In tactical games the AI continues to underperform by a huge margin (webbing units already webbed, etc.).  In the strategic game in all the many hours I've played I've only seen strategic spells cast on me 3 times (and yes I understand we don't get obvious notifications but I pay attention to this) -- and this is on ridiculous and insane difficulties.  The AI should be tornadoing/tremoring/freezing my armies just about every turn, I've never had that happen to me.  All the lovely spells in the game and many are not used by the AI...

I have to play at insane to be challenged, and the only difference between insane and easier difficulties is that enemy forces are more/bigger -- not smarter.

Given we have no MP, AI is all there is and it needs major improvement.  Civ4 doesn't have great AI but it's several orders of magnitude better.

LH/FE/WoM are good games with great potential that is currently unfulfilled.  You have talented people working on them.  My post is intended not as a negative against what your team has accomplished but instead a compliment that they are a great team and LH can and should be -- given the talent of the team -- better.  That's the problem with having talented folks -- we expect a lot from them!  In many ways it's easier to be average than exceptional lol.

+1

Yes, I agree that the AI can always use improvement.  On a number of levels.  It isn't bad, but there are a lot of missed opportunities, which I think need to be pointed out as we find them, preferably in a very descriptive manner if you don't mind sharing.  This is of course assuming Brad will note the 'missed opportunities' and take a crack at increasing use of spells, etc..  Also, sharing how we regularly 'beat/outsmart' the AI units is probably a good idea.

AI is by far the most important feature in games (it provides the challenge, based on the parameters available), and getting it right takes a lot of patience and trial and error, especially in a game with as many moving parts as this one.  This is why I was happy when Brad took a back seat on design to focus on AI work.  That being said, Brad is also, I'm sure, fairly busy doing CEO type stuff, so it might not hurt to have another dedicated AI designer on the Stardock team that can share the workload with Brad, assuming E:LH sales justify a fourth go round.  Plus, a different perspective might lead to new approaches to some of these issues, and crack open/resolve a few problem areas in the process.

Re AI, if people are beating the AI beyond challenging, then we need them to explain what basic things they're doing to beat it.

They're not beating the AI because spiders sometimes web an already webbed unit. There is some more fundamental strategy being employed that I'm not aware of.

I can only make the AI play an optimal version of how I play the game.  I have no doubt there are plenty of people who are better at the game than I am.  Therefore, I need those players to explain, in general terms, what they are doing.  

I have only played the campaign on the current build, so if you've made changes since .87 I wouldn't be able to comment.

But if I may be so bold,

what I see the AI doing is settling sub-optimal spots (e.g. settling 3/2 over 4/3/2) and failing to clear monster lairs near its core.  The AI should not be settling 3/2 or 3/1 spots.  The only reason to do that is if it's scouted the entire map and found nothing else.  And often it appears that the AI is picking the spot located besides the best spots.

Obviously, the AI should settle the best spot possible.  I am not sure why it doesn't do that.  Maybe there's something wrong with the map it's using?  Shrug.

Not clearing monster lairs is another significant problem for the AI.  And if I were to venture a guess, I'd bet it's primarily due to the way you place monsters when generating the map.  IIRC, you favor the player by moving the higher level monsters away from their starting point.

That means that the human player can grow his first city without fear and has more flexibility on where to expand and can usually avoid high level monster lairs near his first few cities (e.g. until the midgame).  The AI doesn't get that benefit so ends up with high level monster lairs near its starting and second cities that it cannot clear and so is constantly harassed by roaming monsters.  Even if they're not attacking the AI's cities, they will end up destroying improvements and killing units by moving at random.

If that is the case, then I would suggest removing the monster positioning bias on difficulty levels higher than Challenging.

The mistakes you point out the AI makes there might make a difference in a challenging game. But in a hard or later, that wouln't be the difference.

There must be fundamental differences in play style going on.  I mean, after all, if what you say were the key thing, I could just bonus the AI with more resources and call it a day.  Clearly, there's something else going on.

I can't beat the AI behind hard. I know from the stats that very very few people can beat it above normal (not even challenging).

Derek and I have discussed this and what we usually see players doing is creating uber-custom sovereigns and uber-custom factions and then Ctrl-Ning until they get the perfect start-up spot.  Which is fine, it's a single player game. But those victories aren't due to "bad AI" that's due to the players giving themselves a huge advantage.

i think it's more about very powerful (overpowered?) unit designs. it's not terribly hard to design a unit with insane initiative and lots of armor ignoring bonus damage (just ignore armor beyond leather, use init/fire/frost jewelry and find a spot with 3-4 essences and enchant it with heart of fire & aura of grace); then add on top of that the init boost from fortress level 3 and the "fast" trait and the bonus damage from finesse/enmity/bloodthirsty (you get enmity/bloodthirsty for free from the warrior temple). those guys often get two turns for each turn the AI gets (considering that high level AI love to slow down their units with heavy armor) and since much of their damage ignores armor (fire/frost; also armor penetration from spears) the enemy units don't really get much of a chance to fight back before they are severely crippled or dead. and then you can also add some powerful traits from a custom faction that boost the effect even more (like quick & enchanters) at that point, it doesn't really matter what units the AI fields and what it tries to do wit them - many (most?) of them are dead before they can do anything at all

@Frogboy MadDjinn is a well known strategy beast (he's streaming live various strategy games and collaborated with develeopers to help improvements / bug fixes etc) and he plays only on ridiculous diff. I'll give him a poke to contact you (he's very sharp and notices AI mistakes strategy wise, short-long term etc).

I think it has more to do with AI priorities than with it's building path. For example am playing  game on expert world / AI, huge, everything normal only quests plenty, 10 players. 2 custom factions, one similar to Resoln - The Scryers but with enchanters instead of spiders (my faction) and a kingdom one The Seekers - focused on no penalty for traveling, chainmail, archers and assassin tools (amarians as race). What happened in mid game when I met everyone was something like this: Pariden and myself were bottom of faction power, with about 100 score. The Seekers were top of list with 400+ FP, followed by the normal Kulan ~300 (I have yet to play a game post 0.9 where he doesn't make it to top 3), Yithriil, Altar, Capitar with 200-300 each. I'm somewhat in the middle so everyone DoWs me at a point, and I'm pretty much just spending time running from one corner of my realm to other to defend / take cities back, wasn't too lucky with expansions until AI's locked me in only 1 stable and 1 warg den, 3 crystal mines and 3 metal, it doesn't help that tech wise I'm up there keeping up, even with trading I can't make more than 1 good unit for like 30 seasons, or upgrade 1 piece of gear to other units (126 crystal cost for a decent archer unit or defender - have to include fire cloaks or am toasted as I have no air heroes and everyone is making companies of mages left and right). Anyway I can't even imagine how much they expanded and cleared the map, but I bet at least 80% of map is covered by AI territory. They are all fighting eachother or have alliances etc, their troops design is quite good (can't really find anything I wouldn't do myself if I had the resources / production), so is their mix up etc. While they do lack in tactical fights regarding some spells usage (why would you cast earth grab when all my units are mounted hence immune to prone?), can't say they didn't make good use of buffs / heals / debuffs or damage spells (was quite surprised when I fought kulan 4 man army, I said umm okay so you send 4 heroes to fight my army? then he goes: haste on kulan + diamond skin + guardian wind and other air dodge spell vs ranged, growth etc you know the drill, was lucky Asha's infection landed on first cast to spread the blinds & slow).

I am holding my own for now and even stealing a city here and there from close neighbours (on top of that I got the event that spawns the spider queen in middle of my small empire, 4 epic armies with 2-2.5k hp when my freeze was on cooldown, beat them but this was really tense - I admit my sovereign was level 15 with -1 turn on spell casting and master death & water - as I was already fighting Altar and trying to delay Magnar on other corner). But I have no clue how this will pan out. I have a hard time trading for crystals (they either have metal and don't want mine, my money are not enough to afford to pay 1k for 50 crystals etc) which severely limits me in building decent troops that I can leave to defend the kingdom if I want to venture further. 

So there is no problem - in my game - with AI expanding - it actually does it at an insane pace, even in early game his heroes had double my level and gear I could only dream of. They don't struggle economic / territory / tech wise (unless they are at war with couple other AI's and start losing stuff fast), they build good armies, and use spells (I have magnar spamming every couple turns oppression on my cities for example, Kulan used several strategic damaging spells when I tried to enter his territory to break some stuff etc).

Maybe on smaller maps / different settings things are different.

Well that was my best guess.  I think with monsters wandering around, even if they're not targeting the AI, they probably still destroy stuff a lot of stuff just at random.

I think about this subject a lot.  But I can't tell exactly what's wrong with the AI.  Maybe they're hamstringing each other?  Though honestly I see the same behavior in 1v1.

do those stats count people who stop playing when the game becomes boring because the monsters are dead and the AI is not competitive?

I don't think I am cheesing to win (though I do not play on Ridiculous -- just Hard/Hard).  I use the standard sovs (though I did play a game with the sov that can recruit beasts and that was ridiculous).   I don't abuse the Ctrl+N.  I don't think my strategy is anything to write home about.  I try to recruit two additional army units to go with my sov & new champ.  I harvest low level critters and do quests for XP.  I scout to find viable 2nd and 3rd expansion locations and clear out the monsters in those areas while my pioneer is moving there.

Unless there's a gorgeous spot (5/2/2 or something), I tend to stop expansion at 4 cities.  And then I go questing.  Still, I don't think about the AI at all because they're not a threat.  Maybe that's the problem?  The AI is not pressing good players because good players have high power rankings.  So good players can choose their opponent and strike when its convenient.

Perhaps the AI shouldn't pick on weaker players.  When dealing with the human player it should be sympathetic to lower power rankings and scared of higher ones.  And it should declare war on players it sees trending upwards too quickly.

The monsters destroying improvements / attacking cities is a bit random (it feels to me, if they don't have a unit to chase, they just wander, if the rng path generated for next move is overlapping / intersecting an improvement or town in their way, they will attack). I had a fel dragon army (yeah he joined up with 2 spiders) sitting inside a village borders wandering back and forth in Altar land (he was next to a city of mine) for over 30 seasons and never attacked it. And his lair (dragon nest resource or w/e is called) was inside his lands which he disturbed when expanded there. When I settled near an ogre lair, 2 turns later my city was gone. So it's rng when roaming monsters will attack players (AI's included) improvements / outposts / cities due to their movement behaviour.

I'm terrible at the game, but I have won on challenging, without any custom factions/sov.

I actually think it is the tactics and unit design.

On my "challenging" win, I decisively won two wars where, according to the end-game graphs as well as my civ score, my army was much weaker than that of my opponent.  One of those, I did have some help by when I timed my attack (I attacked a faction when he was already busy with another war), but the second of those was a massacre against a Yithril army about twice as strong as mine on paper, with no help, while distracted with another war against Tarth. 

I have never seen an AI build mage units (which are absolutely amazing), he doesn't take maximum advantage of swarm bonuses, and- perhaps most importantly- he very rarely casts things like strategic wither, or frozen bones.  Also, his army tends to be much more "split up" than mine; I'm generally able to dismantle it in a whole crapload of relatively small pieces, whereas mine tends to be 2-3 powerful forces.  Whether it's because he isn't being aggressive enough about researching max army size, or whether it's because his troops are geared too much towards quantity over quality, I can't tell.

I like this approach, but at the moment LH doesn't seem to be working like that, which my current game exemplifies. I'm currently playing a game on hard as Gilden. Yithril started a war against me when power ratings were as follows:

Yithril 227

Me 109

Resoln 65 

Altar 56

Pariden 32

One thing is that other civs don't seem to be doing very well. They've all settled only 3 cities (I've settled 6 and Yithril 8).

Anyway, with that sort of difference in power Yithril should roll over me. He will not, due to his complete military incompetence. I have only one strong stack that really counts. He first assaulted a city defended by it with an only slightly weaker stack of his, which I duly defeated.

Then he arrived in my borders with 3 strong stacks travelling together on one hex. I had to attack them and defeated the first stack without losing any troop (I used an artifact whose name I can't remember to take over one of his juggernauts, which was crucial). However, my troops took some damage and if he attacked me in his turn with remaining two stacks, then maybe I could defeat the first one, but there is no way I could defeat both. Then the war would be probably over. Instead, he decided to wander around aimlessly, letting me recover hit points and choose a convenient moment for the next battle.

In the meantime, he sent another stack into my territory, which is heading for my capital, but which I will defeat with no problem, because his other two stacks are not putting me under any pressure. At the moment he's got 180 power against my 145 and I have no doubt I will win the war. 

So, I will defeat an opponent of twice my power ranking, more settlements and at least five strong stacks with my single one, because he absolutely no clue on how to use his several armies together. I hope you can improve on this kind of AI behaviour and if I can help you in anyway, just ask.

I think the lack of coordinated army attacks is the problem.  Human players can handle one maybe two stacks with only one army but if the AI brings 3 or more then they need to have multiple stacks to defend unless they turtle within their cities which invites the ability to disrupt infrastructure.  If multiple army stacks can coordinate attacks this will greatly improve AI efficiency instead of one stack at a time marching to its demise.

I spent the last hour writing a detailed response, and the server apparently ate it.  Not sure I have it in me to recreate the post...  Maybe later I'll try.

Edit -- great, this short content-less post goes thru, the detailed lengthy one didn't.  lol at me

I agree about utilizing armies on the strategic map. I was fighting 15 armies sent to attack me, I had 5 or 6 armies. First, 13 of those armies were stacked on the same tile so I just used that fire spell which attacks ALL 13 stacks and hurts them enough. I then whittle away at them. I use my 5 or 6 stacks to attack them and win each battle. They now have 9 or 10 stacks left. They should have all focused on maybe attacking 2 or 3 of my armies so that they could defeat them and then I would be left with only 2 or 3 armies and would probably lose although I would have almost wiped them out regardless.

Also, they just aren't defeating me in tactical battles, I can crush them with even or even slightly lower power ratings every single time and my army is still capable of fighting another battle without having to recover after that. If the power rating is significant I will lose.

I don't think it should be possible for 15 armies to lose to 5, I guess it is possible but it should be very difficult, and it isn't.

I am not saying it isn't fun though. I had lots of fun fighting those battles. Oh, and I am playing on challenging.

Frogboy, you don't honestly think you can develop an AI that can beat serious players straight up, do you? Most people play the game casually, but they also didn't know that you are collecting data on their games. Winning on challenging is as simple as developing a mage sovereign to cast fireball in 1 turn, building a mana reserve and some troop to buffer the sov, and using it to obliterate opposing troops. Almost any coherent, focused strategy will consistently beat the AI, which seems to pick traits in a breadth-first traversal of every tree. It is not that the AI just makes double-webbing mistakes, it doesn't do most things in a coherent, focused manner. The only advantage the AI has is that most players aren't trying to win the game, they are thinking about leveling up, questing, and taking down wildlands, instead of cutting the AI's throat at the most opportune moment.

I personally believe that it is not possible. I wrote a thread (and replies) about it here: https://forums.elementalgame.com/438446/page/1/#3300617 . If people are playing this game to win, either the game can be reined in until it is simple and uninteresting and the problem may be tractable or the AI can cheat and have the illusion of player-like play, but in a fun, commercial strategy game the human mind has too many advantages dealing with the crazy complexity of the game for a fair AI to compete (or even a fair AI with reasonable bonuses).

Maybe you meant the vast majority of players who play the game, because most everyone is not serious. That is doable, and you have done it well. But I don't think you could seriously hope to catch the people playing at ridiculuous and above, they will just find the new best strategies and adapt them in real time. Unless you develop a learning solution, your AI must play your moves that you intend (and actually work in the general way they are intended) ahead of time while you were blindfolded to the actual gameboard. The human mind will win that every time in a game this complex, in my opinion (with the caveats I mention in the thread).