I'm just using a wolf and lasher with Unstable Injector and they are fast enough to grab a capture point each on most maps, so I can confirm a high performance speedster isn't required, just something reasonable.
I thought you said you don't do the frigate dance.
Isn't that more like the frigate
grab? Once you get the DP and use it to deploy, you don't have those ships taken away when the capture point is taken back by the enemy, so you just have to get there first with enough spare time to fill the bar.
And that's really the only way to do it without sending the whole fleet to the point, because frigate vs. frigate battles entirely in the AI's hands are indeed nothing but a coward dance. The only real option for making the AI co-ordinate attacks (like a frigate wolfpack, for instance
) is manual command micro, which one doesn't have the command points for. An "assault" order tends to result in frigates breaking off into 1v1 duels with enemy frigates, which would be okay but is of course ruined by the "officer in every ship" problem. This is trivially easy to rectify - you can adjust maximum officers allowed and even their monthly salary modifiers (and even control how many merc officers the AI uses) in settings.json, and
theoretically being able to modify AI behavior via officer personalities should go a long way towards rectifying the 'ol frigate dance. What this does to Alex's intended command point economy I don't know; I don't understand the XP system and how it relates to story points enough to understand if "100% bonus XP" for mentoring officers means the story point expenditure is effectively free...?
Most of the balance issues we've been kicking around since the update dropped are a simple matter of values tuning; in fact we could fix them ourselves by opening a .csv file and editing some values. (The ease with which Starsector can be tweaked is one of its best features.) AI is the one exception to this, as the AI currently lacks functionality to support what Alex is trying to do with frigates/destroyers in the later game. For that matter, the AI doesn't very well support "battle line" formation, which is the source of 80% of dissatisfaction with ships expressly designed for such. I rather suspect that one upgrade could solve both problems; you could call it a "squadron" or something, where ships have a general idea of wanting to fly line-abreast with each other.
(EDIT: This would be even simpler if the player was allowed to designate them. Even better, one could make those designations at deployment, and cancel them in combat, but not re-form them. IRL, squadrons tended to break up in big naval fights as chaos reigned and couldn't really reform until afterwards.) Frigates and destroyers would be biased towards wanting to attack the same target (perhaps with a modifier based on their speed, so SO ships will be more confident in their ability to get in - since they also have the ability to get
out again-) and the same modifier would encourage cruisers and up to instead pick their own target from the opposite number; aka "battle-line targeting." This would not only solve some issues for every size of ship, but also for individual ships in their weight classes. Assuming the AI already compares own ship speed to nearby hostiles when deciding whether to kite or duke it out, when given a slow low-tech hull with no SO mod, it'll probably come to the right conclusion more often or not.
Another, option would be to simply let players dictate "personalities" to ships even without officers in them; kind of like a "null" officer, with officer personalities over-riding that choice. A "doctrine" setting, if you will. The enemy fleet AI could also use this; setting frigate wolfpacks to "aggressive" when they're dispatched to take a capture point, for instance. You could even let players change the "doctrine" setting on the fly if you want, with officer AI overriding it in various cases. Figuring out the right balance of automation vs. player control here would be the hard part; it'd probably be relatively simple to implement - the aforementioned "squadron" idea, on the other hand, would be another matter.
The difference is that there's a lot more factors to consider in combat. Yes, fundamentally it does boil down to "advance to shoot, retreat when threatened". But when it comes to answering the question "should I advance and attack?", the answer requires quickly assessing the current situation, which changes from battle to battle and from moment to moment. "Alright, there's half a dozen ships nearby, and I want to attack that cruiser that's currently having a flux fight with one of my own. My flux is low, so I'm good on that front. That bomber wing over there seems to be going for someone else, and that other cruiser is backing off to vent after getting hammered, so I don't need to worry about those. There is an unoccupied destroyer nearby, but it has no officer and has already expended all its missiles, so I should be able to dart in for a quick salvo to tip that flux fight in my guy's favor and have enough juice left to win against the destroyer when it comes for me."
This is a phenomenal summation of why AI for a game like Starsector must be a nightmare to design - every new possible maxima increases complexity by an order of magnitude. It's nigh exponential. Fortunately we don't need the AI to be "can duel with players" good; the player has enough problems being the only intelligent person (literally) on the battlefield. It even fits lore-wise; anyone who can go and found three or four new colonies, given the state of the sector post-collapse, must have potential comparable to the Lion of Sindria himself. The AI only needs to fight, well,
itself most of the time; for the same reason the player is obligated to take more than his flagship around; numbers matter. The escort AI is actually
good; I tried using it to create "wolfpacks" and it didn't work because the escort AI
properly escorts; covering its charge's tail and flanks and consistently placing itself to block and run off any pesky destroyers looking to ram Reapers up one's tailpipe. It's constraining, for sure, but it's
reliable, and really, that's all you need. So much of real-world military affairs boils down to doctrine because even though it can be constraining, it's
predictable; and you can plan for and around the limitations of it much easier than you can try to control the chaos of battle in the midst of it (which is more or less accurately represented by having limited command points.)
"Fixing the AI" is impossible, but making it suitable for gameplay dynamics desired, I think that's very doable. Not
easy, it never is, but definitely doable. And the simpler the adjustment, the more doable it'd be.