• We have updated our Community Code of Conduct. Please read through the new rules for the forum that are an integral part of Paradox Interactive’s User Agreement.

EU4 - Development Diary - 24th of October 2017

Good day all. As we reach ever closer to the release for Cradle of Civilization which, as you know will be released on...wait, we haven't announced the date yet? Well keep your ears to the ground and you may hear something about that very soon™. In any case, running up to that we have a military mixed-bag of features coming with Cradle that we want to talk about today, as well as a guest appearance from our AI Maestro @Gnivom .

Firstly in Cradle of Civilization we are bringing the ability to Name Generals. Does what it says on the tin: when recruiting a general, you will be able to give them a custom name or roll the die to generate a random name for them. This courtesy is, of course, extended to Conquistadors, Admirals and Explorers.

Name Generals.jpg


As an additional free quality of life change in 1.23, the number of pips your leader may receive is detailed in the tooltip, showing the random spread they will get as well as any bonuses you may get from Ideas and other such modifiers.

Staying on the Military Tab, we can also show the interaction to Slacken Recruitment Standards which will let a nation gain manpower in exchange for allowing somewhat less qualified men into the army. 2 years' worth of manpower can immediately be made available at a cost of 5 professionalism. Naturally, a nation needs to have at least some standards of Professionalism in able to slacken them.

slacken.jpg


Another small addition we have in Cradle is the ability to have compact battle results. This has been requested by the community in the past so we decided to add it into the expansion for those who don't need more than the bare minimum information punctuating the slaughter of war. This can be toggled on and off in the options menu

compact.jpg


Finally, sticking with the military theme, let's talk Templates. They are pretty awesome and remove a lot of micromanagement of recruiting armies. In Cradle of Civilization this is taken a step further with the ability to Conform to Template. In the Unit View, an army can conform to a template that you have specified earlier. Upon selecting a valid template, all superfluous units will detached from the army while required units are built and march directly to merge with the unit.

conform template.jpg


As this feature requires the ability to make templates (an Art of War Expansion feature) to be of any use we are unlocking that feature for owners of Cradle of Civilization too.

Now I'm handing the mic over to the Maestro of AI @Gnivom , to talk about himself and the changes to the AI in v1.23 Persia Update.

###########

Black News Indeed

I have worked on EU4 since February 2016. For the last year or so, I have had the great honour of being the master of its AI. But all things must come to an end, and I have decided to move on to other adventures, outside of Paradox. It is therefore my equally great honour to present to you my successor, @mikesc . These last months, I have taught him what he needs to know, as I was taught last fall. This will be my last week, so today is a good time to talk some about the AI.

AI Improvements
There has been much AI work put into 1.23. I will mention a few improvements here, and for the rest you are referred to the upcoming changelog.

Building AI
This may well be the most notable change, at least for those of you who often play into the late game.

The AI has always been horribly bad at deciding where to build buildings, which means an experienced player could easily be many times richer and stronger than the AI after only 100 years, with the same development. This has now turned around completely. Instead of deciding buildings based on scripted rules, it now calculates precisely how much it will gain from building each building, convert all values to ducats, and consider how big a treasury they need for the return on investment to be worth it. The reaction from QA after their first handsoff game with the new building AI was “You have created a monster”. Indeed, late-game AI is extremely rich.

It probably undervalues manpower and forcelimit at the moment, but that may be for the best.

This doesn’t only make the game harder however, as an expanding player will get a lot of free buildings.

Power Balance Threat
One feature in the AI that I believe was added by Chaingun, is called Power Balance Threat (PBT). It means that each AI determines another country, if applicable, that it considers a long-term strategic threat/rival. It will then try to weaken this country in subtle ways, such as sending gifts to or guaranteeing their neighbors. Allies can also be picked as PBT. The algorithm for determining PBT focuses on large and/or quickly expanding nations. Since this very often targets the player’s nation, the algorithm has been explicitly prohibited from selecting the player’s country, except on Hard and Very Hard. My understanding is that this was to avoid the perception of the AI targeting the player because of it being the player. This block has now been removed on Normal difficulty as well.

You may now see even your allies act to slow down your growth, and some countries getting worse attitudes towards you, but don’t expect a huge difference.

Other Changes/Improvements
Many people have complained about the AI’s use of Edicts, and correctly so. From 1.23, it will spend a lot less money on Edicts.

AI aggressiveness has been slightly tuned down, especially when in debt.

Many small improvements have been put into the army AI. For example fixing many cases where it pathed through, rather than around, enemy armies (often leading to just cancelling the movement anyway).

Many more improvements will be found in the changelog.

I am currently spending my last week at Paradox trying to make some major improvements to army AI, that will be incorporated after 1.23.

Cheats

My personal opinion has always been that the AI should cheat as little as possible - that the player should play the game at equal terms with the AI. I believe the EU4 AI follows this philosophy more than any other strategy game I have played (at least on normal difficulty).

The cheats that do exist are typically not to make the AI more powerful, but to work as lubricant where the AI is not as precise as the player (which could be to avoid more complicated and bug prone code, or unnecessary computing power). A good example is the extra diplomat, which is only used for one-time actions. The obvious exception is the Lucky Nation bonuses, which are not available to the player.

So, let’s mention some of the cheats that do (and do not) exist:

There are maybe a dozen places in diplomatic AI, where it is nicer towards other AI in specific cases that make sense but could potentially be exploited by the player. It may for example call in allies as if with Promise Land clicked, without then being bound to actually give land. It will however try to give if possible.

The AI does not cheat with dice rolls in battles (or sieges, or anyplace else that I can think of)

The AI no longer cheats with Fort Maintenance, since a few patches back.
EDIT: There is still a bug that they can cheat anyway sometimes. See https://forum.paradoxplaza.com/forum/index.php?threads/1-22-2-vassal-not-paying-for-forts.1052067/ .

The AI doesn’t cheat with Fort Zone of Control. Nor was it ever meant to, although there has been at least one bug in the past where it could in rare cases.

Other than the Fort Zone of Control, the AI/Player cheats currently mentioned on the wiki are correct (to my best knowledge).

Additional considerations:

  • The AI may sometimes look up underlying numbers that are not available to the player.
  • The AI sometimes restrains itself to get a more historical outcome.
  • The AI is not allowed to pause the game, and has to play hundreds of nations simultaneously. Nor does it use the extra time it could get when the player pauses, or plays on speeds lower than 5 (so that behaviour isn’t affected by game speed).
/Gnivom

###########

Thanks a lot for that Gnivom. Next week we will have another guest speaker from the team to talk about a huge project we've undertaken for the game which, so far, we have managed to keep completely secret and will be rolling out with 1.23. See you then!
 
Last edited by a moderator:
  • 1Like
  • 1
Reactions:
Damn, lots of small changes, nothing spectacular, but lots of stuff who will make my next games way more enjoyable and/or less tedious, thanks to those little changes. Thanks !
 
Is there a cooldown on slackening recruiting standards or is it only limited by your professionalism?
 
Next week we will have another guest speaker from the team to talk about a huge project we've undertaken for the game which, so far, we have managed to keep completely secret and will be rolling out with 1.23. See you then!
Could you by any chance tease us with the vaguest of screenshots?
 
Thanks for leaving us with these amazing AI changes @Gnivom. o7 And many thanks to the rest of the AI coding team as well of course. Really curious to see how richer AI countries will play out. Might just play on hard right now to check out PBT.

Next week we will have another guest speaker from the team to talk about a huge project we've undertaken for the game which, so far, we have managed to keep completely secret and will be rolling out with 1.23.

Whatever could this be!? A multiplayer addition? More historical starting dates? An update to old unit pack models? Vicky 3 with savefile conversion? The possibilities are endless!
 
Building AI
This may well be the most notable change, at least for those of you who often play into the late game.

The AI has always been horribly bad at deciding where to build buildings, which means an experienced player could easily be many times richer and stronger than the AI after only 100 years, with the same development. This has now turned around completely. Instead of deciding buildings based on scripted rules, it now calculates precisely how much it will gain from building each building, convert all values to ducats, and consider how big a treasury they need for the return on investment to be worth it. The reaction from QA after their first handsoff game with the new building AI was “You have created a monster”. Indeed, late-game AI is extremely rich.

It probably undervalues manpower and forcelimit at the moment, but that may be for the best.

This doesn’t only make the game harder however, as an expanding player will get a lot of free buildings.
I'm worried that this will mean even more merc spam from the AI as they will be even richer.