Gaps in the wall

[Faulty Logic]

A list of tactics which rely on deterministic AI behaviour:

Cross-planet waves, raid engine waves, and ships from popped carriers split up and arrive in small groups.

Deactivating the ff over a command station causes AI ships to go toward it, this combined with a low priority target elsewhere in the system means you can endlessly kite the AI by turning the ff on and off.

(Human) Homeworld-related problems: waves against the homeworld never escape to become threat. They also shoot at the massive ff stack over home command instead of the hundreds of turrets that are destroying them.

Threat: bunch of mantis issues about this one, it is the current most visible, most predictable AI behaviour, and thus easy to exploit in a number of ways, from baiting, to permanent paralysis, to getting them to stream in a few at a time to AI systems with your waiting fleet.
http://www.arcengames.com/mantisbt/view.php?id=7016
http://www.arcengames.com/mantisbt/view.php?id=5001
http://www.arcengames.com/mantisbt/view.php?id=9396
http://www.arcengames.com/mantisbt/view.php?id=9465
http://www.arcengames.com/mantisbt/view.php?id=9466

[Kahuna]

Would you like to have some wine with your cheese?

[TechSY730]

Cross-planet waves, raid engine waves, and ships from popped carriers split up and arrive in small groups.

Since these all spawn a bunch of freed ships at the same time, this could be solved by making their spawned orders (in this case, go towards a planet) be group moved together, or at least have a high chance to. If something interrupts them, then they can fall back to standard threat behavior, but at least they will approach in a way that is harder to take out piecewise.

As I mentioned elsewhere, when the sources of ships come from disparate sources, finding a good "grouping" for fleet management is actually a difficult problem.

EDIT:

Oh, and thanks for making a thread about the big remaining ways to game the AI. These sorts of things have been bugging me for quite a while.

[TechSY730]

Oh, can I add a new one as well?

9467: Have the AI consider firepower of planets adjacent to the planet they are considering to enter
This should make it harder to bait threat through simple techniques like just moving ships to the planet next door.

[Faulty Logic]

Yeah, at this point it shouldn't be suprising that any AI-related bugs will be tricky to fix. Natural selection in action.

Oh, and thanks for making a thread about the big remaining ways to game the AI. These sorts of things have been bugging me for quite a while.

You're welcome. Can you think of other AI-behaviour-related problems?

Edit: ninja'd.

Yes, new suggestions/implementations/etc. are welcome.

I think logic similar to the exo-speed cheat should trigger when ships are deployed from a carrier, except going with the slowest.

I think AI ships should only retarget every 20 seconds or so, to avoid endless kiting.

For the human homeworld, I have no idea, but allowing the humans to have additional cheese-potential for the additional frequent AI-fleets-on-homeworld is fine.

For threat, these mantis issues sound easy to implement (as most things do, until you try).

[TechSY730]

Just FYI for those following this thread:

There's not a whole lot I can do about the underlying behavior routines without ripping up a ton of code; the actual "what do I do?" logic for normal AI ships is largely untouched from like 2 (nearly 3) years ago, and I didn't write it.  In general it does a very good job at providing a "bedrock" of challenge upon which I can build crazy stuff like hybrids, exos, special forces, etc, but tampering with it specifically is something I only do sparingly.  That may change in the future but for now I'm looking for lower-hanging fruit.

Understandable. These sorts of behavioral changes tend to be a big undertaking to implement, and with the expansion work going on, I understand why much of this will need to be postponed for now.

Still, thanks for sharing why not much progress has been made on the freed ship logic behavior front; messing with the relevant code seems like it would be a much bigger undertaking than any of us players were expecting.
I was wondering about the odd silence on these types of issues, so thanks for giving us an update.

[Kahuna]

JUnderstandable. These sorts of behavioral changes tend to be a big undertaking to implement, and with the expansion work going on, I understand why much of this will need to be postponed for now.

It's ok. I have a hotfix for this: Just don't exploit these "gaps".

[orzelek]

JUnderstandable. These sorts of behavioral changes tend to be a big undertaking to implement, and with the expansion work going on, I understand why much of this will need to be postponed for now.

It's ok. I have a hotfix for this: Just don't exploit these "gaps".

That hotfix sadly won't work if you want to really play 10/10 and add new bug report

[Kahuna]

JUnderstandable. These sorts of behavioral changes tend to be a big undertaking to implement, and with the expansion work going on, I understand why much of this will need to be postponed for now.

It's ok. I have a hotfix for this: Just don't exploit these "gaps".

That hotfix sadly won't work if you want to really play 10/10 and add new bug report

It does work. That's why 10/10 still needs a buff. I'll win my current double 10diff Special Forces Captain game (patch 5.076) next weekend. Without cheesy stuff.

[TechSY730]

JUnderstandable. These sorts of behavioral changes tend to be a big undertaking to implement, and with the expansion work going on, I understand why much of this will need to be postponed for now.

It's ok. I have a hotfix for this: Just don't exploit these "gaps".

That hotfix sadly won't work if you want to really play 10/10 and add new bug report

It does work. That's why 10/10 still needs a buff. I'll win my current double 10diff Special Forces Captain game (patch 5.076) next weekend. Without cheesy stuff.

Agreed. 10/10 won't really be "fixed" until these sorts of issues are solved, or at least enough so such that it doesn't make major parts of the game trivial. Sure, you can buff the AI, but buffing the AI enough through "MOAR" to overcome all of these limitations without changing logic at all would require so so much buffing it becomes just plain unfun.

EDIT: Some of the changes, like the special forces change, are indeed making steps to a smarter AI. These sorts of things can fix these issues "enough" until a more thorough "code digging" can be done.