AI War 2:The Grand New AI

From Arcen Wiki
Jump to navigation Jump to search

Known Issues

  • Multiplayer is disabled until post-1.0 while we focus on tightening up the single-player loop (more info here).

What's this phase all about?

After lots of feedback, we're ready to take it to the next level! Major new functionality and improvements across the board are incoming to make games more interesting than ever, and we're also going to start building toward multiplayer.

BETA Version 1.101

(Not yet released -- we're still working on it!)

  • In the prior release there was a thing with "growth versus stagnant" multipliers. Those are gone now, reverted to how they were before.
  • Also in the prior release, there were a lot of things that didn't get cap increases anymore, but now they work like strikecraft (and thus like they always did, although with some extra clarity because the cap scaling is still iterative as noted last release).
    • Frigates are unaffected by all this, as their scaling was essentially just evened out and they were still "growth style" units.
  • uses_growth_multipliers_even_if_stagnant has been removed, as the mechanic it was working with is now gone.
  • Added the ability to mod out or deprecate ship lines in starting fleets or otherwise by setting their max cap to 0, which then takes them out completely.
    • Thanks to Flypaste for requesting this ability.
  • Fixed a bug where display_name was being incorrectly blanked out on partial records when reading xml.
    • Thanks to Flypaste for reporting.

BETA Version 1.1 So Much Stuff We Can't Even, But More Is Still Coming

(Released January 1st, 2019)

Hi there! To play this on Steam, please right-click the game, hit properties, and go to the betas tab. Then go to the current_beta option and it will download this. We could really use feedback from a variety of folks as we build in the remaining stuff that we have planned for the 1.3 build that we plan to officially call all this once we're sure it's stable. Please take a moment in particular to try the new parallel processing and see if you can run it and how much smarter all the factions are at high speeds if you play at those speeds.

  • Add the "Extreme" category of Quickstarts, with 2 default, 1 nearly default, and 2 bonus scenarios.
    • Thanks to CyberKun for suggesting the category.
  • Ion Cannons and Orbital Mass Drivers will now be automatically targeted by units again, but are set to a low priority.
    • They still seem to like shooting these first however, if they happen to be in the way.
    • This lets factions destroy these again rather than being shot to bits, ignoring them.
    • Feedback is appreciated - this is a fairly opinion heavy thing.
      • Thanks to Ovalcircle for reporting.
  • Updated the "Various Weapons" fleet name styling with DEMOCRACY_DEMOCRACY's latest changes.
  • Fix a bug where the Random Any ai type could pick 'random easier'
  • The Tech History menu (viewable by clicking the Science icon in the resource bar at the top of the screen) now shows the total science you've spent.
  • The "Show Entity Strength When Hovering" tooltip has been removed, as we are now going to do this by default. It was strange not showing it.
    • Ships now say if they are showing their own strength or the strength of themselves as a stack when you hover over them.
  • Targeting behavior of ships that are freely moving around (pursuit mode, whether of the player or other factions) are now vastly smarter in how they deal with non-combatants in general (things without guns but that need to die).
    • They'll still shoot at them just as frequently as ever IN PASSING, if there's nothing else to shoot at or that thing is the most important priority to shoot at based on the various metrics. But they no longer will specifically pursue those sorts of things with an excess weight just because they happen to be close.
    • If there are non-combatants with a high priority, then ships should still beeline for those just as well as ever, it's worth noting. And if they're not in pursuit mode, then nothing is different at all.
    • The net effect of this is that ships will generally approach and destroy dangerous things like great-turrets before they start messing about with an unarmed guard post or a gravity generator or whatever else. Survivability goes way up.
  • Nanocaust construction centers and Marauder Outposts now warp in over time (a short time interval, like 30 seconds to a minute) rather than warping in instantly. This gives their enemies a bigger window to counter attack in.
  • The Tech Menu now tells you a very approximate strength increase for both defenses and ships after upgrading.
  • Added the ability to hack Global Command Augmenters to re-roll provided turrets. (Thanks donblas)

Optional Advanced: Massively Parallel Faction Processing

  • New setting in the personal settings: "Advanced: Run Massively Parallel Faction Processing"
    • Only matters on the host in multiplayer. This lets the factions in the game be substantially smarter and faster in how they react to things, and is most notable at higher simulation speeds (10x, etc).
    • But all this may come at a cost that your CPU can't handle; we're still testing that and would love feedback on the difference you see in simulation speed (the percent under the game speed multiplier) if you turn this on and off.
    • You will have a much fiercer foe at 10x speed with this on, versus an almost braindead one without it. At regular speed it won't matter much between one or the other.
    • We're hoping to build toward this being the default for everyone, but for now we don't want to inflict slower framerates on everyone just so that 10x speed runs smarter. One step at a time.
    • Right now this is off by default, and needs a lot of testing, but amazingly with the current design it did not slow down at all in a relatively-small-number-of-factions game on Chris's i7-6700 HQ. Much more data is needed.
  • You may experience some strange and awful bugs by enabling this, at least at first, as the factions have never been parallel in quite this way before. This may tank performance or cause lots of strange errors, and in most cases we can find and fix them if you send us the errors or the savegame with the bad performance. In most cases it would be two threads from different factions trying to use the same list or other collection. To report: https://bugtracker.arcengames.com/
    • Thank you!

Fireteams

  • Fireteams are a concept developed by Badger to allow non-player factions to flexibly coordinate a variety of simultaneous actions. It's the smartest strategic AI mechanism in either AIWC or AIW2 (in theory at least). Fireteams were developed to power the Scourge, a feature for the first DLC, but we can use the same basic logic for other minor factions to give those factions a nice intelligence boost. We are backporting R&D done for an expansion-only feature to the base game because the base game isn't hard enough yet. Mwahahaha.
  • Allied ships using fireteams (just marauder raiders for now) will show you what that fireteam is planning on doing in the verbose hovertext, to make coordinating between you and your allies more possible.
    • The simplest description for how Fireteams work is like this. All units for a given minor faction are assigned to small fireteams of 1-4 strength, so there will be lots of fireteams. Each fireteam determines its own objectives independently. If multiple fireteams have the same objective then they will coordinate their attacks.
    • This is currently available for the Hunter Fleet (via the 'Assassin' type in the Game Lobby) or the Marauders (via choosing "Intelligence: Tactician" in the game lobby).
      • The Fireteam implementation for Marauders and Hunter Fleets is brand new, so I expect some teething pains and would love feedback (with save games)

All-New "Adaptive" AI Type

  • Add in a new Adaptive AI type, which alternates between available types every so often. They work similarly to Random, in that there is "Adaptive Easier", "Adaptive Brutal", "Adaptive Any", and so on.
    • Change interval is in XML, and is once per hour. It won't change any existing units, but reinforcements/wave composition and so on will change.
    • There are currently no Achievements for these AI types
    • Thanks to Flypaste for the suggestion

Balance Tweaks

  • Orbital Mass Driver reload time 9s -> 6s.
  • Hack Ion Cannon and Mass Driver now takes 2 seconds, rather than 15.
  • Ion Cannons and Orbital Mass Drivers that the player either hacked or bought now die to remains.
  • Replaced Military Command Stations engine stun with paralysis.
    • Engine Stun on this thing could be fairly exploitable, allowing it to hold off incredible amounts of enemy strength.
    • However, that was only possible if the enemy could be slowed at all. If they couldn't, it tended to fall relatively easily.
    • Swapping it to paralysis should let it not be as strong in the old ideal case, but also improves it in the worst possible case.
  • Dire Nucleophilic Guardian damage 4,000 -> 8,000, shots per salvo 6 -> 3.
  • Reduced range of Tesla weapons very slightly (10% ish?) and bumped up the splash size by 15% ish.
    • What this means is they effectively act more or less the same, except will not waste their initial shots on the first singular ship to come into range, and will tend to hit more beyond.
    • This was split between both the range and splash size to avoid making the visual effect too much larger. Shouldn't have much impact on gameplay, however.
  • Dire Tesla Guard Post now has the fortified effect on it.
  • Hydra heads and the like cost no energy.
    • This was the cause of mysterious brownouts.
    • Thanks to DEMOCRACY? DEMOCRACY! for reporting.
  • Reworked the way that wave compositions are filled to be more efficient and ensure that absolutely as much of the budget is spent of a wave as possible.
    • Thanks to Puffin for discovering that this was not happening.
  • In the normal non-stealth-planet themes, the stealth guard posts are now 1/50th as likely to appear as before. There were vastly too many of these showing up in the galaxy prior to now.
  • Dire Tesla Guardian damage 2,500 -> 5,000, now has the same durability as the other Dires.
  • Reduced range slightly of MLRS, Concussion and Nucleophilic Guard Posts.
    • The main contributors to the "small world problem".
  • Increased the cost of normal Guardians by 50%.
    • This is in preparation of the next update, in which these'll be getting buffs. It's done in two stages to avoid current saves suddenly getting much harder.
    • With this, old games will be similar, and new ones will have the correct Guardian count for when the buffs hit. So for a time, Guardians'll be a bit easier.
    • This is part of making Guardians scarier individually, as right now they can feel a bit flimsy and kind of just too close to Strikecraft overall.
  • Okay, so this is a kind of funny story... cough.
    • Essentially, various ships and turrets and whatever have modifiers to kill other specific types of ships or structures with certain characteristics, right?
    • So something is really good at killing structures in general, or high-energy-usage stuff, or items with thin hulls or whatever else. Easy enough.
    • The problem was, we had these ships getting BETTER at killing those same things at higher marks, while also getting a general damage buff.
    • Puffin's example, as he was the one to realize this:
      • MK1 Pike Corvette does 45 damage, 3x to high armour, 2x to high hull remaining, for...270 damage.
      • MK6 is 189 damage, 5.5x to high armour, 3.5x to high hull remaining, for...3,638.
    • As he put it, "I feel like I know where the "late game is easier than early" problem is."
    • We may need to rebalance some of these units to generally have higher specialized damage or higher damage in general, BUT, we've now removed the increase in damage multipliers from the following categories of stuff:
      • Strikecraft (yours and enemy of course), Turrets (ditto), Frigates (mostly yours), Guardians (mostly enemy), and Guard Posts (mostly enemy).
    • We've LEFT this monstrosity of multiplication on a few specific things:
      • The Rorqual Hegira Ark (yours), AI Fortresses (enemies), and any battlestations (yours) with weapons.
      • This may be something we want to give to a few other special units of the AI or you, we'll see, but it would be specifically for Very Special Units.
  • Metabolizing Gangsaw armour 60 -> 90, reduced count found in ARS and Fleets.
  • Consumer Starting Fleet now uses normal Gangsaws instead of the Vicious variant.
  • All normal Turrets (i.e not Plasma or Beam) have armour of 80mm.
  • Doorkicker Starting Fleet now uses Vanguard Hydras instead of the normal one. Also has 25 instead of 20.
  • Nucleophilic Guardian speed 600 -> 1,000.
  • The "singular freaky surprises" of AIs now will only appear on planets between marks 2 and up, rather than every planet owned by the AI. No surprises TOO early in the game, then, with golemite and similar.
    • Thanks to Puffin for suggesting.
  • GCAs now have a higher chance of having an extra "OtherDefense" item at the end.
    • Thanks to donblas for a suggestion in this area.
  • Reduced the cost a bit of Weaken Turret and Weaken Guard Post hacks.

Ship Cap Calculation Refinements

  • The way that the caps for ships are calculated is now vastly different, and a bit more expensive on the CPU (but worth it).
    • Essentially before it was using a floating point number to get a result, and then rounding up. So very often this would lead to things not actually increasing in a manner that you'd expect, or frigate counts not increasing at all for several marks before finally increasing.
    • Now it applies each multiplier sequentially to the original number, and then rounds up each time in between, so you always get the expected result.
    • This means that, for example, the mark 3 frigates of things with cap 1 now have a cap 3 instead of 2. The end result is usually pretty much the same at mark 6 and 7, or close, but the intervening marks are now more accurate and we shouldn't have people with lots of confusion on their caps not increasing like they would expect.
  • There are now three different ship cap growth styles, rather than one:
    • Stikecraft (works about the same as it always has, although will give you slightly more now, working as expected because of iterative math).
      • This was already pretty much working fine.
    • Frigates (much lower growth rate than before, but ends up in a similar place numbers-wise, however gives you what you'd expect rather than something surprising now).
      • Numerically this was balanced fine, but people couldn't understand why it jumped around strangely because of rounding with fractions.
    • Everything else (does... not gain cap by leveling up).
      • Wait, wut? Hold on a moment, though: one of the most annoying things ever is having to go back and put out more mines or turrets later on on all your planets because you just increased your caps. That still will happen from GCAs, but it's at least less common because general techs won't cause this to happen.
      • Besides which, performance can be a real pain with turrets if there are just excessive numbers of those. How about we increase the strength without that? Same result, but game runs better and also less micro for you. For more about that, see below.
  • On the mark levels, rather than just having a multiplier for hull/shield points and attack, it now has a "growth" one and a "stagnant" one.
    • "Growth" are the same numbers as before, and apply to frigates and strikecraft. Aka nothing changes as these have both a power/hull increase AND a ship cap increase as they go up.
    • "Stagnant" are used for ships that used to get a mark level bonus to ship cap, but no longer do. So to keep them equal in power while smaller in number, it includes their hull or attack power multiplier multiplied by what their old ship cap multiplier would have been.
      • The end power on these is absolutely identical to what it used to be, but it's for turrets and similar where they are now not something you need to go in and place extra every time you upgrade a tech that improves turrets. See the above notes.
      • This obviously does also apply to things like tractor beam arrays and minefields and so forth, and given that AIs don't use the human-style caps for those that does mean that those are just plain stronger now on AI planets when they are (rarely) used.
        • If we want to change that, we can make AI versions with the uses_growth_multipliers_even_if_stagnant flag from below, but that doesn't seem super important for anything but turrets.
  • Added a new uses_growth_multipliers_even_if_stagnant flag which, when set to true, causes something that would normally grow attack bonuses and hull bonuses per mark level as if unit cap was stagnant (thus larger bonuses) and change it to grow as if it had an increasing cap (thus smaller bonuses).
    • This is used for the Great-Turrets so that they don't get crazy stronger at higher mark levels, since the AI doesn't use normal human-style caps anyway.
  • Any fleet leaders (golems, arks, battlestations, citadels, etc) now act as though uses_growth_multipliers_even_if_stagnant is true. Thus they'll scale as they always have. Same with guardians, guard posts, command stations, king units, and drones. Also LargeShips and SmallShips that are used by third party facions.
  • All of the AI structures now explicitly have uses_growth_multipliers_even_if_stagnant on them to avoid them getting accidental crazy buffs of any sorts. This mainly benefits AI Fortresses and the like, which could have been very scary otherwise.
  • Fleet Capacity Extenders now only work for strikecraft, as their increasing a frigate could be absolutely crazy in terms of the impact to both your power (in a good way) and your economy (in a bad way).
  • Stationary planet-bound forcefields, tractor beam sources, gravity sources, and tachyon sources now upgrade in the same way that frigates do (aka they behave more or less like they always have, but with slightly more clarity -- they get more unit cap as they go, and the lower "growth" style stats increases instead of the higher "stagnant" ones).
    • People do love having more forcefields, that's for sure, and the ability to field more tractors and so forth are similarly not something we want to condense down like we would turrets.
  • Minefields now have their own special_entity_type and also a rollup for easy iteration over them. Some clever AI routines may need that sometime in the future.
    • Additionally, minefields now continue to get a larger ship cap at the strikecraft-style rate of addition, and regular growth stats. Got to have enough for coverage, of course!

AI Great-Turrets

  • All of the AI turrets were already different from the player turrets in that they generally have half the range of the player counterparts. Now we're taking it even further.
    • Mainly because turrets can't stack, and thus will grind the game to a halt if the AI builds full caps of them at the sort of strength and costs that we had them at previously, we've condensed them to being essentially like stacks-in-one right from the start.
    • All of the AI turrets are called Great-Turrets instead of just turrets, and they have 5x the cost and firepower and health that the player versions do (while still having half the range).
    • This makes them truly terrifying even in small numbers, but it's something that is actually slightly in your favor compared to what it was when they were separate since they had more shots and thus were less likely to overkill your ships (just your framerate).
    • This shouldn't really affect existing games too much, because the AIs barely used turrets before -- mostly -- at least compared to what they do now.
  • For any games older than 1.026, it will now remove 4/5ths of the turrets from AI planets, since AI turrets are now a lot stronger and more condensed being great-turrets.

Attack Ranges And The Small World Problem

  • Added a new galaxy option to the balance section: "All Ship Attack Ranges"
    • Lets you adjust the effective range of all of the ships in the galaxy -- yours and enemies. Does not affect 'infinite' sniper ranges, naturally. Also does impact tractor beams. The default is 0.8x in order to avoid the 'small world' problem where all the enemies are fighting you at once on a planet.
    • This does make it so that you need to be slightly more strategic with placements of your own turrets, of course. Your turrets naturally have 2x the range of their AI great-turret counterparts, though.
    • If you want to crank this as low as 0.1x (bad idea) you can, or you can crank it all the way up to 1.3x (also probably not the best thing).
    • Chris notes: This is a softer nerf than we originally thought of, just to play it safe for now. But if people are interested in trying smaller values, you now can.

Spire Frigates

  • Lance weapon of all variants (plus Laser Pulse weapon) 12,000 -> 10,100.
  • Triple Lance variant damage 4,000 -> 1,800, now hits 15 targets instead of 10.
  • Other variants Lance damage 6,000 -> 2,650, now hits 15 targets instead of 10.
  • Laser Pulse damage 4,500 -> 3,500.

AI Reinforcement Budget Caps And So Much More

  • The five "balance_initial_ai_defenses" xml entries in our central constants have been completely removed, and equivalents are now on the actual AI types instead.
    • This lets us completely balance the AI defenses around AI types if we ever want to tune that.
    • These are also per-reinforcement-point now, rather than per-planet, and they are also directly in AI cost units rather than "caps" which have to be converted to those units by a mysterious multiplier.
  • The entire way that we calculate what the defensive budget cap at a planet is has been rewritten from scratch. It was giving absurdly high numbers previously, such as 443 strength for the AI home planet's cap at the START of a game. That was just way too much for the start to be sitting at, cap-wise.
    • Now things are balanced completely around the number of reinforcement points, like the original game was, so if you go in and neuter a planet it will absolutely cripple their ability to reinforce. More power to you, the player! Yay!
    • The overall caps are also far lower than they ever were in the past, this is worth stressing before you read the next two bits.
    • But there remains a very tiny (laughably small, really) percentage bonus that the AI gets as the AIP rises. Eh.
    • There's also a NEW fairly small percentage bonus based on time spent in the campaign. The way it works out is a 48% increase in budget after TEN HOURS of playing at difficulty TEN. After 10 hours at difficulty 7, it's a much more modest 12% increase. This gives you some time pressure, but it's really nothing major and is still less than the AI could theoretically start out with prior to now. And with your new ability to absolutely cripple the AI by neutering guard posts at no AIP cost... we may need to make some of these multipliers higher, who knows. But probably now, we mainly have other plans.
  • Special Forces Ninja hideouts no longer count as reinforcement points. That was old AIWC-style thinking. Now it's just guard posts and the command station itself.
  • We've been experimentally tuning the numbers a bunch on these, too.
  • planet_dire_guardian_count has been changed to planet_dire_guardian_percentage.
    • This lets us be a little more flexible with the marks of planets getting percentages of dire guardians rather than explicit numbers,
    • The AI homeworlds are always full-dire, still.
  • The way that guardians are spawned on initial planet creation, and/or on reconquest, is now completely different and respects the budgets and budget caps.
    • Previously it was adding guardians kinda-sorta twice, and thus often making the really high-mark planets absurdly overpowered in general but in particular on AI types like Royal.
    • Not only was this logic doing the above, but it also was interpreting the budget caps entirely incorrectly, thus more or less ignoring them.
  • Added a new setting to the Debug section: Write AI Budget Info In Planet Tooltips
    • If something seems off with the planet strengths and reinforcements, then this is a way to see what is going on under the hood.
  • Improved the performance of how the game calculates tachyon, tractor, and gravity sources in the rollups. It's now substantially more precalculated than before, no longer having to iterate over any systems.
    • There's a small benefit to the strength counting code in this, too.
    • And actually, map generation in general seems notably faster with these changes in, despite the below change.
  • The GetAIToPurchaseCostPresentForBudgetType() method is now far slower, but much more accurate (not lumping strikecraft and guardians together, for instance) and always up to the instant accurrate (not having to wait for the strength counting code to run).
    • This should solve several issues with the budgets on reinforcements being ignored or improperly overly strict, as the case may be.
  • Found a place where it was seeding extra guardians per guard post, and took that out.
  • During reconquest seeding, the AI no longer gets a bunch of guardians to go with every guard post.
    • Guard posts in general are not expected to be tied to guardians, and guardians will eventually be vastly more powerful than they are now.
    • This was causing some planets, like AI homeworlds, to get VASTLY more guardians than they should, blowing way past their budget for guardians.
  • The way that AI reinforcement budgets are calculated is no longer based around the number of reinforcement points, but instead just a flat amount for the planet, which gets reduced (severely) by any lost reinforcement points.
    • Thus no matter how many reinforcement points there are, you don't have to suddenly worry about super high budgets on the AI planet, or strange spikes in budgets just because there are more guard posts.
    • Actually what this means, though, is two things: firstly, each reinforcement point is weaker on a planet with many of them (due to same number of units spread around more); but secondly, destroying a reinforcement point on a planet with fewer of them will do more (due to it reducing the cap of the planet by a larger percentage).
  • There is a new ignore_guardian_restrictions_in_waves that we use on the royal AI type, which lets it use its full capacity in sending waves of guardians at you.
  • When creating the budget for turrets and non-turret defenses, those will no longer increase based on the mark level of the planet or the fact that the planet is an AI homeworld.
    • This keeps them from having way too many of these things on high levels and next to nothing on lower ones.
  • When creating the budget for guardians, it's no longer doubled for the AI homeworlds.
    • This keeps THOSE from becoming so absurdly high.
  • There is a new multiplier_for_starting_guardian_budget_on_reconquest on AI types which makes it so that there CAN be some guardians added back onto planets, per-AI-type, during a reconquest of that planet.
    • For most AIs this will just be 0.015, which is an eighth of an eighth (the significance of that will be relevant in a bit).
    • This lets the AIs MAYBE get some of their guardians back right as they capture a planet, partialiarly moreso later in the campaign, but often it will be too paltry for them to get anything.
    • For the Royal AI, it is set to 0.001 since it has such a higher budget of guardians in general. This may mean they practically get nothing, ever, in those circumstances; but better than too much and we can always tune later.
  • There are then other new multipliers on AI types: multiplier_for_game_starting_guardians_from_total_budget, multiplier_for_game_starting_turrets_from_total_budget, multiplier_for_game_starting_non_turret_defenses_from_total_budget, and multiplier_for_game_starting_strikecraft_from_total_budget.
    • These all affect how much of the AI's budget for each planet they start with when the game is at the very zeroth second still in mapgen.
    • In the past, we always gave -- for some reason -- fully HALF of the budget cap to each planet, which did not give much room for them to expand and get really huge if you ignore them for too long. ;) Most any planet could do would be to double in power, aside from things like the Hunter or other third parties coming in and making it temporarily stronger.
    • Now we've got this set to 0.125, which is an eighth of the full budget or a quarter of what the old starting strength was. This lets us multiply the final budget by 4x to still have the same starting strength (and we have done that), but now the starting strength can get 8x stronger if you leave them alone for THAT long. And that's before you get into the (very small, as noted in prior notes) increases based on time elapsed and AIP.
    • So the AI has more growth potential, BUT this is measured against the fact that your neutering of them is incredibly powerful.
      • If they had 4 guard posts and a command station, and you kill the guard posts for 0 AIP cost, you'll reduce their max cap to 1/5th of whatever its max would have been.
      • If they had 10 guard posts and a command station, and you kill the guard posts for 0 AIP cost, you'll reduce their max cap to 1/11th of whatever its max would have been. Wowzers.
      • In both cases this keeps you from neutering things to be TOO absolutely weak, but also gives an incentive for popping guard posts when you can that are not related to your main objectives but might threaten you in the future. Aka a lot like the first game.
  • Added the fields always_is_ai_turret_spawn_point and always_is_ai_non_turret_defense_spawn_point, both bools, to game entities (ships only).
    • This lets us override the logic that says "only primary keys of modulus 6 get turrets around them" and things of that nature.
    • This is now used for the AI overlord, to always let it be surrounded by both types of things (phase 1 only).
    • Note that these fields only work for ships that are already a spawn point (mainly guard posts and command stations).
  • The Turtle AI type now has 4x the turret budget of other AI types.
    • Fortress Baron has 3x the turret budget.
    • Peacemaker has 2x the turret budget.
    • These heavier-turret variants are certainly interesting to run into and give the game a very different feel. These are the more defensive-oriented AIs, after all.
  • Added a new is_reinforcement_location bool field. This lets us make things other than command stations and guard posts into reinforcement points.
    • Now that reinforcement points don't actually cause the AI to get more budget, but rather just have it spread out more, it's far more interesting to have this.
    • Plus it makes it so you can do even harder neutering to certain planets, since the total number of reinforcement points will be higher by definition, which is good for you as a player.
    • But it also makes it so that isolated units with this flag on it will now be more protected than before, possibly with turrets around them, which is in itself quite interesting.
    • BUT that still means that forces are being spread more thin around the planet, potentially, so there are pros and cons to you strategically in general with this.
    • AI structures that thus now act as reinforcement points:
      • Warden Fleet Base (again -- back to the same as the prior version that was last public anyway).
      • Data Center
      • Fortified Data Center
      • Major Data Center
      • Plasma Eye
      • Ion Eye
      • Troop Accelerator
      • AI Fortress
      • Raid Engine
      • Alarm Post
      • Magnifier
      • Instigator Base
      • Astro Train Depot
      • Risk Analyzer
      • All of those new ones above also now have always_is_ai_turret_spawn_point="true" on them. Yet more texture in the planets, as well as giving them more inherent defenses.
  • Guardians, as well as anything marked as a reinforcement location, can no longer be transported or put inside a guard post.
    • Older savegames may have some of them in there.
    • This causes FAR more guardians to be out and patrolling around now than were before. Some of them were just sitting inside guard posts in a very boring fashion before, rather than doing their patrol thing from what we can tell.
  • The tooltips for planets on the galaxy map now show you how many reinforcement points the AI has there out of how many total there have ever been there if it's an AI planet.
  • Added modulus_for_turrets_at_reinforcement_location_lower_is_more_frequent (default 6) and modulus_for_non_turret_defenses_at_reinforcement_location_lower_is_more_frequent (default 3) as fields to the AI type. Previously these values were hardcoded in, but now they can be changed for various AI types. This will make turrets appear around more structures for those AI types, or non-turret-defenses do the same.
    • Turtle now uses 1 and 2 instead of 6 and 3.
    • Fortress Baron now uses 2 and 3 instead of 6 and 3.
    • Peace Maker now uses 2 and 3 instead of 3 and 3.
    • This makes it so that their turrets are more spread out on their planets rather than being clumped in super-crazy places on just some locations.
  • Added a variety of new levers that determine now how budgets are distributed intra faction or inter faction, nor what the caps or starting amounts are... but instead affect how bonuses (or penalties) work to various AI types. These can be regular floating point numbers with three decimal points of precision (0.001)
    • multiplier_for_budget_reinforcement, multiplier_for_budget_wave, multiplier_for_budget_cpa, multiplier_for_budget_warden_fleet, multiplier_for_budget_reconquest, multiplier_for_budget_hunter_fleet.
    • These don't have any effect on other parts of these AIs -- there are other mechanisms for choosing how much budget goes where, but this lets us set general scales on a per-AI basis on certain types of budget. This is pretty importance for us to be able to balance things reasonably so that there are things like appropriate growth rates of AI budget without you feeling like it starts out with too much, or has too low of a final cap, etc.
    • Most of these are just set to 1 for now, but multiplier_for_budget_reinforcement is 2 for almost all AIs. multiplier_for_budget_wave is 2.2 for the Vicious Raider, and multiplier_for_budget_reinforcement is 4 on the Turtle.
    • We'll undoubtably balance things more using this in the future, but for now this is a nice new lever for us to be able to use to give AIs their own unique characters as well as simply making them not have anemic defenses 30 minutes in even if their defenses 1 second in are smaller now.
  • The way that the AI tries to spend its funds on reinforcements DURING gameplay is absolutely reworked from scratch.
    • It was doing a strange thing where it would kind of roll a die and depending on the results try to reinforce certain things with more or less priority, and this often led to some funky results. Potentially it resulted in under-spending at times, and sometimes it may have meant a larger portion of the budget went to non-turret defenses than was at all warranted. And sometimes there were no turrets being put in at all. The reasoning was sound, really, but it was old and didn't fit with the modern economy of the AIs in the game.
    • Now it instead goes through a complicated set of calculations of percentages of what goes where, within random tolerances, dividing out budgets in a certain priority between strikecraft, turrets, non-turret defenses (NTDs), and guardians. If something will get lower amounts of funding, then other things naturally will pick up the slack in a pretty sensible order of priority. It then tries to spend its money more wisely once it has sub-budgets allocated, and rolls any excesses into guardians first (to give more of a chance of buying one), and then into strikecraft (to try to fully deplete the budget if at all possible.
  • Changed multiplier_for_starting_guardian_budget_on_reconquest to be 0.125 instead of 0.015.
    • This will substantially hurt. The AI is kind of taking a beating on some of the econ changes lately, so this seems warranted.
  • Changed all the various multiplier_for_game_starting_whatever_from_total_budget to be twice what they were before, or therefore a quarter of the max starting budget in most cases rather than an eighth.
    • This should eliminate the sense that you're able to just bumrush planets in the early game before they can reinforce. We could have upped the actual caps of the AIs, but then the late game would have swung too much in their favor if the game runs really long.
    • We'll see what happens; if the starts are still too easy, then we'll be upping the starting caps instead of the percentage again. More than a quarter of the cap at the start seems like planets will just be far too cramped.
  • The "defensive budget multiplier" that comes from the AI difficulty is now split into ongoing and initial. This lets us make the lower-difficulty starting amounts not SO anemic that they are pathetic, and then as we get into difficulties higher than 7 not have them be so ludicrously overpowered right at the start. Difficulty 7 is also toned down to 90% of its ongoing defensive budget when it comes to initial seeding, now.

Completely Revised Way Of Seeding Guard Posts and Command Stations

  • There's a new AIGuardPostAndCommandPlacer set of xml data that lets us define a lot of different ways that guard posts and command stations can be arranged on planets beyond what has been the basic prior to now (one one each metal generation point).
    • These can be varied by mark level ranges of planets, and have different sets for the homeworlds of AIs. They can easily be modded in or added to like map types or whatnot.
  • The way that human homeworlds have their home command station seeded is now a bit more randomized than before, but also does a better job of keeping itself away from wormholes.
    • This should make your starts slightly less predictable.
  • The AI Homeworld was set to always put its command station dead center. That's no longer the case. This will make some games easier, others harder, and is fine with Chris.
  • Over the years, the concept of what a "guard post" is has changed quite a lot. Originally in the first AI War, they were literally just a gathering point for other ships. Then in 4.0+ of that game we started making them scary weapons in their own right.
    • This sequel has always had the armed variant, but that causes some problems in balance with them sometimes overshadowing the actual ships that are at their location, and in many cases making it impossible for you to fight just a single guard post or two at a time with the way that their weapon ranges were so large that they could pick you off.
    • Thus we've hybridized the old and new concepts, making a new "Unarmed Guard Post" type (and Ghost and Swarmer variants), and then adding in a chance per post for it to be armed versus not armed now.
    • We've split it out so that normal and home AI planets each have their own chance, per AI type: chance_out_of_100_for_armed_guard_post_normal_planet and chance_out_of_100_for_armed_guard_post_home_planet.
    • Bear in mind that, prior to this build, it was always 100 for each one, essentially.
    • Old numbers: Now for most AIs it is 15 and 60, turtle is 60 and 90, Fortress Baron 50 and 80, Peace Maker 30 and 60.
    • Please note that these are not the pure percentages of guard posts on a planet that will be armed: rather, at each guard post it essentially rolls a 100-sided die and if the number is lower than what is stated there, it will be armed.
      • So you'll see all sorts of funky distributions, ranging from completely unarmed guard post planets to completely armed ones.
    • This DOES make things easier for the player, but mainly it gives other units a chance to shine and makes the armed guard posts more interesting in that they are more rare.
      • Ideally we'll be able to make guardians scarier in their stead, but one thing at a time right now. This release has a lot of changes already!
  • The way that unarmed guard posts were seeding has been changed yet again. There's now a 100% chance of armed guard posts on mark 7 worlds and homeworlds, and then a per-AI setting for what the percentage chance is for every mark level below that. The percentages are generally much higher than they were before, making for far fewer unarmed guard posts as the game progresses and in general fewer overall.
  • The "initial reinforcements" that get seeded into/around reinforcement points now happens as the absolutely last thing, which means that things that should have turrets or whatever around them from the very start, or which are not guard posts but act kind of like them and are added normally during the "special entity seeding" period of mapgen will now get their initial followers properly.
  • Usurpers now have to be on the target planet for 10 seconds before they can reclaim it, otherwise you never even see the usurper if you don't have a defending force.
  • When a reconquest of a planet happens by the AI, it now uses only armed guard posts regardless of the ratio of armed and unarmed it would have previously used.
  • Previously there was mostly just the AI seeding guard posts one metal deposit spots, which is boooring. Now that we have the ability to do more, we have a variety of new types:
    • Guarding Wormholes Generally: this can start at mark 1 and go all the way to mark 7 (not homeworlds, though).
      • These are highly aggressive stances that make it hard to get into the planets, and on very high marks they wind up putting 2-3 guard posts PER wormhole. Talk about ow. If there are too few spots for their posts at just wormholes, the rest are spread out very ambiently.
    • Guarding Wormholes Harshly: this can start at mark 3 and go all the way to mark 7 (not homeworlds, though). They should be VERY infrequent based on their weight. These are deathtrap worlds.
      • These start out with 2 guard posts per wormhole, and only go up from there -- by mark 7, it's freaking 5 guard posts per wormhole. Just entering the planet is leaping into death, more or less. But that's part of the fun of having a rare extra difficult guard post design, and bear in mind that it doesn't make for more reinforcements and so the total firepower of the planet doesn't get a boost per se (well, guard post firepower itself aside -- which can be huge on higher marks, admittedly, but... well, have fun with those!).
    • One Single Mass: this is another super rare one, and can start at mark 2 and up, but not homeworlds.
      • This puts all of the guard posts pretty much in one big clump somewhere on the planet. Depending on randomness, the clump can be kind of strung out, or more circular, or very tight. In many ways this makes them into one super-guard-post, which is potentially like trying to crack an old-school fortress in AIWC. These have randomized ranges of guard posts per mark level of planet, with 2 on mark 2 to, from 3-4 on mark 3 up to 9-12 on mark 7. This is another intense one.
    • Spread Out Guard Posts
    • This is a variety of ones that are much more lightweight and which get seeded much more commonly: basically just randomly putting guard posts around the whole planet, but pretty far away from one another to make them easier to approach. The numbers vary a lot, and most planets will use one of these.
    • Guarding Metal Deposits: moderately low chance of coming in on any mark 3 and up, but not homeworlds.
      • Places one guard post on every metal deposit, and then no more. This is similar to old behavior, except that it doesn't try to add more if there are fewer metal spots than an "ideal" for the mark level. Consequently the difficulty of these will vary a lot, but stealth guard posts remain easy to find because they're just on the metal spots (that was a big giveaway in the past, cough).
    • Spread Out Guard Posts And Two By Command
      • These ones start kicking in on mark 4 worlds and up, and one of the two homeworld variants has this, too.
      • This basically puts two VERY close to the command station, guarding it in a ferocious fashion, but then the rest spread very far around.
    • We can always code more of these as we think of them, and these are easy to add in as part of mods or expansions or whatever as well.
      • Extra logic could be added to seed specific extra structures at certain guard posts if we really wanted to, although that's not my favorite and it's not by default possible in the moment. But the possibilities for future additions are pretty endless if we have sufficiently interesting ideas. A lot of the other things are better off being a bit random is the current feeling, though, as it plays into these nicely.
  • Added the ability to have certain guard post layouts only be for certain AI difficulty ranges (based on the max AI difficulty, not the difficulty of the AI at that planet).
    • Additionally, based on the max AI difficulty, the ability to have a certain added weight to guard post layouts per difficulty above some cap. For now we're using that in the following ways:
      • (Standard weight for something appearing neutrally frequently is 100, please note.)
      • GuardingMetalDeposits (bit hard) base weight was 20, but it now gets +5 for every difficulty 6 and above. (6=25,7=30,10=50 etc).
      • GuardingWormholesGenerally (quite hard) base weight was 10, but it now gets +5 for every difficulty 5 and above. (5=15,7=25,10=45 etc).
      • GuardingWormholesHarshly (super hard) base weight was 2, but it now gets +3 for every difficulty 6 and above. (6=5,7=8,10=17 etc).
      • OneSingleMasss (super hard) base weight was 2, but it now gets +3 for every difficulty 6 and above. (6=5,7=8,10=17 etc).
      • ThreeSpreadOutGuardPosts (super easy) base weight was 100, but it now gets -10 for every difficulty 5 and above. (5=90,7=70,10=40 etc).
      • FourSpreadOutGuardPosts (super easy) base weight was 100, but it now gets -10 for every difficulty 5 and above. (5=90,7=70,10=40 etc).
      • SixSpreadOutGuardPosts (easy) base weight was 100, but it now gets -10 for every difficulty 6 and above. (6=90,7=80,10=50 etc).
      • SixSpreadOutGuardPosts_AndTwoByCommand (moderate) base weight was 50, but it now gets +5 for every difficulty 6 and above. (6=55,7=60,10=80 etc).
      • EightSpreadOutGuardPosts (moderate) base weight was 100, but it now gets -10 for every difficulty 7 and above. (7=90,10=60 etc).
      • EightSpreadOutGuardPosts_AndTwoByCommand (hard) base weight was 50, but it now gets +5 for every difficulty 7 and above. (7=55,10=75 etc).
      • TenSpreadOutGuardPosts now only appears on difficulty 5 and up.
      • TwelveSpreadOutGuardPosts now only appears on difficulty 6 and up.
      • AIHome_TwelveSpreadOutGuardPosts now only appears up to difficulty 8.
      • AIHome_TenSpreadOutGuardPostsAndTwoByCommand now only appears on difficulty 6 and up.
    • All of this combines to give a much more richly-textured scary landscape for the higher difficulties, while keeping things notably simpler at easier difficulties even at the level of a cursory glance.
      • Looking at a difficulty 5, 7, and 10 game, you can see that this is NOT a subtle effect. As we theoretically add more types of seeding patterns, or other people think up seeding patterns, we can make an even richer tapestry.
  • If you have map gen logging on, you can now see entries for "Guard post style for planet" that shows what mark the planet is and what kind of thing it is seeding.

Bugfixes

  • Adjusted the "BUG: Ship is on burningAndDying list but has not exploded" message to instead just mark the ship as exploded, as there are too many ways to get to that state.
    • Thanks to ArnaudB for reporting.
  • Fixed a nullref exception that could happen in DetachVisObjectFromSim on ship squads.
    • Thanks to ArnaudB for reporting.
  • Fixed a nullref exception that could happen in StartReconquestDefenseSeeding for AIs types that do not use non-dire guardians for some reason.
    • Thanks to ArnaudB for reporting.
  • Fixed a nullref exception that could happen in BattlefieldVisualSingleton.HandleSquadAndShipUpdates, and made it so that it self-repairs when it can or self-reports when it can't. Either way it continues to draw things and update them properly rather than having the display get all strange if such an exception occurs.
    • Thanks to Daniexpert for the report.
  • Improved the "find a safe space to be" code for ships so that they are now more likely to find a CLOSE space to be.
  • For guard posts in particular, it does an extra good job of trying to keep it close to the intended point (whatever that is in the given schema).
  • Finally fixed a really annoying bug wherein command stations that got shot to remains and then rebuilt did not count properly (thus not letting anything else be built or rebuilt at their planet) until you saved and loaded the game.
    • Thanks to Badger, ArnaudB, Asteroid, KDR_11k, Puffin, Sigma7, Ryulong, Core, and GrimerX for reporting and helping to find it.
    • Special thanks to Ryulong for the recent savegame that let us reliably reproduce this.
  • Fixed a bug where some achievements relating to killing specific AIs was not triggering because it was trying to match the internal name to the display name. Now it checks both display name and internal name to be doubly sure.
    • This makes Full Ensemble in particular now properly trigger when you beat them.
    • Thanks to ArnaudB and DEMOCRACY_DEMOCRACY for reporting.
  • If you have existing saves where you already beat the game, loading those savegames will now trigger the achievements for you that you properly earned (just load them, unpause, and wait at least 2 seconds).
    • It's much better this way, versus having to dig up a savegame from right before you won, etc, if you're going back to get achievements that you earned but didn't log because you were offline, or there was a bug, or whatever factor.
  • Fix a typo in the hover planet text
    • Thanks to Sigma7 for reporting
  • Improve the readability of some Instigator base hovertexts
    • Thanks to Daniexpert for reporting
  • Since it is now possible for non-combatant ships to contain lots of combatant ships (unarmed guard posts on the AI side, but frankly also your own transports), the game now takes great pains to make sure that it isn't overlooking the combatant-nature of ships contained within non-combatant ships when checking things like strength.
    • Compared to prior versions of the game (where all guard posts were combatants) this doesn't fix or change anything, but it might make player combat strength more accurate in the case of transports if there was some issue there (though there's no sure evidence that was a problem).
  • When ships are dying, you no longer get a bunch of red "null fleet" messages hovering over them, unless you have debug info turned on for tooltips.
  • The game should now say "release" to show a more or less detailed tooltip on the various screens if you're already holding down the button to make it do whatever it's doing (you can invert the behavior, so which is held and released should match that).
    • Thanks to Admiral for reporting.
  • Clarify the tooltip for Risk Analyzer Notifications
    • Thanks to Thotimx for the bug report
  • Completely reworked how the long range planning threads keep track of how long it has been since they have run, so that they should probably run at appropriate intervals when the game is running faster than 1x speed now. Previously these were running slower the faster the game ran, which made the AI stupider at 10x speed than 1x speed.

Fixes For XML Modding

Example tiny mods and discussion are here: https://bugtracker.arcengames.com/view.php?id=22433

  • Several different things were not staying properly on ships when using an is_partial_record="true" to mod them:
    • exp_to_grant_on_death
    • watch_planets_at_X_hops
    • tags (ow this would cause many problems)
    • capturable_seed_weight (again very problematic)
    • capturable_max_per_galaxy
    • capturable_can_seed_at_all
    • Thanks to NRSirLimbo and ArnaudB for reporting.
  • Fixed a couple of things that would complain they could not be found properly on ships when using an is_partial_record="true" to mod them (but in reality these were carrying over fine from the base version being modded):
    • ship_or_structure_explosion_sfx
    • ship_or_structure_explosion_if_on_other_planet_sfx
    • Thanks to NRSirLimbo and ArnaudB for reporting.
  • If outgoing_damage_modifier or incoming_damage_modifier entries are duplicated in the list of rows for a system or entity (respectively), whether that be through mods or just in the original definition, it now only uses the last-to-apply version. This last-to-apply one is always going to be a mod version if there is a mod version.
    • This makes it so that you can't have two different modifiers on the same weapon that both are based off of the hull health of a ship, for example, but I don't think we're doing that anywhere and it would be messy in the UI if we were, anyway.
    • This then fixes the issue with various weapon bonuses and the like duplicating themselves when you are using an is_partial_record entry to mod a weapon.
    • It's worth noting at this point that xml modding is inherently slightly fraught in terms of minor bugs or oversights like this which can have major repercussions. Please just let us know on a case by case basis what you find, ideally with the xml you're using and a savegame that shows the wrong behavior, and we can likely fix it up pretty fast.
    • Thanks to ArnaudB for reporting.
  • Both ai_ship_group_membership and fleet_membership now properly adjust the existing membership when there is a duplicate entry entered for them (from a mod or whatever else) linking a given ship type to a given template or group. It now adjusts the weight and so on in the template or group, rather than adding a duplicate entry.
    • Note that if you change the category for a template, it will assume you mean it to be a duplicate and will add that accordingly. There really should never be any valid reason to mod the category of a ship, but if you must then you can just 0 out the min and max of the old one maybe?
    • Thanks to Flypaste for reporting something next to this, although not exactly this.
  • Creating a mod to replace or redefine parts of the fleet design templates (like the starting fleets) now works properly rather than creating duplicate entries or extra counts or both.
    • For an example of a mod that works with this new function, the raw xml bits are here: https://bugtracker.arcengames.com/view.php?id=22136
    • Note that if you choose to change the category of ships from HERE, that will work fine. You just can't change categories of fleet template items from within the game entity side.
    • Also please note that this change now prevents you from having duplicate lines of the exact same type of ship in a starting fleet. So something with two lines of fusion bombers would not work anymore, for instance, even with no mods in play, and would just take the count of the last fusion bomber entry and use that.
      • This doesn't stop you from using multiple duplicate ship lines in-game of course, but it does stop it in the starting fleet designs, none of which (that are still active) were using that function anyhow.
    • Thanks to Flypaste for reporting.
Retrieved from "https://wiki.arcengames.com/index.php?title=AI_War_2:The_Grand_New_AI&oldid=37573"