- Redone how the difficulty building cost discount is applied.

From now on, the difficulty itself won't give AI characters any modifiers that need to be negated - rather, it will fire a monthly pulse from every independent AI ruler, which will make them and all their vassals, except for the players and any vassals of theirs, have the appropriate discount for buildings for 32 days, which is extended back to 32 days each month if still valid (longest months are 31 days, event needs 1 day to apply).

In other words, the player should now never even see the modifier, except for up to the first month (up to 32 days, average 16) after they vassalize someone, or after one of their vassals dies and their heir inherits. It was more like a couple days beforehand, but this is much cleaner, and I can always shorten that time by adding more pulses, though I don't want to overdo it before first testing if it doesn't slow the game.

I could just make building stuff in the barony in the meantime impossible, but that messes up with AI building logic, since it would basically banish the buildings from its long-term planning if I do that, since it'd see them as no longer valid to build, and spend its gold on something else.

- AI will now (in simplified terms) retreat from a battle if it judges itself to have <60% odds of winning it, rather than <40% (going the other direction turned out to produce much better results)

- Logic changes to make defenders and attackers that are losing less agressive, and more coherent in their army movements. I've extensively tested it in the following scenarios :

Big vs Big, with advantage on one side - Seljuk invasion of Armenia :

Around 50/50 odds of either side winning, despite the overwhelming early Seljuk superiority in numbers - because the Byzantine defenders avoid engagements not on their terms, stall for time, and take time to regroup after each defeat, meanwhile the Seljuks have no choice but to continue sieging and taking attrition. Either they succeed in accruing 100% warscore, or eventually they lose the vast majority of their special troops, , at which point they cease being able to push forward, and the byzantine army starts pushing them back.

Overall very happy with this scenario, and I've in fact achieved worse results than the Byzantine AI when I tried!

Medium vs 2 Medium in separate wars - English threesome :

England competently switches between fighting off the two sides when able, and taking breaks to replenish its troops whenever possible. Sometimes it does so when it can't afford to timewise, leading to a defeat, but that's fine, especially since Norway/William together have great numerical superiority. I'd say it handles the scenario okay, though definitely not as well as a good player.

Big vs 2 Medium - HRE war on France + William's England, both sides evenly matched :

The defender starts sieging the border before HRE can gather all the troops (as intended because mustering the levies takes time for such a large realm) - and then the HRE tries to siege the wargoal, at which point France&England move together and a big battle ensues, which the HRE usually loses. Even if it doesn't, the French/English move back to regroup while HRE tries to siege, and they beat it back with replenished numbers. HRE loses 90%+ of the time, which is great, since its pretty damn strong. AI logic is fine there, too.

Small vs Small - 4 county minor Dukes, evenly matched :

There is a lot less manuevering around and battles - mostly decided by sieges, unless one side can catch the other in favourable terrain. I'd say its not ideal, but not bad either.

Tiny vs Tiny - 2 counties vs 3 counties, evenly matched (Ural scenario) :

Playing as Ural and trying to siege the enemy capital is mostly a binary outcome - either the AI lets you do that because it thinks you stronger (even numbers), and hence loses the war because it is unable to siege down the third province of yours before you siege its second and last and get 100% warscore - or, it starts with a slightly larger army and wrecks you. I'm a bit unhappy with how this one turned out, but you can't have everything, and you only ever spend half a year as a 2 county count, anyway, so this can be sacrificed in the name of bigger wars being more challenging.

- Reduced added flat occupation score per barony to 37.5% of what it was, since the AI now knows how to siege better.

Example : 10 baronies occupied out of 200 total (5% of the realm - keep in mind one siege can net you as many as 3-5 baronies) :

Vanilla values : Meant 5% + 40% = 45% occupation score, multiplied by a 90% multiplier -> 40.5% warscore.
Now : Means 5% + 15% = 20% occupation score, multiplied by a 90% multiplier = 18% warscore.

This change makes large realms substantially tougher to occupation in wars, making battles more important and decisive for the outcome, as they should be. The change doesn't really affect small realms (counties/duchies) much, but it does affect wars like the 1066 threeway for England.

- AI now always sets the marshal on "organize levies". Previously, it was using it as the backup option, for some reason. 75% of the time it was on "increase control in county" (pretty useless 99% of the time), and 25% of the time on "train commanders", which can be an okay option but lets face it - organizing levies is the optimal one the vast majority of the time, and is never bad. The previous logic also caused all realms to pretty much always have less levies than they really should, which made the game easier for the player, particularly early on.

- Slightly lowered enemy capital sieging priority for the defender, which seems to make it behave smarter in very small wars (counts vs counts),

- AI will now (in simplified terms) retreat from a battle if it judges itself to have <40% odds of winning it, rather than <45%.
- If the AI chooses to stand still in defensive terrain and wait to be stackwiped in a situation that is deemed hopeless, it will now only do so for 12 days, down from 30 (and then won't do so again for 90 days if it didn't, in fact, die.).