Military Strategies

[TechCor]

Hey all,

I am getting close to working on military tactics for my AI, and I would like to get some opinions from some of our top strategists here on Outpost Universe.

While many players probably use a "deathball" macro type (mass Thor's Hammer tigers) or rush laser lynx, I actually want to look into some more micro type tactics, since the AI will more or less be able to handle multiple groups simultaneously. The goal is to be interesting, offer rush, mid-game, and long-play strategies and styles of play. The AI can assess combat situations and retreat, evade, and harass, or use frontal assaults if substantially stronger than the opponent.

On top of this, I want to really make use of as many types of units as possible. I am aware that some units are kinda "duds", or become useless with later research due to poor balancing, but perhaps they can be used under very specialized circumstances without sacrificing a win. If someone has deep insight into the specialization that each unit would be good at, please post about it!

Also helpful, if any of you awesome people could send me links to videos of multiplayer games where players are on an even skill level, I would much appreciate it! This would also help give me an idea of base priorities as well as tactics, so that the AI can keep up with people like you! Videos of a variety of play lengths, styles and maps are also welcome.

Tactics

Bomber-Escort Group
About 1-3 Starflare or Supernova with 4-6 EMP (optional).

The idea would be to run this group with other stronger groups to a set distance away from the enemy base. The AI will assess weak points in the enemy base and attempt to move bomber groups towards poorly defended critical infrastructure such as mining and power. If the AI determines that defenses will arrive before detonation, the group will retreat.

This group will typically not attack without other groups making a push, and therefore defenders would likely need to split forces to engage the flanking bomber squadron(s) while dealing with primary forces. This tactic will also trigger if the enemy makes a major attack. The bombers will attempt to navigate around the attacking forces and hit major structures or the CC.

This group will typically run with lights off, but will turn lights on if enemy units are on an intercept course, or it is clear the bombs will reach the target before any defenders can stop it.

Harassment Group
About 3-4 structure damaging lynx.

This is an early game tactic for pissing players off. Instead of directly attacking your defenders, this group will actively avoid them while maneuvering towards other undefended units and structures. The goal is not to kill, but to do as much damage as possible. The group will only retreat if it is no longer possible to attack without taking damage. It will then hang out and patrol around your base, especially between nearby mining beacons. This group can grow in size and begin attacking lightly defended targets, if it determines it will win the battle. If it grows large enough, it will disband and become part of an assault group.

Assault Group
Any number and mix of units.

This is the main attack group. It primarily performs map control and defense. When this group is stronger than other enemy forces, it will engage them. If this group is stronger than ALL of the enemy's forces combined, all other groups are disbanded and join this group for a frontal assault.

If the enemy is splitting forces, this group may break into separate assault groups to attack and defend multiple points. The split groups will always be stronger than the enemy's groups. If splitting the group will result in a group that is too weak, the split does not happen. Instead, the assault group will attack based on the following conditions:

A. If the enemy is on an intercept course with a base, and it is possible to destroy the enemy base before they destroy its own, attack the enemy base, otherwise defend the base.

B. If the enemy is splitting forces, but is not heading towards a base, attack the enemy base is winning is possible, if not, attack the nearest weakest force. Otherwise, find optimal map control position (between own base, enemy groups, and enemy base).

[Hooman]

Don't forget tube cutting 

[TechCor]

Don't forget tube cutting 

I mean, I'll admit the things I listed might be considered cheese tactics, but that's so dirty, it never even crossed my mind!

That led to me looking it up, and discovering SirBomber ranting complaining that Eden sucks. Acid clouds < ESG, Railgun < RPG, EMP missile end games, etc. Interesting, to be sure, but hard to breakdown into tactical descriptions.

It sounded like there were many fun multiplayer games back in the day. If only people recorded some of these...

[Crow!]

You can find one series of 1v1 matches on Pie Chart from a while ago if you look on youtube.

To my knowledge and recollection, the early game is generally defined by the Stickyfoam weapon, which ensures that vehicle losses are minimal, but which can kill any struture with impunity if given enough time.  The midgame is defined by the ESG and its ability to encourage poking, since its effective range against an enemy that isn't paying attention is 9, but it does poorly when chasing into an opponent who is also using ESGs.  The late game is defined by the EMP missile and its ability to break stalemates.  High speed units like supernova lynxes and spiders suddenly gain some extra relevance, but having a tiger backbone to the army is also needed or else you lose any fights where EMP missiles didn't exactly hit their mark.

Eden's window of relevance is extremely narrow.  They can just cross their fingers and hope the enemy doesn't notice a starflare attack in the early game, or they have one moment just after the advent of Thors when the enemy will likely not have EMP missiles yet and armies might still be small enough that the Thor does better than the ESG.  Aside from that one weapon, Eden has inferior DPS and zero tactically interesting options that Plymouth doesn't also have.

Is your target for this AI a multiplayer opponent or a single player opponent, by the way?  If single player, just be aware that when making a stage with an Eden opponent, it'll need to have access to / start with more stuff than Plymouth would to be a challenge.

[Sirbomber]

That led to me looking it up, and discovering Sirbomber ranting complaining that Eden sucks.

I don't appreciate this...  and you could at least spell my name right if you're going to say things like that.

As I suggested to Crow when he started work on his rebalance mod, if you want to do something like this, you should probably get some firsthand multiplayer experience.

[TechCor]

I don't appreciate this...  and you could at least spell my name right if you're going to say things like that.

Sorry about that. I meant it in jest. I guess it would have helped if I linked what I was reading.

Also, this wonderful gem, is so funny that not sharing was a crime worse than capitalizing the "b" in your name (IMO, of course).

As I suggested to Crow when he started work on his rebalance mod, if you want to do something like this, you should probably get some firsthand multiplayer experience.

I suppose getting my ass handed to me a few dozen times would teach me something, but there is a bit of a difference in this case. Mastering APMs and optimizing real-time decision making is something a human has to get good at. Bots, OTOH, can multitask and react better than any human. The difficulty is going to be having the AI adapt to changing conditions.

At Crow's suggestion, I looked a bit more and managed to find Leviathan's stash. I think studying some videos for build orders and unit tactics is going to work out much better. I can play against the AI until I can't beat it due to its superior reaction times, then I can play it against other people to discover its flaws and "train it". It'll be a lot of work, but a lot less than trying to become an Outpost 2 master. I'm not really interested in PvP personally, so no point going that route.

Whether it can compete with pro players is somewhat of a side goal. Currently, I would be happy if it could pick up from any starting condition, survive, build, and attack successfully with virtually no setup by a mission designer. The idea would be that if the AI isn't good enough, just give it more starting resources. Most players will figure out a bot's weaknesses and exploit them eventually anyway.

[Sirbomber]

Yeah, fair enough.  That said, I've mellowed out; bringing up old stuff like that is embarrassing.  Moving on...


When I say "play multiplayer" it's not an invitation to "come get owned".  Being able to read, react to, and predict your opponent is an important skill, especially in a game where you have perfect vision of the map at all times.  A working, practical knowledge of the game will give you more insight than the theoretical knowledge you'd pick up from watching replays (especially since I'm pretty sure 99% of those replays are all 1v1 Plymouth mirror matches on Pie Chart) or playing the campaign/colony games.

[Vagabond]

For multiplayer, I typically just play as Eden and split my tanks 2/3 Thors, 1/3 EMP. Makes it easy to play if not that effective. I always liked the idea of playing Plymouth but trying to be quick enough to take advantage of stickyfoam, ESG, and EMP missiles in tactical situations became tedious for me. More interested in the colony building I suppose. I also like playing with morale on in multiplayer, which is probably strange to others...

-Brett