This parry effect you're describing sounds like some kind of force wall you're lumping in front of you. I think it would have to exist on the solid layer so it can absorb effects and operate independently.
I'm saying that a bullet would have to hit the parry block attacked to the ship 2 turns after the block is activated to be reflected
Sounds complicated. How would you figure out what it does and how would the enemy AI be able to predict several turns in advance to use it? What happens if you just spam it?
As an example: The ships spam astro-fac for two turns when they use it. I found that simply reacting wasn't enough to block a threat - the gun that made the bullet is still aimed at the ship. But add a load of other blocks and the ship starts tooting it's guns like crazy, just to feed the astro-fac's demands.
I'm probably misinterpreting your idea, but factoring multi-turn events into an AI that's looking after a group of weapons that do different things is a world of nope.
yeah, i should have considered AI in my idea.. Speaking of AI, what if you made a real AI? Like an AI with a neural network?
Most of AI is about assigning a score to your goals. You need to know what that score is before you can apply a solution to it.
The game currently uses several heat maps generated by "threats" and acting entities. The ships poll their blocks and react to immediate threats and opportunities.
The main flaw is that they act from the core and move towards your core. I need to add getting the whole glom to look at each direction it could move as well as what advantage is gained by firing. I'm thinking of a democracy. That's more or less what's currently happening but a lot of blocks have veto, and just override any other concern. By building a score for each block's possible actions, I could then adjust the vote based on which blocks have the best utility.
This is something I'd have to do before even considering a neural net because I need data to put into that neural net. I doubt I'll use a neural net, I wouldn't know where to start - and I say this having programmed several classic neural nets in the past.