters interact with each other, though. I spent a ton of time instrumenting and testing the charge attack.
You can see some of the thinking in the debug display above: the orange guy is moving toward the pink disc which is offset to the left side of the yellow player (because orange is planning to make a counter-clockwise swing). The player's not moving, so its predicted position (the tiny pink disc) is centered in the yellow disc. The orange guy is predicting that based on where he wants to be and how fast the distance is currently closing, he should launch his attack when the gray disc touches the yellow player disc. If the player were to lunge toward the orange guy the gray disc would expand accordingly, triggering the attack earlier.