In a fairly close siege for the AI Cultist city, the AI chose turn after turn to embark and disembark its troops over the adjacent coastline for no reason. When they tried to break my siege, a significant portion of their army was on a boat and couldn't help.

Maybe update the OP with a feedback summary of things so far?



some thing i noticed:



a) When at war the Roving clans often send me messages that war isn't good, but don't offer a truce just that message. when i click 'start negotiation' and offer a truce they refuse.

Is this a bug, or the roving clans AI spams this message because they rely on trade victory? but then they choose to refuse on other parameters.



b) will an AI decide to truce/make peace based on what other enemys they are at war with? how smart are they on judging their strength compared to other factions - e.g do they use information similar to the statistics in empire view? this example was over several turns.



The AI necrophages were destroying the AI broken lords, but the Broken lords wouldn't truce with me(i wasn't going for elimination/expansion and had enough regions so i offered) and decided to keep fighting me as well, so got weakened on two sides and knocked out (necrophages did the most I assume, they looked to be the ones who eliminated them)

- is the AI dumb and only takes into account interaction one faction at a time? can they make assumptions/projections about things, like fighting multiple enemies is harder?

- Is this because they considered a better chance of surviving taking over my territories and losing their's to the necrophages?

- did they misjudge their capabilities/strength ?

- are they tricked/get greedy so won't truce if you have weak (or sacrifical) territories but actually have strong armies elsewhere?



against AI I can frequently can move armies around via roads and get terrories back they take, and destroy any armies left behind with a couple of my strong roaming/questing armies. is this ok?



c) a necrophage army stayed for several turns as a ship in an inland sea after i got close to them, and then took over their territory. is it exploiting this to not get attacked, or dumb AI? why doesn't it move again faster, instead of sitting there for many turns, even after my armies left?



d) in battle why does the AI for your own army's targeting(and i assume the enemy AI) like to attack the unit furthest away of a weakest type? there can be closer units of the same weak type, if they were attacked first it might prevent them doing damage(iniative dependent). there doesn't seem to be any factor due to proximity or damage being done(aggro?) to pull the ai targeting attention that it is a threat.



It's a pain this is the fallback behaviour when the first priority target you manually selected has already died.



e) is the information in the statistics view giving us players an unfair advantage over AI, seeing who is strongest and needs to be taken dow a notch, or who can be eliminated easy, or do they know this too? should it require some sort of diplomatic embassy with each faction?

I mentioned this in another post here, but I'll talk about it here as well since I suspect I noticed this on account of bad AI



-I suspect that the AI may not consider room for expansion when settling their cities. It would seem to me that they tend to settle in the spot that objectively has the most fids in the immediate area and then expand in accordance to what spot gives more immediate fids. This is normally a safe strategy, but there are situations where settling on an oasis of fids leaves you with a barren wasteland to deal with upon expansion. Often times, the best option is to settle in an initially awkward position so you can rope in 2-3 anomalies through triangular expansion.

And I'm not sure the AI takes into account the fact that they could build things that enhance tile output, like with rivers.

When planning to settle, the ai needs to see their starting location in terms of what the city will produce when fully expanded, not just initially. (and, to avoid further problems, should be able to weigh optimal locations against the likely amount of time it would take to get the best tiles while keeping the city happy and bla bla bla like that)



Done right, the more difficult ai could become less reliant on buffs given by harder difficulties to be challenging.



-I also noticed that the necrophages, at least in my playthroughs, have a slight tendency to settle directly on the border of a territory. This is likely a bug, but it actually has strategic potential if it was done better. (as I mention in the other post).

Properly implemented, defensive empire planning could go a long way into making a more challenging ai.

Thank you! Here’s my update on what the AI should do:

• Plan movement of the whole army instead of each unit during the battle
• Move out of a reinforcement tile when reinforcements are still waiting during the battle
• Retrofit its units faster/better
• Take into consideration the best angles to attack (reinforcements, deployment zone, etc.)
• Focus more on strategic resources (on this point, the last update improved a lot how AI prioritizes strategic resources)
• Being able to not include reinforcements if the battle seems to be doomed
• Use the market smarter and more often
• Use diplomacy to stop winning empires; for instance, by organizing a common war against the growing empire with the weaker empires as allies, breaking a peace with a player close to a diplomatic victory, or not trading techs from the 6th era
• Be less exploitable when trading via the diplomacy
• Should not target first the militia during battles
• Have better army mixes
• Make better hero assignations
• Be able to equip accessories
• Use faction abilities such as Force Truce or Vaulters’ teleportation
• Have a better research selection/prioritization
• Handle the units better right before siege and during siege (pulling units out of city before siege and not leaving units to die in cities after fortification is gone)
• Have a better awareness of frontiers: which city should be attacked? From where?
• Not send all its units against one wounded infantry during battle (in other words, the AI is easy to lure)
• Aim a victory
• Not let armies stay into the water for no reason in order not to miss potential reinforcements
• Have a better estimation of its own military power and enemies’
• Choose a spot to settle by taking into consideration the future expansions and not only the city center
  
  

Manu wrote:

• Use diplomacy to stop winning empires; for instance, by organizing a common war against the growing empire with the weaker empires as allies, breaking a peace with a player close to a diplomatic victory, or not trading techs from the 6th era
  
  

I just want to question this one a little: I think breaking treaties simply because a player is close to a diplomatic victory would be super annoying, and make an already rare victory condition even less appealing. The ideas related to other victories make better sense.

I just want to re-emphasize for the sake of consideration, when picking a spot to settle, it may be worth while to have the ai at least consider settling closer to the edge of a region for that sake of having a future city be able to support it in event of a siege. Especially if an empire is focusing on a military victory and wants to make their own cities a bit more defendable

I just want to question this one a little: I think breaking treaties simply because a player is close to a diplomatic victory would be super annoying, and make an already rare victory condition even less appealing. The ideas related to other victories make better sense.

I'm not sure, it may be best to see what happens with this kind of hostility implemented. Points toward diplomatic victory are still earned while in war and cold war, all be it very few, and by the time one gets close to diplomatic victory, other players will be close to other victories, meaning people will want to team up to take out other winners, which requires a level of cooperation sooo....

Stick it in, see what happens. Human players would respond like this so why shouldnt the ai?

A lot of these recommendations regarding settlement are of questionable value.



Long-term planning gets a lot of real-life human players in trouble when they neglect their immediate future for the late game and lose their capital early because of it. Settling near borders may allow militia support, but it also means that closed borders are less meaningful.



A lot of AI improvements could be minimized at lower difficulty levels-- for instance, I think it's appropriate for AI to deny players diplomatic victories at lower difficulty levels, but never at Endless difficulty*. Preferably, AI "intelligence" could be separated from AI bonuses. The current AI is often a spoiler for multiplayer games, enough that it's common to see agreements ("nobody touch the AI") when dropped players lead to AI-controlled empires.



But Manu's latest list strikes me as pretty complete-- or, at least, ambitious enough.



*edit: oops, i said this backwards. The AI should deny diplo victories at higher difficulties, but possibly not at easier ones.

natev wrote:


A lot of AI improvements could be minimized at lower difficulty levels-- for instance, I think it's appropriate for AI to deny players diplomatic victories at lower difficulty levels, but never at Endless difficulty. Preferably, AI "intelligence" could be separated from AI bonuses.

Agreed. If it´s possible to make behavior and bonuses two different aspects of the AI, it would greatly enrich the game. It would make sense that lower level AI makes more naive mistakes, even most things mentioned in the thread. I think the ideal would be that top level AI made absolutely none of them and actually had no/little bonuses because it didn´t need them.







I´d also like to perhaps add to the point I tried to make with the example of my cult game.

I´m not sure if "Have a better estimation of its own military power and enemies" was about that - but while that´s a good deal of what I meant, ultimately what I thought to be the issue was that the AI seemed to not understand/not value my friendly actions towards it.



Obviously, the relation between player´s/AI´s armies has to be a factor when it´s measuring whether the player´s actions even matter to it or not. But once estabilished the player´s military superiority, it should be able to then collect information about battles involving me and its enemies that happen within its sight to use this as a diplomatic modifier. The game already awards influence for battles in allied territory, for example. But it should be able to set the ground for alliances during the first half of the game while people are just at peace (or not) with each other.



It should be able to recognize a trend in its diplomatic relations with a player in a way that a long peace matters more than a recently agreed one. This would even help balance Drakken´s spam of "War - force truce - war - force truce - force peace - war - etc...". After a stable period of peace, a player´s army should be able to even station in lands at peace without that counting negatively for diplomacy.



The AI should be able to adopt a free-rider diplomatic stance. Pretty much like 95% of the nations in the world.

Wow!



I am loving this thread!



Thank you, Developers for asking, and, I am loving the responses.



Marc

The big one for me would be AI aiming for certain victories depending on the factions they are playing, according to the factions design.



It would be nice to meet the drakken and know they are going to be diplomatic, and meet the necrophages and feel a chill down my spine as i realise its going to be a bumpy and war filled time. At the moment the AI doesn't seem to focus on its factions 'role', the drakken going on a war spree and the necrophages sitting in a corner not doing much. Breaks the atmosphere which has been so beauifully created with the fantastic artwork and lore, which is a pity.



EDIT:

I notice somone mentioning the AI getting random personalities at the beginning of the game such as 'peaceful' or 'warlike'. If this is the case the option to set the personalities before entering the game could be good. That way people who like the random personalities can keep it while people who want the AI playing to their faction roles can set drakken too 'peaceful' and necrophages to 'warlike'. For people who like to play with all random enemy civs perhaps an option to lock the AI to chosen faction role, so automatically setting drakken to 'peaceful' and necrophages to 'warlike' if they both random into the game. Again by having a tickbox for it in the game menu before you start allows the people who like all random to keep it while those who want AI factions to play to their strengths can have that.

The AI absolutely needs to be cognizant of its favorable win conditions not just because it would make them more interesting to compete against, but because the AI's actions in single player are training for multiplayer.



There is a huge missed opportunity in not giving the AI specific diplomacy dialogue. The threat/compliment dialogue is an excellent opportunity to diegetically include tips and hints for the player about how that class plays the game. Make the player aware that they are competing on multiple battlefields.

Hello everyone. I have learned, that AI does not use privateers (kaper, corsaire, каперов). Any fraction, never. Even if to buy privateers and to return management AI - that AI attacks and kills own privateers.

[spoiler] [/spoiler]

That´s so weird.



I have been attacked by privateers before. 6-unit armies, usually a lot of Marines, nothing to do with any quest of mine, I could only assume it was the AI.

AI settler behaviour:



1) I saw a settler just off the land looking like it was heading towards a region I had just colonised. The AI left it there. Is this right? To me it looks like it didn't kow what to do with it.

Is this a bug with the sea? Or do AI not have a routine to pull settlers back before using them again, if a targeted region is colonised?



2) If the AI have a settler in a region that you are moving an army in (currently at cold war, beginning of game), they will still settle a city even if you can then take it very easy. A) this provides an easy new city but B) It makes it hard to defend uncolonised regions if you are producing a settler to go there. You do get a city, might not be where you want it though, but you go to war.



Is this intended gameplay? It seems like it could be, but is lacking. If these were human players, they would be going 'mine, get away' or 'ok you can have it' etc before the region is settled. I'm not sure if a warning/compliment suffice in this regard.





Aside : something that does seem good to me is the AI tend to settle regions with the strategic/luxury resources first, and leave ones without them. Not sure if I'm imagining this, but this seems like good AI behaviour!

I've got this same issue with the deadlocked settler... It just standed at my border, not knowing what to do. I've filed this as a bug.

I will also mention the lack of AI privateers. This seems to really screw with the roving clans AI in particular. In my current game i have share vision with everyone. The roving clans is behind and obviously wants to attack somone to get forward in the game. It has built a massive army that is just standing next to one of its cities but since they are unable to declare war on anyone. Them being able to use privateers would allow them to attack somone.

@Zenphys: good point! Maybe the roving clans should have only a sort of protective army, and then use the market to buy mercenaries ... But maybe the AI is building all this units because they have *nothing else to build*... So maybe the AI should start making a "to sell" unit... to convert the production into some money, and change the governor's priority to make gold instead of production

I really appreciate this thread, it's good to see that Amplitude cares about our feedback, it's going to make Endless Legend an even better game!



There are already so many good answers here, and some of these Manu already added to the list, but I'll just weigh in with my top concerns for the record.



1. The AI needs to make some sensible assessments in choosing research. They often neglect essential tech, like Empire mint, and they need to prioritize resource extractors when they have potential resources to be extracted.

2. The AI needs to be more consistent when it comes to diplomacy - often times they'll accept an alliance and then break in ASAP for no reason. And at least some AI-personalities should be more compelled to estabilsh trading networks.

3. The AI needs to react to enemy forces and move their own forces strategically to defend their cities.

4. And this is a big, multifaceted issue: The AI needs to be able to play on a satisfying level as all the factions. Right now, playing against random factions is practically broken because the AI is totally useless as some of them.

I think Yuri's list has been covered elsewhere in this thread with the possible exception of number 3.



But I also want to stress how vital number 4 is, because it really seems to be the deciding factor of effectiveness for several of the AI factions. The Necropahges, the Roving Clans, the Cultists and the Drakken are especially needful of AI routines that lend to their strengths.

I didn't read all the ideas from this thread, but here are some points i want to highlight, i hope that post is not too redundant with the discussion.



1. AI diplomacy is too rigid. I mean if you had been in an alliance with an empire for 100 turns, they should'nt mind if a small army walk through their territories searching the ruins.

a/ I find the common border malus not filling its purpose. A friendly empire that have trade routes should not be considered a threat because they have a common frontier. Plus some amount of common frontier is caused by the AI expansion.

b/ Let AI understand aggressive colonisation that is aimed at smothering their expansion. Maybe let the AI plan for some amount of nearby desired zones and adjust diplomacy scores if someone else take the wanted spots.

c/ Converted villages should not be targeted by AI's armies under alliance terms (maybe peace if the AI think highly of the cultists?). Maybe allow AIs to take back villages in their territories at higher difficulties?

d/ Converting a village in an AI territory should be considered an hostile behavior by that AI.

e/ An army from a peaceful or allied empire should be able to visit the territory. A cold war one should be considered threatening.



2. More insight in what technologies are important, and come back to previous eras technologies in case of need. I like the fact that the technologies are not set in the marble but here are a few points that should be focused on:

a/The Strategic ressources extraction should be a primary concern if the AI lean toward developping armies or want to build advanced buildings. But if the AI is focused on other things (or has little need of the ressources and can a little from the market or exchanges) it can be avoided.

b/The Imperial Highway is a central technology for every empire. The movement part is really key and seeing AI without it is always painful for their development, defense and economy. It doesn't have to be a high priority, but most of the times it should be important to get it before era IV.

c/If the AI want to achieve a scientific victory, buying technologies with others empires or researching low eras technologies should be considered.



3. Combat AI. Lots has already been talked about, but i want to add some things, most of that would suit the AI opponent but with minimal GUI could be had to the player weapons as well (when auto fighting):

a/i would really appreciate if the Auto-combat AI would understand better the role of each unit in an army. Maybe allow players to set a role to each unit like tank, range damage dealer that want to be far from ennemies as possible, support that want to primary target ennemies, support that want to be near the frontline and so on? If that can't be well understood by the AI.

b/set some subgoals to the fight

-> lose as few units as possible. That include sub strategies as healing more damaged units, retreating very wounded units from the frontline and so on.

-> Kill the maximum of units without caring about own safety

-> Focus/not focus supports/ranges

c/ A more accurate pre-combat understanding of who has the advantage. I find the current information wrong a little too often (it can be result oriented though, me noticing every time the game told me i had the advantage and losing the whole army with few dead units on opponent's side)



I want to thanks all of you Amplitude people willing to improve your already great product, and be willing to take everybody two cents! AI is indeed a part of the game that can be improved greatly but need lots of work to do well. I like better to fight smarter opponent than an opponent with twice the production or research advantages, and even in multiplayer an AI that has less flaws allow more variety and enjoyability to the overall experience.