Red is not My Favorite Color

[Toranth]

Maybe I need to feed it more raw meat.

I agree. Hacking response low and very low is practically ignorable.

Sort of.  An unattended hack at Very Low currently produces between 500 and 700 fleetships, while a merely Low will produce somewhere around 1000-1500 (this is where other things start happening, like Waves, Starships, etc, which messes up the count).
A single spawn at Extreme, like if you let a SuperTerminal run a few seconds too long, can produce results like this:

Code: [Select]

spawnedType:AIRaidStarshipV quantity:116289 (strength-each:64)

Maybe there should be a difficulty-factor in the Hacking response?  That way, players like Faulty to whom 500 or 1000 threat isn't even noticable can get a nastier response, while the 7/7 players don't have to deal with crazy "Game Over" results.

[keith.lamothe]

Maybe I need to feed it more raw meat.

I agree. Hacking response low and very low is practically ignorable.

Worth a shot, at least.  Though as Toranth notes we don't want to make it drop premature Game Over moments on folks playing in the 7 to 9 range.  But iirc there's already a strong difficulty-based factor in the hacking response magnitudes.  Will check when I have time.

Syncing the reprisal wave was a nice try, but ultimately fruitless.

(disembodied sound of fingers snapping)

Double wave arrives and is repelled, though the defenses weren't completely back up, and I took a bunch of casualties, including the spire refugee outpost, sadly.

Ah well, the refugees couldn't have been anything important, after all.

FS exo imminent. My defenses/fleet can probably handle it just fi -

(hopes up)

I recognized that thought process from basically every time I've lost to an exo, so I built the mkII and mkIII implosion artilleries, and some rams/shield bearers.

Drat.

The home command station had red health when it was over.

It's not a real exo unless your CIC is open to space.

[Toranth]

Maybe I need to feed it more raw meat.

I agree. Hacking response low and very low is practically ignorable.

Worth a shot, at least.  Though as Toranth notes we don't want to make it drop premature Game Over moments on folks playing in the 7 to 9 range.  But iirc there's already a strong difficulty-based factor in the hacking response magnitudes.  Will check when I have time.

Right now the only place difficulty seems to appear is as a linear factor in the base-strength of the response.

From my current Diff 9 game:

Code: [Select]

spawnStrength = max(1,(AIDifficulty * Handicap)) * 2.5 = 22.5For diff 7, it'd be 17.5, for diff 10, it'd be 25.  7.5 points base difference, so diff 10 is only 40% more difficult.
Aside:  Does the TDL/Blade and Riot cheese factor do anything any more?

If you wanted to really ramp it up, you could increase the number of wild-rolls with difficulty - that's a superlinear factor, usually.  Right now its one plus one roll per 50 points of HaP spent.  Changing it to one per 25 on Diff 9 and one per 10 on diff 10 would be a very dramatic increase in response power.

[Faulty Logic]

===Operation Lay Siege====
Just a few more things to do before endgame.

I quickly rebuilt, then sent the fleet out to conquer the last core shield generator, the ever-mildly-annoying E-network.

Hacked a fabricator for mkV teleport raiders.

Captured another artillery golem, and escorted him home.

I'm out of money.

New round of exos. I'm starting to run low on asteroids, but no significant casualties.

Hacked a StarIV constructor with some difficulty, and spent the last of the hacking on 3000 more k.

Repelled the golem exo with ease. I'll want another batch of disposable spirecraft for the upcoming FS exo, though.

Offensive assets:

Shield bearers, sentinel frigates, zelecs, and fireflies to mkIV. Triangles and nanoswarms to mkI. Also full merc caps, assorted reclaimed fleetships, and 20 gatlings.

Flagship, Zenith, Spire, Enclave, Corvettes, and Riots to mkIII(IV). Leech starship mkII.

3 cruisers, 7 destroyers, 28 frigates from FS.

2 fully armed and operational champion fleets.

The resistance fleet.

Assault transports

Teleraids, telestations, tigers mkV. Decoy drones. One cap of bulletproofVs and translocators, but I lost their actual fabricators.

Golems: 2 hive, 2 artillery, 1 widow, 1 cursed.

Spirecraft: 7 penetrators (mkV, 2 mkIV, 4mkII). I'll probably build more before the final attacks.

I'll wait for the 85% FS exo (and 7.022 to fix the UnhandledErrors), then whittle down the first homeworld with teleport/jumpship strikes.

Then I'll get everyone into a massive pile and smash the other one.

[Faulty Logic]

After repelling the exo (this one put most of it's strength into a mkV H/K, so it was pretty easy to implode it), I repaired and sent the fleet out to the former archive world up top.

And then started cranking out core teleport raiders. It took a while, but I destroyed all the guard posts.

Called the fleet back to repel the next set of exos, not quite synchronized. Then the CPA.

Triggered the FS reserve exo, and murdered it when it came to my homeworld with a huge batch of one-time spirecraft/double hive swarm. And all the fixed defenses and normal fleet at home, of course.

Then I took everything, including the golems, to smash the other homeworld without issue. First one was killed by a spare penetrator II.

Victory. This is my first red victory since way back.

Unfortunately, I realized I didn't actually save before exiting, so no screenshot or save this time.

[Kahuna]

Then I took everything, including the golems, to smash the other homeworld without issue. First one was killed by a spare penetrator II.

Well that was easy.

Unfortunately, I realized I didn't actually save before exiting, so no screenshot or save this time.

Hmmm so you don't have any evidence.

jk I believe you. Good job. I'd like to win my favorite AI combo: AI 1: Raid Engine/Oneway Doormaster and AI 2: something. Raid Engine/Oneway Doormaster / Alarmist/Special Forces Commander and then enable the Hunter plot so when I destroy a Special Forces Guard Post I get attacked by Hunter/Killers! It will be extremely brutal and I'll probably die.. but I want to win 10 difficulty Raid Engine at some point.

[keith.lamothe]

Congratulations!  That was a lot of brutality to fight through.

Beating 10/10 with reds... sigh

Seems like the AI just needs bigger hammers at this point, but the hacking response in particular did stick out as needing enhancement.  Anything else, specifically?

[Faulty Logic]

Seems like the AI just needs bigger hammers at this point

Agreed. Waves, CPAs, and the Golem/spirecraft exos could use some more punch. Especially waves against a single choke.

Anything else, specifically?

Hmm, a few things.

Hybrids suffer from outdated stats.

Normal champion response should be cranked up some.

The strategic reserve should grow past its start once AIP gets really high.

AI ships will often stop moving to duke it out with enclave drones. Ion cannon will shoot at enclave drones before other fleetships. And enclaves are still a bit too strong. Drones every 7 seconds?

Gravity turrets should be a near-number one target priority for the AI. I've had ships in molasses that were shooting at enclave drones even when they were in range of the grav turrets. We might want further gravity nerfing, but I'd like to see what that does.

Bug: threat/threatfleet apparently don't consider neutral planets in their targeting. Which means that the best defense of a capturable is an empty system.

[Toranth]

The strategic reserve should grow past its start once AIP gets really high.

This is certainly in need of improvement.  Right now, in a diff 9 game, the Reserve has a cap of about 12,000 strength all game (since AIP is fixed at 200 for non-Lazy).  By the time AIP hits 200, the Special Forces are throwing around 40,000 strength. 
For Diff 10, it's 24,000 for the Reserve and about 80,000 for the SF.

Those numbers should be reversed, if anything.
Of course, I also think the Reserve spawned units should match the tech level of the planet, rather than being Mk V always.

AI ships will often stop moving to duke it out with enclave drones. Ion cannon will shoot at enclave drones before other fleetships. And enclaves are still a bit too strong. Drones every 7 seconds?

Enclaves don't actually do that much damage.  A Mk I enclave can have 5 spawns alive at once, for a maximum DPS of 57,000 per unit, and a cap of 114,000 DPS.  In practice, you only get that if your target is directly on top of the Enclave - Even a short range travel time of 3-4 seconds is about 12-18% reduction in damage by the drones.
Compare that to other starships:  About the same as Raid starships, but without any of the bonuses.  Slightly better damage than Leech and Flagships, but again, no reclaim or attack bonuses.  40% less damage than Plasma Sieges.  Much less than half the Bomber starship DPS, and barely third of a Spire Starship's damage.  Aka, it fits right in the middle of the starship range.


Since damage isn't really their thing, Enclaves have two purposes:
1)  Free units you can throw into Guardpost/Fortress/whatever range to get off 1 shot before dying, that won't cost you any Metal or Reprisal.
2)  A mass of disposable targets that mess up AI targetting priorities.
Note that 2 actually works against human units, too, if you ever attack-move or FRD when there are Enclave Guardians around.  Humans can redirect, though, when the units act stupid.

1)  This is a very long, slow process if you really want to 'siege' an inconvenient fortification.  Taking out bad pair, say a Fortress and MLRS under glass, can take Enclaves litteraly hours.  Some set ups, like most Bunkerer and many Rude Gestures, are so fortified that even a full cap of 9 Enclaves can't do anything.

2)  This is a usually a mixed benefit.  Drones can soak up shots, but killing drones lowers the Enclaves DPS - a drawback unique to this unit.  If you made the AI units completely ignore drones, then kiting and attacking fixed units would be OP.  Otherwise, no matter what, the AI will waste shots on disposable units.

[keith.lamothe]

Normal champion response should be cranked up some.

Yea, I was in that code recently to copy a starting point for the alt-nemesis stuff and I was like "wow, this is being really gentle".  At least in terms of how much worse it is on diff 10 than diff 7.  That is to say, not by much.

The strategic reserve should grow past its start once AIP gets really high.

Meaning that it should gain something once AIP goes past 200 (on lazy it acts according to min(200,aip), on non-lazy it just takes 200 all the time)?  In terms of defensive assets, or offensive ones?  I'd like to avoid having it just spawn an impenetrable wall of MkV stuff on the homeworld.  Though I guess you'd just warhead it

But having it redirect that into wave strength like the reinforcements would be a way of making the AI still get full benefit from the extra AIP without potentially creating a stalemate.

That said, redirecting into waves is probably not going to seriously threaten a good player.  I could just have it stash the extra strength away for the next CPA.  That might be too brutal, but I guess we could find out.

AI ships will often stop moving to duke it out with enclave drones.

Do you have a save where I can observe this?  There's some pretty explicit code to make it prioritize just about anything over a drone, and debugging it will be much easier with a live case.

Ion cannon will shoot at enclave drones before other fleetships.

Ah, yea.  I should just make the drones immune to insta-kill so it doesn't even bother trying to kill that which is going to be dead in a few seconds anyway.

And enclaves are still a bit too strong. Drones every 7 seconds?

Let's try to make the AI a bit smarter about them before we do that.  As Toranth says they're not really all that powerful, they just play merry havoc with the other side's behaviors.

Gravity turrets should be a near-number one target priority for the AI. I've had ships in molasses that were shooting at enclave drones even when they were in range of the grav turrets. We might want further gravity nerfing, but I'd like to see what that does.

Yea, it'd be good for them to prioritize grav turrets in some way.  I typically try to avoid giving very broad ham-handed "this is what you should kill" rules (though I don't mind ham-handed "this is what you should ignore" ones as much) as that can lead to very stark and undesirable patterns of emergent behavior, but I've thought for a long time that gravity was far more effective than it should be due to the AI not taking it seriously.

Bug: threat/threatfleet apparently don't consider neutral planets in their targeting. Which means that the best defense of a capturable is an empty system.

Hmm, do you have a save where I can see this?  The intent of the code is very much to consider all planets (or at least all non-AI planets, and sometimes even AI ones), so I think there may be some case-specific subtleties at work.

[Faulty Logic]

For Diff 10, it's 24,000 for the Reserve and about 80,000 for the SF.

Those numbers should be reversed, if anything.

I disagree. It makes sense for the SF to be bigger. The SF defends everything, including the homeworlds, while the SR is just a homeworld bonus.

Enclave drones do cost you a little reprisal.

Aka, it fits right in the middle of the starship DPS range.

But it has planetary range.

Otherwise, no matter what, the AI will waste shots on disposable units.

Yeah, that's not a problem. It's a problem when the AI stops moving toward my fleet to fight a neverending drone stream.

Do you have a save where I can observe this?  There's some pretty explicit code to make it prioritize just about anything over a drone, and debugging it will be much easier with a live case.

You can see a mild form of this behaviour in the attached. At the core world, the enemies will stop and start, shooting at drones, rather than beelining to my fleet (where photon lances await them).

It looks like they forget about their old target when taking potshots at the drones.

I'd like to avoid having it just spawn an impenetrable wall of MkV stuff on the homeworld.

I don't want that either, but it has a long way up from its current state to "impenetrable." And yes, warheads (though, interceptor).

Hmm, do you have a save where I can see this?  The intent of the code is very much to consider all planets (or at least all non-AI planets, and sometimes even AI ones), so I think there may be some case-specific subtleties at work.

Yeah, the attached has this as well. Plenty of threat, and a completely undefended core turret controller (Monolith, the P9 planet).

[orzelek]

Making AI prioritize grav turrets will be a bit of a.. contradiction. Those turrets had at least one wave of buffs so that they actually survive to have some effect - and now they would get focus fired ?

[keith.lamothe]

You can see a mild form of this behaviour in the attached. At the core world, the enemies will stop and start, shooting at drones, rather than beelining to my fleet (where photon lances await them).

It looks like they forget about their old target when taking potshots at the drones.

Turns out they figured the other stuff wasn't worth attacking due to being waist-high in Decoys.  I've given them a pep talk that it's better to hurl themselves upon the photon lances to maybe get a shot into one of the Decoys (that will be replaced by the MSDs anyway) than to shoot at enclave drones that will be dead in seconds anyway.  In retesting they seemed much encouraged.

Yeah, the attached has this as well. Plenty of threat, and a completely undefended core turret controller (Monolith, the P9 planet).

Huh, quite right, it really wasn't considering neutral planets.  It will now.