Always thankful for all the atention.

wilbefast wrote:
What the AI should do: re-evaluate the value of each dust unit as it is added to the diplomatic agreement. So if I have 2000 dust the first 1000 is worth less than the remaining 1000.

This is cool and important. Is it possible maybe to couple this with a better acceptance for higher volumes of trade, to benefit large contracts somehow? Right now, if you wish to spend loads of influence, you can just make several different very small transactions. I don´t know if this is intended, or if there was supposed to be some balance there already in which case I hardly notice it. I feel it can be abused as a cheap way out of an imminent (and deserved) defeat.





On an unrelated note - why is it that sometimes, only sometimes, the AI will keep attacking a fully fortified, fully garrisoned city several times consecutively? I´m not saying it never sieges, it´s just sometimes it doesn´t and these times it seems to keep repeating the same mistake over and over. Maybe it judges its armies far superior based only on HP and figure the forfications won´t matter (or maybe more precisely - is not able to count them in the math at all)? This would be a terrible oversight, for a Military System that considers HP so highly, the AI should always consider the amount of fortifications exactly like HP.



Save file Necrofage Turn 54 - This is the only thing close to an example I have, it´s the game from the thread I was trying. Blue Vaulters would declare war on me on turn 58 and perform the first stupid attack on turn 60. IIRC it would go on to attack the exact same way another two times, except the last time it brought a hero. On their fourth attempt they started to siege and attacked as soon as I brought a single unit out (can´t really blame it here). Perhaps it started to perform the natural routine after repeating the same mistake three times and losing, what, one third of their armies. Can´t promise it will behave the same again, of course. It´s on Dropbox.

wilbefast wrote:
The AI doesn't follow a predetermined path when it comes to research, though given the same conditions (as evaluated by its heuristics) it will make similar choices, and certain choices may be overvalued. If we push the AI to be more balanced and economy-focused it may be that it doesn't provide an interesting challenge early-on and is simply obliterated. Balance is hard to achieve. If there are any technologies or improvements you feel the AI is choosing too often in most/all circumstances please let us know!

I think focusing on improving the preferred tech tree path for each faction would greatly improve the intelligence part of "AI". For instance, for every single era, technologies that add to the dust yield (Aquapulvics, Empire Mint, +Dust, etc.) should be the default-first/always-researched techs for the Broken Lords. Likewise, diplomatic (Parley, Peace, Alliance, +Inf) technologies should be the default-always-first researched techs of the Drakken.



Each faction should have a default research tree/path that it pursues based on its strengths. Any human player knows that certain techs are essential, some are optional and others are fluff based on the faction's abilities and strengths. There needs to be a separate algorithm for each faction with each assigning a value to every tech as 1) essential, 2) helpful or 3) nice to have (fluff).



For all factions, things like resource/luxury extraction should be essential, for factions that focus on certain resources, any tech that increases the output of that resource should be flagged as essential.



When playing the game and encountering a Broken Lord AI-controlled faction that is in era IV but hasn't researched Empire Mint, I am ejected from my immersion in the gameworld and left thinking "WTF".



TL;DR: Each faction needs a default research tree with priority values assigned to each tech.

Enlil wrote:
I think focusing on improving the preferred tech tree path for each faction would greatly improve the intelligence part of "AI". For instance, for every single era, technologies that add to the dust yield (Aquapulvics, Empire Mint, +Dust, etc.) should be the default-first/always-researched techs for the Broken Lords. Likewise, diplomatic (Parley, Peace, Alliance, +Inf) technologies should be the default-always-first researched techs of the Drakken.



Each faction should have a default research tree/path that it pursues based on its strengths. Any human player knows that certain techs are essential, some are optional and others are fluff based on the faction's abilities and strengths. There needs to be a separate algorithm for each faction with each assigning a value to every tech as 1) essential, 2) helpful or 3) nice to have (fluff).



For all factions, things like resource/luxury extraction should be essential, for factions that focus on certain resources, any tech that increases the output of that resource should be flagged as essential.



When playing the game and encountering a Broken Lord AI-controlled faction that is in era IV but hasn't researched Empire Mint, I am ejected from my immersion in the gameworld and left thinking "WTF".



TL;DR: Each faction needs a default research tree with priority values assigned to each tech.

You´re right about the low level techs sure. And Dust for the Broken Lords is a special case because it´s also food, so they really shouldn´t skip any form of generating that.



I don´t know, however, if Drakken is supposed to always win through influence techs and buildings, for example. Through diplomacy, sure, but unless they´re clearly aiming for Diplomatic victory, I don´t really se them going for a National Museum 100% times. They can generate just enough influence through their heros (which reads: about the same as mostly everyone else trying an infl. oriented game) to invest on scientific, military or economic victory, and I think there should be room for the enviroment to influence them into those directions.



I feel it would become a little too predictable and dull if everytime you met a Drakken you already knew pretty much what kind of game it would try.



Same goes for Vaulters and science and Wild Walkers and production. I think it benefits the game that there´s room for them to either capitalize on their strenghts when the surroundings are in favour, or use their extra yielding to make up on a different type of game that might be more accessible than their original affinity.



As a side note, I´ve comfortably won games I never researched "basic" techs. In the Endless ardent mages game from my thread, I believe I never got Seed Storage even though that´s more often than not an essential one. The board, imo, counts more than a Faction´s traits.

I agree totally with what you suggest: a more dynamic AI, but I think that's too much of a jump to expect right now... If they can make the AI enhanced enough to behave intelligently based on each map, then I am all for deviations in the pattern/path. However, I am pretty sure that the basics (capitalizing on inherent strengths) has not yet been mastered.



Therefore, what you suggest is a more advanced AI, that should be a goal to reach for, but until then, I would prefer an AI that focuses on its strengths rather than one that focuses on a random, un-intelligent path that is identical for all factions.

Enlil wrote:
Each faction should have a default research tree/path that it pursues based on its strengths. Any human player knows that certain techs are essential, some are optional and others are fluff based on the faction's abilities and strengths. There needs to be a separate algorithm for each faction with each assigning a value to every tech as 1) essential, 2) helpful or 3) nice to have (fluff).

This, in addition to the AI actually using the factions' skills based on their inherent strengths. To me it seems that most, if not all AI are running on some sort of generic personality and playstyle script which leads to them getting rumper ravaged if not for high difficulty having them cheat resources out of nothingness. The Drakken and Wild Walker AI always end up topping the scoreboard in my games, probably because most of their unique bonuses are rather "generic" and easy to utilize while something like the Cult with a non-standard playstyle always ends up getting steamrolled by every other faction. And even if Wild Walkers or the Drakken end up performing well relative to other AI factions their playstyle still leaves a lot to be desired.



F.ex.

-I've never seen Wild Walkers attempt to build the victory wonder despite their faction being geared towards it. I've seen plenty of games where they've started the victory quest and at that point they should also be able to build the wonder, but they never do so and are content to just chill around or attack their nearest competitor in an attempt to expand a little.

-I've never seen the Drakken use the diplomatic skills, the very basis which their whole gameplay hangs on. Most of the time they spend the game in cold war or war with other players, they never force peace, never force alliances and they never force a truce even if they have the influence to do so and the AI is acknowledging they're losing all their cities to another faction they're at war with. Every AI gives a message when they feel like they're losing and might even ask for peace, but the Drakken capable of forcing it to happen never do so.

-The Cult never expands outside few adjacent regions' minor faction villages. Instead of using their free minor faction units to wage war or scout for more villages, again the basis which their whole gameplay rests upon, they rely on armies composed solely of their starting supports or cavalry and even those don't go out to explore for more villages to convert. There has never been a game when the Cult AI hasn't lagged behind everyone else. The AI is also really bad at defending their cities and it's even worse with the Cult. If I attack an army that's hanging near their city, but not inside it they'll always use all their city garrison as reinforcements instead of relying on their 1000+ fortification bonus to stall out, as garrisons used as reinforcements lose their garrisoning HP bonus. As such they end up with a city with 1000+ fortification but zero defenders, resulting in an instant victory for the invader when anyone sneezes at its general direction.



Also a complaint in general, but the AI really doesn't seem to play to win. As far as I know every player is notified when someone is attempting or nearing a victory condition, but the AI doesn't seem to care. No attempts whatsoever trying to stop a snowballing faction from reaching scientific, economic, diplomatic, wonder or quest victories. It'd make much more sense if the AI was able to set aside their petty differences when a runaway player is nearing a victory condition. They don't act either because they still consider a snowballing player their friend for some reason or they're too scared to attack alone, but too stupid to form alliances. There really needs to be a Civ V style "we both should declare war on X" diplomatic option for players and AI alike in order to encourage more co-operation against the winning factions. Or just have the AI use the regular trade screen and ask other factions to attack their enemies. Convincing the AI to attack each others is extremely easy as a player, but they never seem to propose that type of deal to players or other friendly AIs.

wilbefast wrote:


A specific question for the community while I'm here: have you noticed any player actions which you feel are missing feedback? That is the AI "should" have said something in reaction to an action I just performed but didn't. Bearing in mind of course that there's a kind of "anti-spam" measure applied to stop players from receiving too many messages per turn, so "minor slights" might not always be noted if there are more important messages to display on the same turn.

W.

I dont know if this has been mentioned before, but I just noticed this myself.

I was playing a game and, early on, I discovered the cultists nearby. I took it upon myself to kill all the villages surrounding their territory before they could convert them. They were totally fine with it, despite the inconvenience I caused them. Understandable if they didn't know who killed off the villages, but I'm fairly certain they had sight on one or two of them.

hera35 wrote:
the AI really doesn't seem to play to win.

Manu made note of this in his original list but I wasn't able to find where he was quoting from - thanks for bringing it up again so I can more officially add this to the list. Philosophical question though: should the AI know that it's in a game with specific victory conditions? By this I mean: is it "realistic" if an AI sacrifices its military or economic integrity because it knows that the end-game is coming up?

hera35 wrote:
To me it seems that most, if not all AI are running on some sort of generic personality and playstyle script

Enlil wrote:
TL;DR: Each faction needs a default research tree with priority values assigned to each tech.

BPrado wrote:
You´re right about the low level techs sure. And Dust for the Broken Lords is a special case because it´s also food, so they really shouldn´t skip any form of generating that (...) [but] I feel it would become a little too predictable and dull if every time you met a Drakken you already knew pretty much what kind of game it would try.

Changing the "personality" of AI for each faction is something I'm working on at the moment, though the AI for Shadows does currently have the priority.



Here's a very quick and dirty draft so you guys have some idea of what sorts of things we're thinking about:



Wild Walkers

• Prefer food, industry and approval technologies and improvements
• Prioritise expansion by colonisation
• Especially interested in buildings wonders
  
  

Broken Lords

• Very much prefer dust technologies and improvements
• Hold grudges and remember friends
  
  

Necrophage

• Prefer military technologies and improvements
• Desire war, almost no "war fatigue" (desire for peace after long war)
• Prioritise building and maintaining armies for expansion by conquest
  
  

Ardent Mages

• Less "war fatigue" (masochistic)
• Prioritise research
  
  

Vaulters

• Prefer defensive and strategic resource technologies and improvements
• Hate enemies' trespassing
• Prioritise research
• Like to explore
  
  

Roving Clans

• Prefer diplomacy and trade technologies and improvements
• Less worried by enemies' trespassing
• Very much desire trade diplomatic relations
• Like to use privateers
  
  

Drakken

• Very much prefer diplomatic technologies
• Desire peaceful diplomatic relations
• Like to explore
  
  

Cultists

• Initially friendly, very quickly unfriendly if attacked
• Less "war fatigue" (robots)
  
  

Again this is a very quick draft - we're still working on putting the system in place to support such drastic variations in personality and, again, Shadows does have the priority. Still if you folks have any specific suggestions with regards to AI faction personalities it's a good time and place to let us know

Those are nice personality traits, I especially like the nuances for war fatigue, makes a lot of sense.

wilbefast wrote:
Philosophical question though: should the AI know that it's in a game with specific victory conditions? By this I mean: is it "realistic" if an AI sacrifices its military or economic integrity because it knows that the end-game is coming up?

Yes, it is. It´s what any player would do.

Perhaps the issue would be more like: is it better for the game? Because while setting everyone to go berserk the moment a player is close to a Military victory might be a nice touch, because the scenario implies the player can handle multiple fronts of aggression and the AI bonuses have "to go through" the players, I´m not sure about the other victories. I´ve seen AI Broken Lords escalate their Dust output absurdly, and apparently, they were just doing it for the kicks and not to prevent my imminent victory. I wonder if I would have been able to win any game whatsoever if all 5 AIs were simply out to stop me from doing so.

wilbefast wrote:




Here's a very quick and dirty draft so you guys have some idea of what sorts of things we're thinking about:



...

A few thoughts based on this list:

• At higher difficulty levels, Wildwalkers shouldn't prioritize approval techs unless needed. Generally the bonuses to higher level AI's keeps approval up easily.
• Due to the "rolling down the hill" affect, if an empire loses their main unit stacks it is easy to lose a number of cities. (Quality >> Quantity). Due to this, ALL factions should prioritize military techs and improvements.
• Ardent Mages should prioritize the use of spells.

Philosophical question though: should the AI know that it's in a game with specific victory conditions? By this I mean: is it "realistic" if an AI sacrifices its military or economic integrity because it knows that the end-game is coming up?

Yes and no. My thoughts here are that in military victories, everyone should be trying to stop the 'warmonger'. In non-military victories, your allies don't care if you win the game, while your enemies should care. So it is dependent on victory conditions.

• AI should do everything possible to stop a military victory as it approaches.
• Approaching Diplomatic/Economic/Wonder victories shouldn't make your allies stop being allies. But it should make your enemies more aggressive.
  
  

Maybe something like these conditions would help to balance out the fact that military advancement is one of the most common ways to any of the victory types. If everyone actually cares that you are taking cities right and left, it might change your playing style or make it more interesting at least.



^Just my opinion.

These developpements into the personnality traits are very interesting and look ambitious.



As interesting features, I would like :



- An ability to for a war alliance and make joint military operations against a very dominant player (especially if he has been a warmonger). Off course this ability feats more with Roving Clans or Drakkons than Necrophages.

Hard-coded AI personality is good. It needs a slider as to how much the AI RP that personality too.



Vaulters should have a preference for climate because of science generation, actually, each faction should have a preference like that.

Thank you for sharing! A few thoughts and comments:



General

• As Chaotic09 mentioned all factions should start the game with a strong , and Strategic resource base so that they can defend themselves and jump start their economy. This should happen before any faction specifc preferences kick in
• Survival concerns (invasion, nearby aggressive neighbor, protect Capitol at all cost) should trump any faction-specific tendencies
• The effect of "envy" should increase throughout the game
  
  

Wild Walkers

• Preferes to fight battles in Forest areas (Forest Rage is a big boost)
  
  

Vaulters

• Spends horde of resources on high-quality units and boosters
  
  

Roving Clans

• Rapidly negotiates Peace in any war that AI didn't start
• Moves diplomacy to Peace as much as possible (so trade routes work)
• Builds Right of Way everywhere
• Privateers really suck after Era 2. Not sure if this is a great strategy.
  
  

Drakken

• Uses Force Peace and Force Truce
  
  

Cultists

• Obsessively defensive of the holy city
• Converts a lot of villages
• Razes opponents' cities to get Stockpiles and not get out paced during the game
  
  

BPrado wrote:
Those are nice personality traits, I especially like the nuances for war fatigue, makes a lot of sense.





Yes, it is. It´s what any player would do.

Perhaps the issue would be more like: is it better for the game? Because while setting everyone to go berserk the moment a player is close to a Military victory might be a nice touch, because the scenario implies the player can handle multiple fronts of aggression and the AI bonuses have "to go through" the players, I´m not sure about the other victories. I´ve seen AI Broken Lords escalate their Dust output absurdly, and apparently, they were just doing it for the kicks and not to prevent my imminent victory. I wonder if I would have been able to win any game whatsoever if all 5 AIs were simply out to stop me from doing so.

The key here is WHEN they try to stop you collectively. Ideally, it should be later game when you are a force to be reckoned with (or on you way to becoming one).



In Age of Wonders 3, you can start your way toward a seal victory, and gradually, the other AIs will grow more hostile the closer you become, until they all declare war or try to compete for more seals/charges. I love that in that game because you literally see alliances created and dissolved, peace made and broken as all eyes turn toward you...and attacks will come to try to stop you. As a human player you have an advantage...your organic brain.



If the AI becomes more effective at countering you due to its increased awareness, it becomes less necessary to prop them up with stupidly high bonuses. Difficulty levels actually become more meaningful at this point, which is desirable.

Propbuddha wrote:

• As Chaotic09 mentioned all factions should start the game with a strong , and Strategic resource base so that they can defend themselves and jump start their economy. This should happen before any faction specifc preferences kick in
• Survival concerns (invasion, nearby aggressive neighbor) should trump any faction-specific tendencies
  
  

Roving Clans

• Privateers really suck after Era 2. Not sure if this is a great strategy.
  
  

True, and very true.



Security is the main concern of any civilization. There´s no point building a beautiful empire for someone else. In the Drakken´s and Broken Lord´s case, playing according to the faction means staying safe. For everyone else, military should come first.



I think that links to a "fanatic" use of privateers. I love being able to have them so early, I love finding a use for them; but this use is always very limited, and it is naturally larger the earlier you unlock it. If a Roving Clans AI is faced with a large space to settle, should it really worry about rushing privateers instead of empire building techs? And later when the space grows smaller, should it still think it´s a great idea in face of how weak they are comparatively? There´s a delicate balance there.



Of course, currently, it barely uses it, but perhaps it´s more of a flaw on the mechanics than on how much affinity the AI should have over the tech.



I´ve suggested in my thread that Privateers are, at some point, allowed to use a certain amount of Dust to reinforce each other. The cost should be pretty steep, comparable to buying an entire modest army perhaps, but I think it would fit into the dust generating game.

BPrado wrote:
If a Roving Clans AI is faced with a large space to settle, should it really worry about rushing privateers instead of empire building techs? And later when the space grows smaller, should it still think it´s a great idea in face of how weak they are comparatively? There´s a delicate balance there.

Exactly! Teaching the AI to use weak strategies isn't going to be helpful. Either buff the strategies (yes please) or have the Ai ignore them.



Mercenaries are a terrible proposition in EL. They are expensive and since you can't upgrade them, they aren't effective. When you can afford them mid-game they are completely outclassed by other faction's Strategic equipment-armed troops. The RC's Mercenary Comforts trait helps make they OK in the early game (Eras 1 and 2), but they are still not worth it. Privateers are even more expensive and also can't reinforce each other so they are even weaker.



Also the buyout mechanic makes the Unit Marketplace pointless as you can buy custom units at any time.



I always play on "Fast Speed" are Mercenary prices reduced in proportion to the discounts regular troops get?



The RC have a general problem (as a faction) that the things they are supposed to excel at (Mercenaries, Marketplace, Trade Routes) are all weak and/or non-critical components in the game

Propbuddha wrote:




The RC have a general problem (as a faction) that the things they are supposed to excel at (Mercenaries, Marketplace, Trade Routes) are all weak and/or non-critical components in the game

I agree that the RC are a great sounding race on paper but in actual execution are really sucky. Even apart from how difficult it is to make the AI use them properly, they still lack unique traits that have a reliable usefulness.

I think it is important that the cultists hate the neighbors, not only the direct neighbors, also the neighbors of the areas with converted villages.



In my experience playing with the cultists you have to hurry to eliminate the neighbors ( I usually play in giants maps), if they expand around you, they will eat you. The cultists should not play defensive.



It would be funny if they give priority to the guardians. And keep wandering powerful armies and their city was an impregnable fortress. But in really i think cultists are a lost cause, would have to improve a lot Legend Endless AI that they were a faction to be reckoned with.





PD: The cultists should also learn the value of stockpiles, should use them when they have, should not trade them (AI sell them to you easily) and should give priority to investigate technologies that improve stockpiles.

Slashman wrote:
I agree that the RC are a great sounding race on paper but in actual execution are really sucky. Even apart from how difficult it is to make the AI use them properly, they still lack unique traits that have a reliable usefulness.

I don't know about that. It's not too uncommon for RC to be able to afford a demon merc with *2 health (or whatever it is) within the first few turns. A unit like that is a one-man siege stopper, an ultra-fast scout, and rush potential, all rolled into one.

I don't play multiplayer, so I can't speak for any faction's balance or power, but the RC in single player are super fun to play with and quite powerful (against meh AI at least).



But it is true that the AI has no idea how to play RC. While it doesn't fare as badly as when it is Broken Lords or Cultists, the way it plays RC is lackluster and erratic at times. An RC AI should theoretically be at peace or ally with the strongest military factions and set up trade with everyone and aim for an economic victory while staying out of wars and disputes, with the occasional privateer raids (which it never does in my experience).



I do think it is necessary for Endless Legend to have different AI personalities, for each faction (alternatively, they could be set by the player, like in Galactic Civ). It's the sort of game that begs for it, considering how different the factions are both in lore and on paper. I am inclined to think that AIs playing each faction like "it's meant to be" would boost its performance.

I feel the AI should most definitely be 'Playing to win'.



This would include getting aggressive and ganging up on any player who is getting close to victory. One of the common gripes I see from people is complaining about a boring end game. Having AI playing to win I feel would go a long way towards fixing that.



At the moment if you are going for a peaceful victory you just get your economy bigger and hit end turn until the game is over.



Having relationships break down and feuds start as different players get closer to different victories will throw some life into the end game and make it an exciting 'race to the finish line' rather than just a click end turn fest.