Some advices on how to improve the AI

[Lama43]

Hello, i've decided to open a separate thread for this sort of little battle of mine that i've been conducting on some topics.
As we all know the EU3 AI is just dumb. To cite some of its problems:
- Non-sense land grabbing (hello Castillian St.Petersburg)
- Non-sense diplomacy (driven purely by emotions, and not very deep ones either)
- "Uneffective" land tactic AI (stacks troops and gets huge attrition some times, while leaving a close-by army to die against superior foe other times)
- Limited knowledge of naval tactics (patched up by cheats like no attrition and knowledge of the position of every enemy fleet)

I think these problems are undeniable, and while in some cases they can be at least lightened by mods, in many others the problem is simply at the core, and cannot be fixed.

Here's a list of some posts i made earlier that give specifical advice on how to fix the problems (i am personally involved with developing games, i understand that it's important to give specific advice)

I think you can introduce a fairly simple goal priority system. For example, there's Russia and there's Poland. Poland holds Russian cultured territory which they have no core on. Let's say that taking those provs takes 5 priority points.

Then, since we don't want to have Russia declare war in case Poland is stronger, we apply a simple percentage, that is the ratio between their strenghts (Russia has 50 regs, Poland 60, the ratio is 0.83, or 83%), so we have now 5 * 0.83 = 4.15 priority points. If Russia and Poland have both 200 relations we subtract some points, let's say 10 (i'm using arbitrary values here obviously), which means that the value is negative and Russia will not actively try to forge a CB on Poland, nor seek to enter coalitions against it. If they had bad relations (like -100) we could add 5 points to each priority against Poland, meaning that Russia WILL actively try to take down Poland. If the value is close to 0 it's possible to make a passive stance - Russia will take spontaneous occasions to take down poland (eg. Poland at war with ally)

And with this system, each country would have a priority on each other, and if that priority is much higher than 0 they will try to attack them, but not if there are more pressing matters (for example a dead Golden Horde for Russia, very easy and tempting target). Different AI personalities would setup different goals, for example Switzerland would have none, England would have the goal of uniting France or the British islands and so on. And when that goal has a high priority the AI tries to achieve it. With some tweaking it would allow a good balance between emotionality and rationality, and no need to write everything from scratch.

This is to more or less fix the diplomacy and also the land grabbing in a way.

Improving the AI should actually be one of the main focuses of EUIV, and you devs can't just say "the AI wouldn't handle it". MAKE IT handle it, it's the same story as with naval attrition, it's enough to slightly modify pathfinding to prefer coastal routes, make the AI seek a port to shelter in ASAP if it gets below a certain health value and other tweaks i'm sure you'll devise by testing. You can even make the AI cheat by calculating precisely how much time it would take to reach the closest port, and decide whether it can procede with the travel or it must return to the port or not. Humans do that anyway, though in a less precise way.

The military AI is lacking as well, i've seen so many times that the AI simply watches its other armies die in an uneven fight, even though it has sufficient forces close by. This is extremely easy to fix, i'm sure there's a factor modifier somewhere which makes AI prefer to keep sieging rather than fight, it's enough to just make the AI fight first and then siege, as it actually happened in real life and makes sense to do in the game. This could potentially solve the problem of tiny armies sieging distant possessions (very pissing off and unrealistic) by making it help its allies on the battlefield (although yes, this would also require to improve cooperative AI).

And this is to address naval and land AI. Let me expand a bit the latter.
In real life, battles were usually massive and in some cases decisive (an example is the Battle of Hattin, where almost all of the crusader forces fought against a big chunk of the muslim forces). The fundamental problem of EU3 is that the AI starts sieging immediately, without caring of the enemy armies. While of course it should eventually siege, this must be done with small forces that get detached once an army advances. The basic warfare should be conducted by assembling armies sometime before the battle would start. In short, the key is to make the AI actually concentrate its forces when needed.

I hope this thread will at least be seen by a developer, i really would like EU4 to be an excellent game, and to make it so it has to overcome the problems that plagued EU3.