LeaderEnemyBoss wrote:

@Abrasax If an empire has no coastal cities that (most likely) means its surrounded by other empires. I dont think its very likely they will declare war on you if you dont share an island?

True to an extent. They could be surrounded by allies, or at peace with their neighbors. There are plenty of reasons why a faction with no coastal cities would want to set sail:

• They need to expand and there are known islands nearby
• Someone overseas has declared war on them and they want to launch a counter-invasion
• They want to wage war for a specific city housing a Legendary Building
• They want to set sail for a specific quest that requires Shipyards
• They need a specific resource that is only available on another island/continent and need to send a Settler overseas
• They are down to one or two cities and sending a secret settler to another land is their only hope for survival

Jojo_Fr wrote:

Vaulters are already good with Fortified cities and teleport.

You can reinforce sieged cities.

I am more worried about the Cultist.

If the guy has 2 armies of 6 units inside + 1000 of fortification (it's doable) + Reprisals capacities (from pearls towers) + two armies of 6 units inside the city, will you really siege the city ?

First turn, you will take huge reprisals damages. In three turns your army will be very hurt.

And the Cultist can still attack at any time and kill you.

That's fine because Cultist city are immense fortress, but it looks like a non doable job.

I think the problem come mainly from the fact that the fortification benefit from any unit inside the city districts. Maybe it should works only on milicians + one stack maximum (so milicians + 8 units max). 

Is there a way to code it so only the units inside each city gain their respective city's Fortification Bonus and any ? And possibly make the health bonus for reinforcing only Fortification/2?

- City 1 (initiating): Fortification 150; 4 Units Garrisoned > these 6 Units gain 150 Fortification each

- City 2 (reinforcing): Fortification 450; 6 Units Garrisoned > these Units gain 225 Fortification each

Just a thought ...

Also, there are ways to reduce reprisal damage and high regen will ultimately negate such reprisal.

Jojo_Fr wrote:

Vaulters are already good with Fortified cities and teleport.

You can reinforce sieged cities.

I am more worried about the Cultist.

If the guy has 2 armies of 6 units inside + 1000 of fortification (it's doable) + Reprisals capacities (from pearls towers) + two armies of 6 units inside the city, will you really siege the city ?

First turn, you will take huge reprisals damages. In three turns your army will be very hurt.

And the Cultist can still attack at any time and kill you.

That's fine because Cultist city are immense fortress, but it looks like a non doable job.

I think the problem come mainly from the fact that the fortification benefit from any unit inside the city districts. Maybe it should works only on milicians + one stack maximum (so milicians + 8 units max). 

Your cultist fear is a bit too much theory crafting in my oppinion. I wont nerf a feature that is intended to buff Fortifications only because of that one theoretical scenario. There is also the fact that you can pillage towers - there is really no reason for letting them stay at all, or is there? Please play with the feature before thinking about how to break it ^^. And again: them having only one city is a big weakness military wise. Its basically a giant bulls eye that says "attack me". If you want to conquer a heavily fortified cultist city, youll have to destroy the villages and maybe catch armies outside. If retaliation becomes indeed too strong, i'd rather weaken that instead of making abitrary changes to who gets fortification Armor that would hurt ALL factions, not only cultists.

Abrasax wrote:

- City 1 (initiating): Fortification 150; 4 Units Garrisoned > these 6 Units gain 150 Fortification each

- City 2 (reinforcing): Fortification 450; 6 Units Garrisoned > these Units gain 225 Fortification each

City 2 would currently not grant any fortification armor since its not a city battle for that city. (The fluff logic is, that only reinforcements, that can join from within the besieged cities walls, gain the Fortification armor of City 1).

Ninakoru wrote:

AI

• Improved Unit Designer logic for heroes
  • AI now ignores the Item Cost of the current design, when comparing with a hypothetical new design (since the cost has already been paid)
  • AIs will heavily favor Iron Talisman on Generals that slow their army down
  • Heros will now heavily favor Items that fit their current role (Governors will mostly ignore Army Items, Spies heavily favor scopes etc.)
  • AIs now take their strategic resource stock into account, when deciding on unit designs, not just their income

Nice changes! About that specific change, something that bothered me long ago when I observed the AI is how the AI evaluates DUST as a normal strategic resource for dust-based gear. As They cost a lot of Gold they usually skip most of the dust items past tier 1 and all dust accessories because they just cost too much dust. This is important because gold items are key when you don't have the strategics for better gear (BTW the gold HP percent accessory is really good and is a staple in my armies).

I already adressed that issue some patches ago, it mainly had to do with that the AI only took their dust income into account when deciding what they can affort, not their dust stock. Also their Dust Accountmanager was capped at a very low value, which often led to it prefering dust 1 or even iron, while sitting on like 20k dust. These are all issues that sadly are basically not adressable with xml modding. Even getting governors to actually buy their damn governor items reliably is very hard with just xml changes due to how the scoring is coded internally. I dont say this to be mean (its not like its your fault) but over the time I noticed that many attempts of your mod when it comes to unit/hero equipment where basically in vain due to how everything works internally in the end (but all of it was very helpful for learning stuff! :) ). I uploaded all the files I changed with elcp just yesterday (with difs to their originals), so If you ever want to have a look: [Link]. You are also welcome to our discord for questions ;).

Yeah, I had lots of headaches due to how AI makes their unit design decissions... With little improvement. Great to hear all this! I'm really excited about this mod, I'll try it fully anytime soon, maybe help a bit with the AI department, but seems you have mostly all covered!

I've seen your git upload, did you upload the unmodded assembly as an initial version?

Thanks for your testing. back to your previous comment, while I have done many things, by no means do I have "everything covered". I personally think there is not much more to do via .xml changes (apart from the occasional tech priority and diplo weight change), but the juicy stuff is in the core files. Another community member for example is currently looking to improve the general battle behavior of units via core file changes, and there is still much more to do - as you have noticed.

Ninakoru wrote:

I tested how goes the game in endless up to turn 55, just some issues I found down the road, please don't take it as criticism, those are just things to improve, or maybe I'm just missing something.

- My buildings sent to a city queue go now to the top of the queue instead of the bottom.

- this happens if you press shift when queueing normally ... when just selecting a building to construct it queues at the bottom for me, cant reproduce

- Forgotten still lag a little behind in tech (tested on endless), however now works much better, not a big deal.

- The forgotten AI - same as human players - have a problem with the sharp rise in tech cost. Due to their spying activity they also often are involved in wars, which end badly for them more than not (due to situational units). All in all I would say this is more a sign of their general weakness compared to other factions.

- AI for some reason sometimes leave city queue empty (very rare cases, seems now new AI cities get their build order).

- I have specifically designed a new routine that catches these cases, but its still a bit primitive, can certainly be improved upon.

- Seems Faction assimilation is still not being reasoned by the AI at all. An AI has as example 2 assimilations, both of them are of 1 village, and have not assimilated but pacified three other minors with 6, 5 and 2 villages (MUCH better options). They should be able to re-evaluate their assimilations based on unit type filling gaps in army composition, and number of villages for obvious bigger bonuses and reduced industry.

- I have not touched assimilation at all ...so yes ;)

- We know something about the bug that makes AI think settler is a combat unit and makes it non-stop? I didn't see the AI falling in such a bug this time.

- I only saw that bug happening when cheats where used (give AI tons of dust and see what they are doing with it), so I never saw it in a "normal" game. I personally think it has to do with settlers in a city garrison not being counted as settlers, so the AI thinks it needs more. Since I have created a routine that forces the AI to empty overcrowded garrisons when building new units, its possible that I "fixed" the issue unintentionally with that aswell.

- AI still ignore building legendary buildings, while they shouldn't be a priority, in endless being able to make nearly every single legendary building is something that shouldn't happen.

- I left legendary buildings largely alone for now. Dont get me wrong, its easy to get the AI to build them, but that would basically mean human players never get them on higher difficulties. Its a tradeoff between "intentionally gimping the AI" making the most of the game goals still achievable on higher difficulties. They could use a bump in building them however, I agree.

- I think AI not comparing the design with the previous is an error, in a game I'm testing, some AI have unit designs up to "Version 18" 44 turns into the game. This is a huge money sink if they try to retrofit those units all the way.

- Them updating their unit designs regularily is intended. What happens if you make it too restrictive, is that the AI may even keep designs around they cant afford anymore (too few strategics), that can leave them unable to build units at all (which is worse imho).

-The AI does compare old designs and new designs (try /aidebugmode in unit designer to see the evaluation). 

-Honestly I have not investigated how retrofit happy they are for normal units. I know where the retrofit"jobs" are created, but I never looked into it. If it does indeed waste too many resources on retrofits, I would rather adress that issue directly instead of them not updating their design often enough.

- A forgotten army design on turn 55 (all versions are range 20-26 BTW), only with a strategic weapon and iron accessories and NO ARMOR at all, with all iron tier 3 gear available. Something went wrong in their decision there. The other forgotten also left empty armor and head slots.

- I would like to have a save of this so I can check aidebugmode whats happening there.

- In endless AI doesn't stop making units, going quantity over quality. Turn 45, 68 units for a 7 city empire (near 10 units per city is too much), and 4 more units on the queues.

- Do the cities have (non-wonder) buildings to build? Are they in a war?

- Seems AI doesn't make many heroes, Drakken and Mezari AI have only 1 hero on turn 55. Both Forgotten AIs go to 3 heroes (one is from questline), two spying one sitting on cities.

- They really like to buyout buildings, thats why it takes a while until they get their heroes going (they typically get a bunch of them in the turn 50-100 range). I honestly dont know what's better but i dont think its the most egregious problem.

Ninakoru wrote:

Let me help a bit with what I remember on this issues:

AI "Hero Need" is governed with:

AI/Registry.xml

AILayer_HeroAssignation\MaximumTurnForHero: 120 turns!!

(as the AI closes to this value, their interest on hero rises).

AI/Personalities[AffinityCLASS].xml

AILayer_HeroAssignation\MaximumTurnForHero: 0.8

(I guess this cuts the turns from the registry by a 20% for example)

I don't remember the details, but it should affect both the priority on hero market tech and the priority on hero buyouts. Can't help much more here, I also remember I couldn't modify the behavior tweaking those, so it can be as you say, if they are not able to save money, they don't reach the phase of hero purchase.

On the unit designs, we have:

AI/Registry.xml

AILayer_UnitDesigner\NewUnitDesignScoreModifier: 0.66

This means the AI will only upgrade the design if the new design is considered 33% better than the old design (multiplies the design evaluation with 66% and only upgrades if it gets a higher value)

From what I remembered, it worked fine, but far from perfect, designs are re-evaluated every turn taking into account strategics available at that time IIRC.

BTW "AIEquipmentMaximumMovement" didn't work, so I made a workaround for the settlers to let them use movement accessories.

I know .. I know .. Lowering MaximumTurnForHero isnt that good after a point, because it basically forces the AI to buy heroes until they reach their designated limit when they exceed that point, no matter if its useful or not. Currently the AI develops its economy first (at least in theory) and then begins buying heroes en masse. I personally think thats a good approach. I already basically force the AI to buy heroes if they reach a set amount of money and are low on heroes. I have lowered the threshold of that amount a bit in my internal build now.

The unit design score depends on the weighs assigned in the xmls but also on the the cost of the specific item (items get assigned a lower score if AI thinks it can barely afford them). It is not used at all for retrofitting decisions (except heroes, where retrofitting and design changes are one and the same). I understand gist of the UnitDesign score calculation, I dont see a problem with them getting updated regularly. Setting NewUnitDesignScoreModifier too low lets the AI keep design even when they cant really afford them, believe me, I tried it. 

Hello there!

Your patch is great indeed, I've really enjoyed it. But may I ask you what is your goal? I mean what do you want to improve like first of all?

I like to play multiplayer games and at high level of play you can see some really frustrating balance shortcomings. We've allready tried to fix them somehow, buffing and nerfing some features, but still the the work was not done at all and at last was forgotten. So I ask you if you will be interested to add some good ideas from these threads or to discuss them here at least:

Multiplayer balance mod

Suggestions&discussion part 1

Suggestions&discussion part 2

The rules that shouldn't exist

Thank you for attention!