Let's talk AI...

By on May 20, 2013 6:14:48 PM from Elemental Forums Elemental Forums

Nick-Danger

Join Date 09/2009
+28

I really like LH.  "Time Played" is 163:07!  I've spent the past 3 days on one game, and am only about halfway thru it.  Game is huge world, insane AI, 5 insane AI foes (all empires, I'm a custom kingdom).  Settings are mucho resources and magic, no wildlands, temperate.  Currently turn 264 and I'm (Dangerites is me) hanging in there...

The game is beautiful -- the art is lovingly crafted.  We can easily tell which element a shard is, and if it's developed.  We can see what weapon our Sov is wielding -- not just on the paperdoll but on their tactical representation.  We can zoom in and see the people doing their daily tasks.

Balance is constantly improving.  Spells are interesting.  Exploring is fun.  All in all a lot of good stuff, and a good game.

The one part lagging behind the rest is the AI.  It's not that the AI is terrible, it's that the rest of the game is so good.  There's no MP, so the AI is all we have.  And it's a strategy game.  And we've heard how the AI will make us cry in anguish it'll be that good.  I have suggestions to make that claim more a reality, all based upon my insane AI world and insane AI foes observations in many hours and multiple games, and with the clarification that I'm an average skill player (in Civ4 emperor difficulty is a challenge for me) so insane difficulty should be beyond my ken:

First tactical:

-target selection.  The AI needs to better select the most 'dangerous' target.  Suggestions to prioritize are ranged, Sov, champion, nearest melee (nearest in terms of # of moves to close to melee attack range).  Within each 'category' the unit capable of doing the most damage comes first (so amongst ranged units, an 'ice' mage would come before a 'fire' mage if the targets are immune to fire attacks).

This can be modified based upon the AI leader/race.  For example, a 'mage' AI sov might fear melee more than magic so melee would be prioritized.  Or maybe a 'mage' Sov would most hate the 'competition' so would prioritize magic foes.  This provides a means of making the AI sovs more unique.

-focus fire.  I've seen one good example of focus fire in my current insane AI game -- the fire skeletons at the end of the 'night of the skeletons' event focus fired at one of my units despite my best efforts to save him (moved him back, moved other units nearer the fire skeletons, etc.).  Every other battle it appears that the AI randomly and individually select their targets with no regard for finishing off wounded units.  In one battle of my 6 units vs. 6 shrills I should have been wiped out but instead I won with no deaths because the shrills appeared to do everything they could to not kill any of my units.  A unit with only a couple of health remaining was adjacent to 2 enemy units for several turns and they instead attacked distant units.  No way that unit should have survived.

-moving to maximize attacks.  When I move to engage I try to maximize the number of units I can get adjacent to the enemy.  I'll also move after engaging if it'll help add in another unit to the attack next turn.  AI doesn't do this, it'll just sit there with no regard for adding in more attackers next turn.

-target selection.  Example -- AI unit #1 is adjacent to 2 of my units.  One of my units has 2 AI adjacent (so swarm=2), the other has 3 AI adjacent (so swarm=3).  All else equal (and in this case all else was equal) the AI unit #1 should attack my 'swarm=3' unit but instead it attacked my 'swarm=2' unit.

Strategic

-preparing before declaring.  Been mentioned many times for many versions -- the AI should prepare to attack before attacking.  Create armies designed to counter the target's units, preposition the armies, then attack.

-spells.  I've never seen the AI cast freeze/tremor or similar spells on my armies in their territory.  I've never had any AI cast damaging spells on my armies that can be cast out of the AI's territory.  I've never seen an AI couldwalk in to defend against my incursions.  I've only seen 2 or possibly 3 spells cast on my cities (and I pay attention to this) in all my games, most of which were on ridiculous and now insane difficulty.

-I'm still seeing cities with no production.  Could it be a 'figment' of a unit being produced and no second unit queued for the 'in-between' turn and there will be a unit queued at the start of the next turn?  or is it a true 'no production'?

-I'm still seeing AI create units in non-fortresses.  I specialize my cities -- my capital is usually my fortress, with all the unit augmenting buildings/spells, and this city creates all my units (until I can add in a second such specialized city) as much as possible, units with all the combat bonuses I can pack in (I then try to keep the units alive -- quality over quantity).

-I'm still seeing AI armies sitting still for several turns (not as long as in the past where a unit may sit for 50 or so turns).

-I'm still seeing unclaimed goody huts in AI territory (someone suggested after I mentioned this months ago it was 'strategy' but Frog or someone mentioned that if an AI passed thru an unclaimed goody hut without specifically targeting it the hut would not be claimed, and this appears to still occur).

-reforming armies.  I cast tornado, AI stack is scattered, and most of the units just sit there.  A few will move, but none will recombine.

-coordinating armies.  I'm seeing multiple stacks approach, I'm not seeing them coordinate attacks effectively to wear me down and overwhelm me.

-army composition.  I'm no longer seeing pioneers in armies, but I would like to see units specifically created to counter the target's strengths/weaknesses.  And better composition regarding melee and ranged, augmented with champion spellcasters (combined arms).

-diplomacy.  An AI offers me a tech trade -- wants 1219 gildar, I have 171 gildar and am netting 2.2/turn.

There's more for both tactical and strategic, but that's all I can think off off-hand...  There's a myriad of threads with other points, such as the one on settlable spots, etc.

I mentioned AI needing work in another thread and the response was:

1- some folks 'abuse' CTL-N for a great starting location and custom Sov creation to 'beat' the AI

2- the response to improve the AI could be to increase the bonus modifiers (AI units are bigger/sooner)

Regarding 1, as I detail above, there's a lot of room for improvement.  I'm not talking about Einstein AI improvements, but basic ones like 'focus fire' and 'target selection' and 'occasionally using the spells provided'.  It's not CTL-N beating insane AI, it's the lack of 'intelligence' in AI.

Regarding 2, this is the easy way out.  Anyone can increase a modifier, or nerf and/or remove spells/abilities.  Adding intelligence to AI is hard, but we have no MP, and we've heard claims the AI will be good.  Increasing modifiers doesn't make the AI better.

If the AI remains as is, LH will still be a good game.  Given the skill of the team working on the game it'll just be a lost opportunity to advance the genre (I'd be happy if it would match Civ4's AI...).

Edited in:

Remember, the above is based upon insane world/AI difficulty.  For lesser difficulty levels the above may not apply.  For example, with regard to focus fire -- on the easiest levels avoid focus fire, on the middle levels make it purely random (it happens occasionally), on higher levels it should happen all the time.  This way difficulty is based more on intelligence and less on bonus modifiers.

Locked Post 55 Replies
Search this post
Subscription Options


Reason for Karma (Optional)
Successfully updated karma reason!
May 20, 2013 6:45:00 PM from Elemental Forums Elemental Forums

Great post!

The question really is whether coding time should be used to balance this or whether players should play at higher levels and in turn give the AI bonuses that address this.

I know what the forum response to the choices presented here would be. But it would be very very time consuming to address some of the things you mention.

Now, with the other parts, like the cities with no production part, along with many of the other areas you mention, those are things I think can be further improved on in terms of AI coding.  Posts like yours area  real gold mine for that.

Thank you!

Reason for Karma (Optional)
Successfully updated karma reason!
May 20, 2013 7:08:56 PM from Elemental Forums Elemental Forums

Yay, a quality AI post that appears to be well thought out and constructive and was Frog approved!

Reason for Karma (Optional)
Successfully updated karma reason!
May 20, 2013 7:28:35 PM from Elemental Forums Elemental Forums

While LH has a much better AI than FE. One of the things I'm lookiing forward to is being able to do post-release AI updating.  It's really hard to do a lot of AI tweaking when the game itself is still changing so much.  

Some of the AI improvements in LH are going to need to be back ported to FE as well as the strategic AI in LH is ridiculously better. But as you can see in this post, there's a lot of room for continued enhancement.

Reason for Karma (Optional)
Successfully updated karma reason!
May 20, 2013 7:44:48 PM from Elemental Forums Elemental Forums

Not going to lie to you Frogboy, that is kind of what killed my want to finish my armor mod for FE.  Every time I thought I had what I wanted to do finalized and would start to implement something drastic would change by the time I was able to work on it.  Though I was delayed by school/life circumstances you are delayed by management/life circumstances.

Reason for Karma (Optional)
Successfully updated karma reason!
May 20, 2013 8:11:32 PM from Elemental Forums Elemental Forums

One suggestion: instead of having folks play on harder setting- why not make a small AI cheat that guarantees the AI's a starting spot no worse than what the player has?  I wouldn't consider that a cheat.

 

I've found the game a bit easier than FE, but I assume it's from playing it more- I can tell the AI has been improved, but more improvements can be made.

 

As for the AI improvements being time consuming, I think they are, but I think the sake of long-term replayability, it's important, even when folks beat that- the game will be harder for the folks giving AI bonuses.   You can always improve the amount of cheating- you eventually run out of resources to improve the AI.

 

Then there's the possiblity of a future expansion.  I still think there's room to expand the content of the series.

 

I've not seen the AI use necromancy.  That is useful when used correctly.

 

 

Reason for Karma (Optional)
Successfully updated karma reason!
May 20, 2013 8:56:14 PM from Elemental Forums Elemental Forums

I much prefer AI intelligence improvements over cheating. However if the AI must cheat (which is must on higher difficulty levels) try to make it non obvious. Small bonuses within the existing rules, rather than big obvious bonuses/values. Helps with immersion rather than it being obvious the AI is playing a different game.

Reason for Karma (Optional)
Successfully updated karma reason!
May 20, 2013 9:19:23 PM from Elemental Forums Elemental Forums

First, thank you for taking my post in the spirit I intended -- LH is a very good game (I'd give it a metacritic of 90 as it is now, and I forgot to mention amongst its qualities that its been very very stable) and the loving craftwork that's gone into it shows, and that given the great crew you've assembled I have high expectations.  And that it's not that the AI is bad, it's that the rest of the game is that good.

Good AI is very very hard and time consuming.  I use Civ4 as a goal -- and yes Civ4 AI leaves a lot to be desired, but it's still a good goal.  Civ4 AI has what... 20 years to build upon?   LH has what... 2 years?  Is it an unfair comparison?  Yeppers it is.  Still, it's good to have a goal!

Regarding 'no construction' -- it may be an artifact of 'between turns' or possibly the city graphic not properly updating (see below for explanation).  I've been watching a neighboring city and the good news is that on consecutive turns I saw:

Starting off Fall 229:

1 turn left on the Mill.

Spring 230 (the very next turn):

Mill was built, but no construction, and no 'flicker' (see next regarding 'flicker').

Summer 230:

Again, no construction?  Not quite.  I was watching and I saw "guardian statue (1)" flicker for an instant, before the 'no construction'.  So, it appears a guardian statue was built but nothing else queued -- so something was constructed this turn.  Was the guardian statue queued the previous turn and that just wasn't updated in the city screen graphic?

Fall 230:

Slave pen is queued.  At first I saw a flicker 'Slave Pen (3)', then it changed to (2).  Was the slave pen queued the previous turn and that just wasn't updated in the city screen graphic?

So, at insane difficulty things can be built so quickly it can appear nothing was built that turn.  I can't say if this occurs at lower difficulties -- it would need to be tested at easy difficulty to see if the AI is allowing cities to not build during a turn.

What may be happening (my guessing):

-fall 229 Mill has 1 turn left

-spring 230 mill built and guardian statue (1 turn) started but this isn't updated on the city graphic (it shows 'no construction')

-summer 230 guardian statue built and slave pen (3 turns) started but this isn't updated on the city graphic

-fall 230 slave pen decremented 1 turn from 3 to 2, and this shows on the city graphic.

If the above is accurate, then there's no 'no construction', instead the city graphic isn't properly updating.

In support of this theory, it used to be that changing the order of construction wasn't updated in this same city graphic until the next turn.  I noticed in .92 it now updates immediately.  Perhaps something changed or didn't change?

I can provide a save starting with fall 229 if requested.

I mentioned 'the good news' above... That suggests there's bad news, and there is.  Note the missing spell on the city each turn.  At insane difficulty there should be plenty of mana for city spells.  Kraxis' power rating is 10x mine (3k vs 300), he's got 16k gildar, 3.3k crystal, 5.8k metal, etc. etc. etc.  It's not that a scrying pool was just built either.

So, to sum up, 'no construction' may not be occurring; and city spells need to be double-checked.

Regarding AI improvement priorities (that don't involve increasing bonus modifiers), I respectfully suggest something along the line of:

-some target prioritization work (nothing fancy, and nothing too set in stone as that invites players figuring it out then taking advantage of it)

-focus firing on said prioritization (again, nothing fancy as above)

-strategic spells, especially tremor/freeze and especially tornado on our stacks in their territory, and tornado and the other spells that can be cast anywhere on our stacks in our territory.  Also, spells cast on our cities (and some UI notification of them -- we really need a log that we can go back and review in later turns)

-better AI city placement to take better advantage of resources

Some relatively quick work on the above would likely drop me down at least 2 difficulty levels.

Pre-planning wars, and determining long range attack plans so not accepting tribute from said long range intended targets (which has bought me a lot of time in my current insane game) would drop me down another level I reckon.

Reason for Karma (Optional)
Successfully updated karma reason!
May 20, 2013 9:20:55 PM from Elemental Forums Elemental Forums

delete please, I hit 'reply' instead of 'edit' for my previous post.

duh...

 

Reason for Karma (Optional)
Successfully updated karma reason!
May 20, 2013 10:11:26 PM from Elemental Forums Elemental Forums

Just throwing this out there - it would be pretty cool to see the AI use the raise / lower land spells.

Reason for Karma (Optional)
Successfully updated karma reason!
May 20, 2013 10:15:19 PM from Elemental Forums Elemental Forums

I am fascinated by AI, so I am throwing in two cents.

It seems like if it doesn't affect war-waging, it won't make the most immediate impact. It seems like tactical battle improvements would be the quickest way to increase challenge. If heroes had fireballs and blizzards and went for the turn reduction trait, or in general aimed heavy hitting spells at sensitive targets, the AI would be a lot more deadly. Summoning heroes would be powerful as well, plopping down air elementals. It seems at high difficulty levels with mana bonuses, mage heroes would be a nightmare for the player. I have never seen them do much that was deadly. Focusing hero traits on the power melee skills and using them in combat would be a big help. I haven't played the latest beta, but the AI seemed to pick traits randomly or in an unfocused breadth first way, if they had set paths of trait selections and combat styles (a battle plan to maximize special abilities) they would be a lot more effective. It may be most efficient to wait for the trees to finalize, but the player uses heroes this way, and heroes seem almost useless if they aren't focused.

I think if someone is clearing high difficulty levels, they are using some tactics which kills many troops with minimal losses. Other than heroes, spell staff units and to a lesser extent archers which target the most dangerous troops (the same kinds of units on the other side) are the only kinds of units that can take out better equipped units with little losses. Having a mixed army of some melee buffers with heavy ranged attackers would seem to be the most effective troop-only type of army. Units have simple tactical AI, so targeting is the only issue. Making sure to target any opposing mage who has something like a no-wait fireball would be a must. The AI seems to give the sovereign the least precedence in targeting, that seems too nice at high difficulty levels.

Reason for Karma (Optional)
Successfully updated karma reason!
May 20, 2013 10:47:17 PM from Elemental Forums Elemental Forums

I think the thing that's bothering me with the AI is a lot of the time it appears weak--I don't really know whether it ACTUALLY is weak; i'm having little trouble beating it on challenging although I am using some save-scumming in there so that's an obvious factor throwing things in my favor.

But like, when the AI uses beguile on an archer or webs an archer in the back rank with nobody nearby it tends to make it look like the AI is incompetant, regardless of whether it's true.  Same deal with the strategic spells--I haven't been at war enough to be sure how often they use them, but even if they are winning fine without them it still creates an appearance of weakness when they completely ignore them.  (And like, if they ai could use the raise/lower terrain spells; idk exactly how hard it is to set up, but it'd do a lot to make the AI seem more like "wow this AI is good" to someone who hasn't played the game a whole lot.)

Same with things like strange city placement--IDK if it actually makes sense in the context of the AI player's faction composition, but I recently captured a city that was in the middle of a forest with 4 material, 2 grain, and 3 essence, with the only strategic resources right next to the town center, and the AI had made it into town for some reason; which seems like a really odd approach; although maybe they already had sanctums and forts or maybe the ironeers hate sanctums (although I can't imagine they'd hate forts) but it seemed like either of those would've been a better choice.

IDK I haven't really played LH that much yet so I can't say too much specifics, but the less than optimal tactical choices does bother me a bit (though I kind of doubt it'll actually effect the overall skill of the AI much if it's improved, especially if its simming any battle they have that isn't with the player).

I would definitely prefer a decent AI that doesn't cheat (or cheats subtly to shore up some areas where AI can't keep up with humans) than just going straight for bonus resources, even if the former isn't actually as challenging to play against.

Or like, for comparison to similar games, I completely gave up on playing Warlock because the AI just... couldn't handle some things and it made the experience seem... idk uninteresting because the AI wasn't capable of providing a challenge, I guess.

Sorry this is probably a bit disorganized but anyway my thoughts on the AI so far.

Reason for Karma (Optional)
Successfully updated karma reason!
May 20, 2013 11:05:45 PM from Elemental Forums Elemental Forums

If you don't mind fielding questions, there are some things I am extremely curious about. How long did it take to design the AI system? Do you use any statistical learning in your design? How do you determine and modify the weights on priorities? Finally, if you had not gone with a fair AI approach and instead just gave the illusion of playing a human-like game without playing by the rules at all, how much time do you think you would have saved (in this game and over the years)?

Reason for Karma (Optional)
Successfully updated karma reason!
May 20, 2013 11:07:44 PM from Elemental Forums Elemental Forums

I don't think the AI deals with necromancy/skeletal hordes well, or the skill is OP.  (assuming the former)

 

The AI tends to do after skeletons, even when it can flank around them and get to the squishy stuff.

 

 

Reason for Karma (Optional)
Successfully updated karma reason!
May 20, 2013 11:21:56 PM from Elemental Forums Elemental Forums

My biggest gripe with AI is that it does a really bad job of evaluating power. Part of this is just whatever formula is used to calculate actual army strength is off... but regardless there is way too much emphasis on masses of weaklings, and too little on the game winning stacks of doom. More granular is the poor evaluation of individual unit power, such as a bear cub being medium, but the most powerful champion accessible in the game being "strong". 

I would love to see this single function be reworked to compare the top <group size> units of a faction relative to the same from other factions, then the second best groups, etc. Some power should still be given to masses of armies, but if a player has a single group of uber units, they should be feared by the AI, not insulted and attacked.

Likewise, all AI seem to be opportunistic assholes who love everyone who is stronger than them and hate everyone who is weaker than them. This can be effective, but it needs to be mixed up. I would love it if a powerful faction declared war on my enemy who (mistakenly) thought he was stronger than me just because he didn't like that I was getting picked on. As stands, no AI can be remotely trusted unless they are far weaker than you.

Reason for Karma (Optional)
Successfully updated karma reason!
May 21, 2013 3:06:16 AM from Elemental Forums Elemental Forums

Ok, I'd like to amend my above post.  I've now seen an offensive spell cast on one of my armies while in enemy territory:

Ouch!

Very very cool.  Keep it up!

On the other hand...

Krax declared on me.  You can see one of his cities just above my city 'Hold':

His city had no defenders, just the 5 city defenders (2 archers, 3 militia).  They were easily dispatched and the city taken.  Probably not for long as Krax is nearby and approaching with an army about 3x the power of my best army:

I wasn't worried, however, as Karavox had to cross a few squares of my territory so I could cast some spells to delay him (freeze, tremor), to split up his army (tornado), and to weaken him (flame this and ocean wave that and falling stars and all those cool spells we have).  At least I wasn't worried until I saw:

"What's all this then?" I asked myself...  Never seen that before.  I looked carefully at Karavox and all his items/skills/abilities, and see nothing to account for this.  Same with all of his 8 companions in the army.  Did a google search and saw no mention of this other than here (mods is enabled but no mods are installed).  Is this normal for Karavox -- an ability or spell or...?

So, in summary:

-cool offensive spell cast on me by AI in his/her territory -- yea!

-AI declared but left adjacent city undefended (when he has plenty of powerful armies running around) -- boo!

-AI sov immune to offensive spells -- wtf!

Reason for Karma (Optional)
Successfully updated karma reason!
May 21, 2013 5:05:21 AM from Elemental Forums Elemental Forums

There is the "Tower of the Mage" level 5 conclave city improvement.

It says: "All Units trained in this city gain 10 Magic resist and the city is protected from offensive spells."

Maybe that is a misspell and it means the units trained are protected from offensive spells which would make it so useful (The written effect is pretty meh).

I cannot explain your issue otherwise.

Reason for Karma (Optional)
Successfully updated karma reason!
May 21, 2013 10:42:11 AM from Elemental Forums Elemental Forums

Is there an ophidian or Obsidian golem in his army? 

Reason for Karma (Optional)
Successfully updated karma reason!
May 21, 2013 6:36:40 PM from Elemental Forums Elemental Forums

Thank you N1ghthavvk and davrovana for the replies!  davrovana you are correct, there is an ophidian in the army.  It seems the most likely explanation.  I hadn't considered that the presence of a magic immune unit would protect the entire army.  Seems a bug to me and will be reported as such.

Thanks again to you both 

Looks like I'm in a bit of trouble in my insane/insane game now  lol

Reason for Karma (Optional)
Successfully updated karma reason!
May 22, 2013 1:35:30 AM from Elemental Forums Elemental Forums

Now the Ophidian from the Quest does have a bonus! It was always too useless to me. Maybe now I'll actually try to feed it long enough to not die every fight that I autoresolve.

Reason for Karma (Optional)
Successfully updated karma reason!
May 22, 2013 8:52:11 AM from Elemental Forums Elemental Forums

Now that the AI casts strategic spells, ophidians may be an essential part of an invasion force. 

Reason for Karma (Optional)
Successfully updated karma reason!
May 22, 2013 10:12:04 AM from Elemental Forums Elemental Forums

Really interesting post. I agree with pretty much all of it. I merely play on Expert, but I should probably raise the level; with anything other than a rubbish starting position I'm pretty confident of taking any AI opponent out with one decent stack. The game is fun without more of an AI challenge, but it would be even better with a better AI.

1) +1 to the AI only building units in fortresses. The AI should have a target it wants to spend on wages compared to income (e.g. 90%) and should build units in fortresses until it reaches that point. The target should go up when it's at war, and if it's a long way off it's target (say half) it should build units in non-fortresses, and (crucially) it should put these catch-up units at the front of production queues. This would deal with one of my biggest AI gripes, which is that if you kill two of the AI's stacks the war is effectively over; it never really tries to replace lost units fast enough. It is pointless building the Tower of the Witch in 30 turns if you're going to lose your cities in 10 turns.

2) +1 to the AI using stack attacking and immobilizing spells properly. On high enough levels it normally has mana to burn, and it should put immobilizing stacks which are within one move of its cities as a very high priority. Damaging spells are fine if it's planning to actually attack that stack. If on the defensive (say it has an enemy stack within its borders which is more dangerous than any of its stacks) then it should consider using movement spells to get its units together to either attack the danger stack or defend its cities.

3) +1 to the AI specialising its heroes. A lot of the best skills are at the end of the tree. Once the skill progression is stable this presumably becomes a lot easier.

4) The AI should also be better at deciding what to build in its cities. Production bonuses and to a lesser extent growth bonuses effectively stack; if you do them early you get the benefits later. It should prioritise buildings which raise production, either directly or through lowering unrest. It should know whether -10% unrest or +1 production per material is better. It should consider buildings which increase zone if the city is not contiguous to its capital. It should consider building growth enhancing buildings if they can be built before it would go up one settlement level anyway.

5) When on the defensive (e.g. it can see a stack in its territory which it doesn't have any stacks big and close enough to attack it with) it should act defensively. Prioritise building units at the front of production queues, move its units to defend cities near the threat, etc. At the moment I can leave my stack outside the AI's capital and it will eventually take its monarch on some pointless journey, allowing me to take a city which would be extremely hard if the monarch was there with accompanying troops.

Any AI enhancements would be gratefully received. I would have thought "standard skill choices" for heroes would be a very easy win when the skills tree has settled down. A better tactical AI could also make quite a big difference, and is presumably easier to test than some of the strategic changes I've suggested. I would have thought though some strategic improvements could be made without ridiculous amounts of effort.

I never really expect AI to beat a human fairly, and Legendary Heroes AI is certainly not terrible. It would be nice if we could eliminate some of the stupid mistakes, though.

 

Reason for Karma (Optional)
Successfully updated karma reason!
May 22, 2013 2:31:28 PM from Elemental Forums Elemental Forums

In response to the people complaining about target selection (specifically with Shrills, Spiders, ect), it was my understanding from previous AI discussions from Brad that the AI itself was different for differing enemies. Aka, more bestial enemies favored a more random target selection, intelligent monsters more focused, with the best calculations reserved for faction AI.

Reason for Karma (Optional)
Successfully updated karma reason!
May 22, 2013 8:18:55 PM from Elemental Forums Elemental Forums

Quoting davrovana,

Now that the AI casts strategic spells, ophidians may be an essential part of an invasion force. 

Quoting N1ghthavvk,

Now the Ophidian from the Quest does have a bonus! It was always too useless to me. Maybe now I'll actually try to feed it long enough to not die every fight that I autoresolve.
It's even more insidious than I first thought!  It turns out that the ophidian in the one stack not only protects that stack from offensive magic (except for tremor spell, that still worked) but if you target a non-ophidian army with a multi-square AoE spell such as starfall, and the starfall AoE would include the ophidian-containing army, the starfall spell can't be cast on the non-ophidian army.

The non-ophidian armies are subject to single-square AoE spells of course.  Still, seems pretty broken to me...

Reason for Karma (Optional)
Successfully updated karma reason!
May 22, 2013 11:09:53 PM from Elemental Forums Elemental Forums

Another example of where the AI can be improved.  I'm at war with Yithril, as is Resolyn.  Screen below shows a Yithril outpost, one of my armies to its right, a Resolyn army below mine (with the spider), and a Yithril ogre below the Yithril outpost.  The Resolyn army came in from the left (I'm 99% certain of this) but didn't capture the Yithril outpost.

Next, is the beginning of the next turn:

My army hasn't moved, the Resolyn army moved northeast then north.  Again, the Resolyn army didn't capture the Yithril outpost when it easily could (and should) have.

Here's a screenshot after I moved that turn and captured the Yithril outpost, same turn as the second pic:

There's no reason I can fathom why the Resolyn army didn't capture the Yithril outpost in the turn of the first pic, and certainly no reason why it didn't in the turn of the second pic.

Resolyn is at war with Yithril, Resolyn is getting its butt kicked by Yithril, and Resolyn forgoes an opportunity to capture a Yithril outpost in what used to be Resolyn territory.  I understand that on the easier difficulty levels this might be desirable, but the game I'm playing is insane world and all insane AIs (and I'm the sole Kingdom vs all 5 default Empire AIs).

Yes, there's a lot of good AI in the game.

Yes, good AI is hard and takes time.

Yes, this is one example of how the AI can (and I believe will) be improved in upcoming versions.

Yes, this isn't intended as a criticism, but rather as a notification of where improvement is possible.

LH is a very very good game.  I see that a review gives it 9/10.  I think that's accurate and well deserved!

Reason for Karma (Optional)
Successfully updated karma reason!
May 23, 2013 7:51:24 AM from Elemental Forums Elemental Forums

This is a great post, Nick-Danger!

Brad, given how expensive it is to script the AI, it would be sweet to open up the AI's scripting to the modders. Didn't the Civilizations community do a ton of work on it to make it better?

Reason for Karma (Optional)
Successfully updated karma reason!
Stardock Forums v1.0.0.0    #108432  walnut2   Server Load Time: 00:00:00.0001375   Page Render Time: