Is it just me, or does France play too defensively?

[Chaingun]

Yes, but the AI is considerably more passive this patch than I can ever recall.

If this is the case, you might wish to poke @Gnivom. It's typically necessary to check that the AI still has a "proper" amount of aggression towards the end of a development cycle, because other changes tend to mess with its tuning for war declarations.

However, let me stress that what actually happens in a single campaign is extremely random. One needs to have a lot of games to discern any real patterns. Especially, war declarations are quintessential RNG, weighted by game factors to make the AI's chance of declaring war higher or lower.

If every country played to max aggression pretty much every OPM start would be impossible.

This also a key point. Basically, the AI can't treat the game as Risk (that many players do) because of how fatiguing that would become for most of the player base. There are numerous gameplay mechanisms to make the AI expand up to a certain point but little further (most apparently country specific missions). Admittedly this is somewhat tangential since France seems under-aggressive compared to their "script" here, but see the RNG point above.

 

[Chaingun]

Sure, if their calculated war declaration probability is high, they will tend to declare predictably. But it can also lead to things like Ottomans not expanding whatsoever solely due RNG in very rare cases, so it's still not deterministic.

Yep, the AI goes along like a lamb to slaughter. I was working towards some higher degree of responsiveness for AI countries during my time as AI programmer on EU4, but it somewhat disagreed with the game design view that the AI should be that lamb, so it's in some ways being reversed rather than fixed when a problem appears (e.g. condottieri).

The fact of the matter is any complicated AI one adds tends to have large open flaws that are exploitable. E.g. the fact that AI armies now try to stay alive has become that as of 1.18 (can't comment on 1.19), and hence needed further development because veteran players would unavoidably detect and exploit flaws in the more advanced behavior. (And then we have the usual topic about behavior being oriented towards the common case, but where AI will go full R-word in some instances.)

It really takes a bit of time for players to reach proficiency in exploiting the AI (and it's highly individual with a range of some people picking it up incredibly fast, others never doing so). So you will find that what's true for you is not true for everyone.

 

[Chaingun]

Diplomacy war declaration AI (dogpiling, refusal to declare on stronger player) and troop movement AI (troops running away) are two very different things and must be clearly separated in a post to avoid confusion when discussing them.

I can't tell you much about 1.19 but I guess army movement issues are recent (1.18+) whereas the other issues you speak of are far older.

On army movement, it's pretty clear what's wrong from a human perspective, it's more a matter of working further on the code - what's simple for a human is algorithmically complex because army movement is impacted by many conflicting factors, e.g. enemy forces, attrition, regiments needed for siege, etc.

On diplomacy, it's a longstanding issue and there are few easy solutions, because any simple behavior will be exploited. AIs don't speak to each others and it's one reason for their inability to do anything against a strong player nation. Should they be made able to? Opinions are divided on the subject, game designers generally want AI to behave according to simple transparent rules - that will easily be gamed by the player. The problem is, if AI are able to coordinate but player <-> AI is not, then the player will be alone and it will be all the more of an "AI targeting player" experience. Of course, game designers could put effort into improving the alliance system (because a fix number of alliances inevitably falters when the player owns a whole continent).