Github: https://github.com/Kondikka/expertai



Advanced configuration has been overhauled to give you the AI sandbox features that are possibly only ever used by me... Every non-country specific setting can be configured from a single event and there is a summary of all settings. Some settings have been removed, some added.

Country specific configuration is accessed through the new decisions which are only visible while you have the main configuration event open (event EAI.0). Every country that exists can be configured now, both difficulty and army. Summary of the current AI focuses is displayed in the army configuration event and all the focuses can be configured through a single event. Any focus can be changed without having to reconfigure all the focuses at the same time.

There are new options to configure the AI army with: light armor attachment, 20w army, 10w infantry focus, 20/40w armor sub-focus, land doctrine path sub-focus, fleet composition focuses, heavy fighter focus. Want to find out what happens when Germany invades the world with Mass Mobilization, 10w infantry and 40w light tanks filled with Leopards? Go ahead (not too bad for the AI actually until it invades USSR...).




Desperate Defense mode

Desperate Defense has two modes: normal and high. The bonuses are higher than before even on normal. It starts at >0% surrender progress and maxes out at >50%. With high mode, the bonuses are much higher from the beginning. Recruitment time reduction is always -85% (takes a month to train a division). You can see the specific bonuses in the config event. The point of this mode is to have a more realistic game where the AI does not build a zillion things even before the war has begun but it still receives the help it requires to put up a fight when it is losing.

Reinforcements mode

Tried every other difficulty setting and it's still not enough? Enable this and the AI may receive free divisions AND equipment. Would not recommend this for your first 1.5 game, but you can of course toggle this on later in the game. You can see the number of reinforcements spawned per country and in total in the configuration event for funsies/testing.

How much equipment: 500-5000 units of the latest unlocked equipment/variants every 20 days at most in most of the land equipment categories including tanks until the country has a positive stockpile.

How many divisions: 4-6 every 20 days at most. It will receive units as long as it is below the current division cap for this country (which depends on various factors but a tiny country should not receive 100 divisions for example).

What kind of divisions: mostly infantry with a small chance of motorized or armored units (if the AI has access to them). Templates roughly match what the AI currently has. Experience levels are a randomized and weighed toward regular.

Where: to simulate reinforcements arriving from somewhere else and not just popping up on your screen, the most preferred location is close to the active front line but just out of sight of the invader while the least preferred location (usually for small countries) is any state that is owned and fully controlled by this country.




I've overhauled how the AI chooses invasion targets. To be more specific... I've forced the AI use an event/ai_strategy driven system (as with other things in the mod) instead of using the vanilla's hardcoded invasion AI. This is to address the lack of focus on invading specific targets (i.e. diluting invasions between invading several countries the point where they are ineffective), the lack of coordination between AI countries (if the Allies invade Europe, all the faction members should join in and try to invade the same target to increase the chance of success) as well as generally invading when it is a bad idea due to various circumstances.

It's not perfect and it is fairly experimental at this point but at least in historical games it does seem to make the Allied landings much more successful and it is far more dynamic than the scripting I had done previously.

Here's what it does now:

a) The AI looks at various circumstances before deciding to invade anything: it compares fleet sizes, army sizes, makes sure it is has free manpower and isn't being invaded at home etc. Among the eligible target countries it will prefer nearer targets.

b) AI will pick just one country to invade at any time, which should give it a much better chance at successful invasions. It will check if any progress has been made on the invasion after a certain period, and if not, it will try another target if available.

c) AI countries with allies in a war will follow the AI faction leader or the overlord, i.e. the leader chooses an invasion target and everyone else attempts to support it. This should result in concentrated and massive invasions when a faction sees an opportunity to invade (like Allies invading Germany). Likewise on the Pacific theater, Japan picks one target to take out at any time which should make it more efficient in securing the nearby islands until its fleet gets wrecked.




This is the ability to influence some of the things that your allies do. Since I've already created several script-based systems that control the AI through flags, allowing you to control these systems through the shiny new decisions system is fairly simple.

Coordinate Production

You can check what your ally's current focuses are for free but changing it requires command power (10-100 CP, scales to the relative strength of countries, 50% discount for subjects/free at the start of the game and 50% more for overlord - if you're a very weak minor, you won't be able to make changes to your overlord due to >100 CP cost). This works exactly the same way as reconfiguring a country's army focuses in the config event does but this a part of the gameplay (kinda wanted to do this a year ago when I added the army focus system but this obviously wasn't possible without decisions).

You can do things like:

-As Germany, tell Italy to focus on only building X ship class.

-As UK, decide what type of divisions your subjects create. Maybe you want to take advantage of Raj's manpower but you don't like the 40-width divisions it makes so you tell it to switch to 20-width infantry instead. You can also set up their doctrines (but not after they have started researching doctrines of course).

-Check what the ally may need when lend-leasing (if it doesn't focus on building medium tanks then there is no point sending it any) You can also get estimates on their current stockpiles.

Coordinate Naval Invasions

Designating an invasion target costs 100 command power. This makes all your allies invade the target country wherever they can for 6 months. Playing as UK and you wish you could tell USA and the rest to prep for a massive invasion in Europe? You can do that. You can't specify a target state or anything but it's better than having your allies sitting around doing nothing. This does not force the AI to invade brainlessly, it still considers various factors before suiciding just because you told it to.

Coordinate Lend Lease

Playing on the Allied side and things are going badly for USSR on the east front? Designate USSR as a lend-lease target for your entire faction and your allies will prioritise lend-leasing USSR over anyone else. Again, whether it actually does lend-lease depends on various factors (the target must have a deficit for one thing).




Thanks to 1.5 I have far more control over the specific equipment the AI produces which means I can make the AI do all sorts of fancy stuff.

Navy: before in this mod the AI was only able to build DD, CL, BB, CV in order to make things simpler for the AI. Now it can build all the ship classes (except super heavy BB) with clearly defined logic for when it should build them (i.e. don't build the early ship types, prioritise capital ships in these situations etc.).

This is built into the AI army focus system of the mod so that the AI commits itself to specific fleet setups, prioritises their research, variant creation and so on instead of trying to research and build all ship classes. You can of course configure this for yourself through the config event or the new coordination decisions (you really want GER to build submarines? That's possible... however it's still not a great idea).

By default, countries will go for fleet setups that usually match their starting level 2 ship techs: all build DD and CL, FRA builds BC and CA, GER builds BB and CA, ENG builds CA and CV.

I've added several strategies for certain countries to build X many carriers or battleships during the early game to beef up their navies, a lot of specific research/production prioritisation depending on the number of docks, as well as the ability for the AI to switch to more expensive fleet setups if they have tons of docks (late game GER or SOV for example).

Air: the change is less drastic here but similarly the AI can now build heavy fighters and transports. USA by default will research/build heavy fighters. Majors with more than 25 MIC will build transports (no more than a couple of factories allocated) but there is a slight problem because the AI won't deploy them until it has enough for a 100 plane wing.




Scripted a new system for the AI where it puts a part of its PP into a reserve that is only used for changing laws. This is a workaround fix for various PP prioritization issues for the AI but it does not require AI "cheating" by allowing it to go to negative PP to switch laws. This is a rather major improvement for the AI even over the old system.

Consider your average democratic/neutral country that has no manpower when at peace (France, Finland etc.) and can't raise manpower laws due to limits. When they get into a war, they don't have any political power to raise conscription laws because they spend it elsewhere which means they are royally screwed as it can take half a year to get enough PP.

With the new system, they can instantly raise their manpower law and start mobilizing as soon as the war begins because they have 150 political power reserved just for such occasions. For example, France which at the start of the war has about 700k troops could not raise its manpower law before because of abysmal PP gain so it would be stuck at that amount for a very long time. Now however, if France is allowed to sit around until early-mid 1940, it can double its army in size.

This reserved PP is also used for the other laws but the AI will prioritise saving it for manpower laws if it feels threatened.




-Ripped out a lot of old scripting and rebuilt it in a more convenient/effective way if needed anymore.

-A lot of improvements to the AI army focuses system in general. AI's choices are actually more random now (multiple choices fired on the same hour = bad because the random seed depends on time). More logic to make the choices more sensible (especially if you toggle on random focuses).

-AI presets i.e. what each country builds by default are defined in scripted_triggers/EAI_ai_presets.txt. Instead of reconfiguring a lot of countries in-game every time you start a new game, you could simply overwrite this file with your custom preset.

-Some diplo tweaking regarding lend-lease, volunteers (X country won't do it if the target country's enemy has a NAP with X/faction member, other stuff)

-Reworked division caps for countries. 1.5 AI is generally better at queuing stuff so I don't need to control it so heavily.

-A ton of scripting to make the AI micro its factory assignments better in various situations.

-Scripting that should help the AI defend its home better when losing as well as avoid some bad situations where it sends troops to places where it shouldn't. Don't expect miracles.

-Construction: A lot of rebalancing and tweaking the scripted AI construction system. AI can build infrastructure for resource gain (in addition to building it for construction speed or supply). It may do so in vanilla as well but in the mod this is specifically scripted into the AI so that minors with lots of resources for example max out their resource rich states quite early for trading and that others try to increase resources that their army focuses need. AI builds anti-air where it has lots of resources if the enemy has strategic bombers.

-Added host check event for multiplayer which I haven't tested.

Gameplay:

I would prefer not to change gameplay in this mod but let's just say that I have strong enough feelings about these issues that I don't want to wait for them to be addressed in patches. :)

-Fort cost nerf and division XP effect nerf from the tech addon mod are now in the main mod. XP combat bonus from 25% to 10%. Fort cost from +500 to +1500 cost per additional fort level (total cost for a lvl 4 fort is 11000 instead of 5000).

-Equipment/manpower reinforcement speed nerf is also in the main mod now. What it means in practice is that sprinting all the way to Moscow should be harder to do now that equipment doesn't arrive instantly to your front lines (and the further you push, the harder it becomes). This may need some balancing but of course this is meant to make the game harder and longer, not convenient.

AI template design process:

Much more efficient, partly from me tweaking the scripts, a lot due to 1.5 fixes. I was able to remove most of the AI's event-created-templates-in-exchange-for-XP that give the AI a starting point from which it develops its own design but certain designs still cause trouble for the AI that I just couldn't fix without events.

AI's army experience generation:

-Made sure minors and most majors get the land doctrine advisor first for the XP.

-On historical mode, all countries take national focuses that give army experience much earlier to make up for the AI's lack of exercising so that the AI can upgrade its infantry templates sooner. Countries with generic trees do this on ahistorical too.

Makes quite a difference in the templates especially for minors.

AI builds:

-AI builds cavalry occupation divisions now that in 1.5 it knows how to use them. It builds two 5xCAV divisions per every occupied state. Regular 5xINF garrison divisions are built two per every controlled state.

-AI is more willing to upgrade its motorized divisions to mech now that it's cheaper. (tank divisions still always use mot, tl:dr it would cause problem for the AI).

JAPAN & CHINA

-♥♥♥ focuses on building only infantry divisions early on. It uses Grand Battleplan+Infiltration which uses templates with logistics that should help in asia.

-China (and the warlords) builds 10w infantry and uses Mass Mobilization (5x inf first, 6x inf after width reduction). Previously I had to limit AI China when it was against AI ♥♥♥ because Japan did so poorly but this is not needed anymore. Most of the time Japan appears to win.

UK

-ENG builds 20w infantry and does not have early armor focus to help it get more divisions out so that it can *hopefully* defend its home a little better.

Germany

-Builds its armor divisions more towards soft attack.

-Builds 40w armor with 26w infantry (normal mode).

USSR

-Builds 40w armor with 20w infantry (normal mode).

-Uses Deep Battle.

For the time being, I have nerfed certain minors from building too many units during the first half of the game or so if they are against AI Japan or Germany: LUX, BEL, HOL, INS, MAL, PHI. Reason being that they would otherwise build fairly powerful armies in this mod that the AI GER/♥♥♥ can't deal with it reliably. I want to make sure that the game progresses fairly historically on historical mode instead of GER getting stuck at Maginot or ♥♥♥ not invading the islands even if it only happens 15% of games.

AI may do some army focus changes for itself such as:

-stop building light tanks in super late game and switches to modern tanks

-build more expensive navy when it has lots of docks

-a minor may switch to 40w infantry when it has lots of infantry equipment, manpower and army experience on challenging mode

-start submarine research/production when it starts researching submarine doctrines (super late)

Research:

Tweaked the AI's research to be more sensible. Much more prio on inf and art lvl 2, less on support weapons. More on air doctrines. Very low prio on expensive stuff like planes, ships and armor if the country has low industry. Less wasting research time on armor variants ahead of time. Other stuff.

Templates:

-A ton of templates removed and added.

-A country with Mass Assault that wants to build armor makes inf+art+harm/marm divisions now instead of using mot+spart.

-Grand Battleplan templates all use logistics company now except for the 20-width types.

-Added logistics/signal to all 40w templates that didn't already have them.

-Removed recon from a lot of templates. Maintenance company added to motorized templates.

-Reduced template widths because the offensive field marshal trait was changed.

-Medium armor designs use more SP-ART except for the pure MARM/MOT/MSPAA design.

-Other tweaks.

1.5 National Focuses:

Tweaks to make sure the AI progresses through the new focus trees efficiently and reasonably historically with historical focuses. Germany now always does the pre-war annexations on historical dates to give it time to get other important focuses early.

Decisions:

-Some tweaks to make sure the AI gets the no-brainer war/industry decisions and doesn't waste PP on the less important ones.

-Tweaks to avoid ahistorical stuff on historical mode.

-Tweaks to make the AI uses various generic decisions to keep its stability and war support up: anti-ideology raids, improve worker conditions, promises of peace, insitute press censorship etc. Good stability is worth around one industry tech so it's worth investing in. Also has the side effect of giving it some protection against random ideology drifting or party boosting as it keeps raiding the unwanted ideologies and sometimes uses press censorship when it has lots of PP.

Variants:

Fixed the AI's variant names so that they use the actual names of the equipment instead of a generic one like "Destroyer Mk 2" (armor variants for majors still have custom names I made like T-34 Model 1942 etc.). The generic AI variant names showed from time to time on the interface which was annoying. The game won't add a proper designation to a new variant every time but it will still be building the best stuff it has.

Added AI variants for all the new stuff the AI can build: h. fighters, all the ship classes.

General diplomacy:

-Spanish Civil War start date is less random, should start in 1936 every time.

-Made sure that GER and SOV will keep the NAP until its time to cancel it instead of doing so as soon as possible. No more cancellations as soon as GER attacks Poland.

-Made sure that ♥♥♥ and SOV, and all the involved minors will create/keep NAP after ♥♥♥ has finished the focus for it as long as it makes sense.

-Script-monkeyed a way for GER to handle the invasion of western Europe focuses better: it will switch to continuous air production focus between the three invasion focuses so that it can immediately start the next invasion when it finishes the previous invasion (getting stuck for 70 days doing a random focus is bad when it should be invading...).

-Added a ton of logic/reworked old stuff for various historical AI focus choices to account for a lot of ahistorical scenarios for GER and SOV focuses (when it is a good idea to declare more wars: dont do it if you have already capitulated or you are a subject, do have a border with the target, do have a army worth something, dont be in the middle of another war...).

-USA/Spain neutrality ideas from the mod are visible (Isolationist Sentiment and Non-Belligerent).

-Improved historical mode border-gore fixes (in specific situations, two Ais will transfer controlled states). On the east front, Axis AI minors will give up their SOV territory to AI Germany so that ROM/HUN don't take over the front.

-GER does not send units to Finland if it joins the war against SOV. Avoids a lot of issues.




It's a new patch and a vastly different version of Expert AI with potentially a ton of new balance issues to address and possibly bugs with the new mod features. Feedback from your campaigns is needed to get things working as fast as possible. The best way you can help is by tagging around and seeing what the AI is doing (for example once your campaign is over, reload an older save and see what the AI was doing at critical moments).

HOWEVER: to reiterate if it's not already clear, this mod is really about improving the AI through the "country management AI", not the "tactical AI". There is a 90% chance that I can do nothing about issues relating to how the AI uses its units. There are very few things that this mod does change regarding where the AI sends its units (invasion target countries for example) but the mod does not really change the tactical AI beyond some define tweaks (which can improve some highly specific issues but will not change the underlying logic in how the AI moves its stuff – no mod can do that). If you want to comment on this stuff, I'll read it but I will not reply unless it's something I can change. You'd most likely be better off reproducing the issue in an unmodded game and reporting it to Paradox.

If you have a problem with an AI country doing poorly for example, I need the following information (unless this is really a no-brainer issue) like: historical mode on/off (ALWAYS INCLUDE THIS), which DLCs you DON'T have, date, situation, what led to this situation, who you are playing. Much better if you can find out what the cause is and a possible solution.

If you have an idea about how to improve the country management aspect, I'll be glad to hear it (best to avoid anything too meme-y or too complicated for the AI to handle). Ideas like "AI should do X" are worth far less than “instead of AI doing Y, it should do X for reasons A, B, C". I would rather hear suggestions from someone with several hundred hours of gameplay experience and a thorough understanding of the game because editing this stuff is hard and time consuming and many aspects of HOI4 are misunderstood by the average player.

Of course, anything you can already do by changing the AI's army focuses in the mod is easy to implement. If you think that X country would do better with Y army focus, make the change in-game and try it out.

These are a separate issue: the configuration event, difficulty settings, focus configuration, diplo-decisions etc. If there are problems with these, I definitely want to hear about them. For example if you try to change what your allies are building and it doesn't seem to work or a difficulty bonus should be tweaked or something just seems broken.




-I do not play on ahistorical mode, I have not tested what happens. At all. The AI will definitely perform worse on ahistorical mostly because it won't be smart about taking national focuses (which are extremely important for strategy, needless to say...). There is no special logic added in the mod to make the AI smarter with its focuses on ahistorical (mrkettyo's ahistorical tweaks need to be readded). How much worse is it? I don't know, the mod should still work otherwise. However, if you want to do ahistorical stuff on historical mode, that would work better for the AI (though don't expect miracles). I've added a lot of scripting that should take care of many ahistorical situations plus there are the new strategy plans for the AI in 1.5 added by Paradox.

-The game by its nature is quite random even on historical mode (and this mod also adds more randomness because the AI can change its builds). One game, AI Japan may crush AI China because it surrounded a quarter of the Chinese army. In another game Japan might get crushed because the opposite happened and Japan made a push or an invasion that ended up horribly – you know, just like you can screw up and lose in some game due to a mistake. If this happens and you don't like it, you can usually fix it by doing some reverse-save scumming (reload the game before the AI screwed up) or just by buffing the country which gives it a better chance at compensating for mistakes like these.

-Germany depending on how the game goes does not have a lot of armor divisions for Barbarossa (unless you buff GER). Sometimes it has less than 20x 40 width armor at the beginning and deploys more later. Maybe this is better for the AI as it will still produce a ton of tanks so it can keep them equipped. The reason it doesn't deploy more is complicated but a part of the problem is that it loses divisions in random ways which I can't fix.

-The AI is still reluctant to use its fighters especially against other AIs. Can't fix. I have scripted the AI to reduce aircraft production beyond a certain amount but it will still build a ton and that's intended.

-♥♥♥ vs CHI balance: it's intended that Japan wins the war, however it's a bit random at the moment. I may have to nerf AI CHI when it is against AI ♥♥♥ again but currently it has no restrictions. As I mentioned before, sometimes ♥♥♥ makes some horrible decisions with its units and loses too much. If you're looking for a historical stalemate situation, it's very unlikely. You can try by giving China bonuses but then Japan will likely be too weak to do anything else.

-GER vs SOV balance: not sure who is favored, I see both win. Most likely depends on what kind of an army SOV builds. The war appears to last beyond 1943-44 at least so I'm happy with it. Depends a lot on what the Allies do. With the new invasion scripting for the AI, it should have a good chance at punching through the AI GER defenses.




-AI creates dozens of template designs. This is perfectly fine for the AI to do. If it doesn't do that, then there is an issue.

-AI may have tons of divisions in the recruitment queue. Not really an issue in most cases. The AI prioritises equipment production for its fielded divisions anyways.

-USA/UK may build a ton of units. Intentional. Historical numbers don't work for gameplay.

-GER has a zillion factories and oil production.

-Minors have ahistorically strong armies. Intentional. I don't consider Spain or Brazil or Raj with around a million troops an issue. The mod does prevent minors from spamming 100s of divisions but they are still meant to put up a fight.

-Ahistorical divisions, armies, Germany rushing Pz 4s, not building subs etc. Intentional. Historical EAI mode makes the AI build stuff that is generally historically inspired (like USSR building medium tanks) but forcing it to build exactly historical stuff would just make it bad.

-The AI uses a lot of lend-lease. It's meant to. This mod overhauls the AI's lend lease logic (when, who to lend lease). If you're wondering why it doesn't lend lease you, it's usually because the AI doesn't start lend-lease with anyone that has a surplus of equipment.




I can't answer that for you. Depends on who you play, how good you are and how gamey you plan to play as always. For a world conquest type of game you probably want to use the startup difficulty presets or toggle on various dynamic bonuses. You could start without bonuses and add them later in the game if you feel like it.

As UK/USA/FRA, I would personally never play without buffs to Japan and Germany. As Soviet Union, Germany could be tough even without too many bonuses as I nerfed fort costs and XP bonuses. As Germany, you could try just giving SOV the high Desperate Defense bonus because combined with the nerfed reinforcement rates it may be enough to allow SOV to recover from the initial losses. As China, I would not play without any buffs to Japan due to how difficult the war is for the AI to manage. As Japan you might want to toggle on at least Desperate Defense for China and perhaps even the Reinforcements mode if you want a really harder campaign.