I really like the Elemental series. Pre-ordered way back when... Got 292:56 hours into LH. The skill and dedication of the Stardock crew is very evident in the game. This is one of the better games I've played. And I especially appreciate and respect the continuing upgrades/improvement to the game. Many companies would have long ago moved on, and Stardock has abided by their promise to keep improving the game. That means a lot!
The AI still is getting better and better but it's still in need of improving -- especially in comparison to the artwork (I really love all the detail -- the artists did a great job), the spells, the modability, the... well all the other components. Good AI is especially important as there's no multi-player and the AI is all we got.
This isn't a complaint! It's 'feedback'.
Current game, using Heavenfall's excellent CoS mod (updated) and my variation of Odinlowbane's 'Duke the Wolflord' mods, huge world, Insane difficulty, and only 1 AI (Skeleton King), season 283:
Yep, while there's a few enemy in a couple of their cities, they have 32! armies sitting in one spot. They've been there quite a long time. We've been at war for quite a while, and Skelie has a number of mobs wandering his lands, yet his armies are having some sort of Burning Man (literally?) ceremony...
All the AI cities -- and by 'all' I mean every single one of Skelies 12 cites -- say 'No Construction'.
It's not that Skelie can't afford more units -- he's got 32,226 gildars in his bank.
Perhaps Skelie has built everything possible in every city (if so then why not build growth/mana/wealth/research?).
Most of Skelie's cities have 15 buildings listed but a few have less (and a few a couple more) so he has been constructing buildings.
I've shown in previous posts that cities with build times of 1 turn can be constructing building and it just doesn't show in the 'constructing' list above the city, but 12 of 12?
There's a lot of open land ripe for settling. Skelie hasn't built a new city in a long time.
Given the above, I think this 'all cities have No Construction' thing needs examining.
I appreciate the DLC packs, but I'd like to see a DLC for 'Sooper-Dooper AI'.
AI design is done with the same people that do art design in startups.
However, in larger studios (and Frogboy can correct me if I am wrong for any of this) the games are roughly made as such:
Stage 1
Engine, Gameplay, Story/Quest (basics)
Stage 2
Dialogues, Levels, AI (meat)
Stage 3
World Design, Graphics, Sound (polish)
Now you are going to need some people with skill in both science and art, some that lean heavily on science, and some that lean heavily on art.
If you are going to get by with one person in each of those types, the tech person will probably do Engine, AI, and Graphics. The art person will probably do either Gameplay or Story, Dialogues, and either World Design or Sound. The 50/50 person would do whatever was not taken yet.
If you had a AAA studio with one or more people that they can use in each section (9+ total employees just for making the game) then you can have 6+ of them sit around at each stage and do minor tasks for the people in charge of the current stage, but I think most game studios out there are probably going to try to use the same person in multiple stages if at all possible.
Because of that I think it's not going to be that uncommon that you see the same people who make the AI also making the Graphics, for instance, even though Engine/AI require a lot more of a scientific approach whereas graphics are better done by someone that is closer to 50/50 science/art in their skillset. The polish up process is more artistic and they don't want their science person to just sit around through the whole stage. It really depends on how many people you have, but what it mostly comes down to is cost effectiveness and making the most out of your team.
In any event, I just wanted to throw it out there that I think a lot of dev teams are smaller than people realize and that sometimes people have to step out of their comfort zones during the game dev process as a necessity.
Is Capitar fighting someone else? I've seen similar behaviour (i.e. I don't see an AI's units even though it declared war on me, even though we have borders) when the AI was fighting another faction. This seems fair enough if the AI has most of its army fighting on the other side of the map. It would be nice if the AI realised it was losing a war on two fronts and made more of an effort to make peace with one of its enemies, but that might not be straightforward to program.
I actually think the AI has got a lot better at building attacking stacks. Certainly the AI's initial attack is done in force, and it seems to have got somewhat better at replacing its losses as well. Garrison cities seem to be much better defended. I haven't seen the problem with stacks of armies just sitting there. I'm sure it happens; if it can be reproduced I'm sure the developers will do their best to fix it. Keep giving feedback and saved games!
One of the most recent changes is that the AI definitely casts a lot more spells in tactical combat. Occasionally maybe it goes too far, sometimes hitting someone with your big sword is more effective than casting Candlecloak, but I wanted the AI to cast more spells; the developers looked at this and now it does. It's now a big fan of casting Cloak of Fear, which is quite an effective spell which forces you to change your tactics slightly. As I say, the continuous improvement is impressive.
One thing which still could be improved in tactical combat is movement and focusing melee damage on weaker units. One of the biggest differences between the way I fight and the way the AI fights is that I have "preferred melee targets", and if I can't go round a unit, I will stop next to it and attack it, but on my next move I will move one tile past it, and probably the next move after that I will be able to bypass it and attack those more vulnerable back row units. The AI never does this; the only times I've ever seen it move a melee unit that is in contact with one of my units is that it retreats when its unit is very low on health, and on one recent occasion I've seen a hero break contact and attack a different unit. That's the only time I've ever seen an AI unit move to attack a different unit. Has part of the tactical AI might have changed to allow it to happen? Anyway, it would make quite a significant difference if the AI moved its melee units after contact; it's fairly easy to game it now by moving a high defence unit into position and putting it on defend; you can pretty much guarantee that even if the AI could break contact and kill another weaker unit, it won't.
I'll write more on the particular challenges on LH AI but I will say that if people call the AI in 1.5 "broken" that I think it'll be time to throw in the towel on lh AI. I'll know then that some people have unrealistic expectations.
That's a reality; don't throw in the towel because of them.
I love LH and I think improvements have been noticeable since I started playing it.
Keep up the good work for the rest of us: we are the majority.
_________________________
LH Mods by Primal
XtraDeconstruct
XtraDeconstruct Canons
The AI is not broken. Since you've added the ability for the AI to do summons, I'm happy. I don't know about these other complaints people are having, since I do not encounter these issues in my games. I've got to experiment on the 1.5 diplomacy AI at the moment.
AFAIK, there is a rule that says basically if an AI melee unit is in melee range with one of your units then it can't move.
That means you can go stand a unit by an enemy melee and be 99% sure that melee unit will attack what you put next to it.
The only choice it has is if there are multiple units in melee range, then it should pick the weakest target.
I think that kinda oversimplifies things, but it also makes the game more fun for the player, because it rewards players for learning the rules the AI plays by and rewards players for learning how to play to maximize their own effectiveness.
Actually, D&D and some other games are going towards a similar model since it has been a big problem in the past that units were disengaging from melees and going straight to the back line units. The D&D solution is to let the melee unit attack the disengaging target if that target is making a long move. If the attack hits then the melee can force the moving target to stop moving.
That type plan is harder to make work in LH (because everything always hits) but it works pretty well in D&D where misses are much more common.
For LH, I am fine with the work around that no melee unit can ever move if it is in melee with another unit.
I think trying to implement the D&D system would frustrate players more because it makes it much easier for the AI to kill player units.
- Edit - In a game I am playing right now, I saw a pioneer in an army. That situation is still occurring. It may be that the AI just happened to accidentally stop both units on the same space, but I just wanted to report that it does still happen. It isn't often, but I have seen it even in 1.4.
- Edit 2 - Example
I can think of a lot of areas, endless areas really, where the AI can improve. The biggest time sinks for me have been systems that just don't work right. The path finding continues to be my bane in that it will, for reasons that are unclear to me, simply return an error sometimes OR worse, it'll do something really stupid (like walk a sovereign right through the middle of enemy players to get to his destination.
I've spent a lot of time fixing that kind of thing and it's helped players with their own units quite a bit (since the same stuff affects them as well and pathfinding was a bane that affected human and AI alike).
There's a lot of design stuff that's harder to deal with. For instance, those stacks of doom are just ridiculously hard to deal with. You get a dragon, you win. Period. The answer to a lot of that stuff is to alter the actual design.
AI is good at brute force calculations. Humans are very good at precision strategies. The AI fights with a club. The player with a sword. And FE makes the player's sword very very sharp.
There is still a lot of room to improve the AI, low hanging fruit even. But I do this stuff on my own time because, and this is something some of you just need to accept as true, the percentage of people who think the AI is a push over is small. But those people are my people and I'm willing to work on it to satisfy them as long as I feel like my time is being appreciated in return. That's what I feed on. But it absolutely sucks the fun out for me to spend my weekends working on the AI only to have someone load it up and say "Still sucks!" when I know that objectively, the AI is much better becase I see the difference in play testing.
For example, I spent a lot of time teaching the AI to *intelligently* cast strategic spells on the player. Yet, you rarely see any acknowledgement of this. Play some other fantasy game and tell me if you can think of any fantasy strategy game, in the history of fantasy strategy games, that casts strategic spells even remotely as well as LH does.
Then you have people who post "Well, HOMM 3 had good AI". It makes me want to tazer them. (I liked HOMM 3 but there was very very little to it).
I wrote the GalCiv AI that's considered pretty good. The LH AI is vastly more sophisticated than the GalCiv AI. The difference is that there are so many more "things" to deal with that computers just aren't good at handling.
I can put together an amazing AI army that is simply wiped out by a tidal wave. I can send a group into battle that gets destroyed because the player casts a spell that makes their tank invulnerable for 5 turns. We had (until 1.5) food cheese (I assume it's fixed) where players could load up on healing potions and just keep drinking them in battle. And then we have, (biggest pain in the rear imo) the issue of how hard it is to get things moved around huge maps quickly (not teleporting, I'd be happy with just something that makes units move hella fast when in neutral or friendly territory.
Sorry to rant.
Let it all out!
Hum, what prevents the AI from building its own stack of doom?
Is there any way to have the AI aggressively pursue dragon lairs (and the technology to get dragons)?
The biggest asset for dragons (imho anyway) is their fear spell. Any ways to introduce items (Unit design) that block that spell (to varying degrees) and have these items not available for single heroes (Could make sense, group courage or something along those lines).
One thing is for sure anyway, by keeping the discussion channels open, we can come up with great ideas to help you out. Just come up w/ your problems and we will come up w/ solutions
I, for one, appreciate you spending your weekends on improving the AI!
So basically you are looking to implement Civ 5 railroads that let players get from one end of their empire to another in 1 turn?
I am not going to say anything bad about the AI, I think overall it is quite good.
They garrison their cities with enough force to wipe out all my cities, but I think I might rather have it garrisoning their cities than wiping me out anyway.
I am not such a huge fan of giving the AI huge bonuses to everything production as a way to increase difficulty, but I know it is a reality that there aren't many better ways to compensate for the fact that there is a human controlling one of the players.
One thing I would really like you to fix is that AI heroes for some reason always have like 30 horses/wargs when I gain control of them. That and like 20 different weapons. I don't know how they get all that stuff, but they should sell what they don't need and if they are going to buy 15 horses they should probably equip them. I have seen some of them not use mounts when they had 20+
I agree with this statement. There does need to be a faster travel plan when in friendly or neutral territory (or huge or large maps). Might I recommend a new tech on the tree or attach to one of the existing techs that gives units movement in friendly territory, +2 movement, and in neutral territory +1 movement. Also, as I've seen others say the AI is not building upgrades to outposts and the stables upgrade yields double movement in the ZOC.
Honestly, I've put a stupid amount of time in to LH and I can say with no reservations that the AI is really good for the average player. I'm a pretty good player and I have problems some times with the AI players, especially the damn Yithril.
Do they do some dumb shit sometimes? Yeah. Does the AI make mistakes? Sure. Is the AI still better than most games? Absolutely.
I'd still to see some of the dumb shit fixed. Wither and Tremor on caravans. Web over and over again. Other things mentioned here.
But Froggy, it's damn good. Not the GalCiv of fantasy 4X, but what is?
Not a HOMM game, that's for sure.
Oh yeah, the most annoying quirk is the AI sovereign stockpiling issue.
Can that be fixed?
AFAIK, web is cast at a random target. If it is cast twice at the same target it is probably because it randomly pulled that target twice.
I am kinda fine with that, because the AI is never going to be able to target that intelligently anyway and it saves a lot of man hours to fix other things.
Why not just change the modifiers for the cityhubs such that they increase movement within them for allied forces and retard ones that are hostile? In neutral territory it would be normal speed and then in someones hostile territory their cityhub modifiers will take over.
I want to throw my 2 cents in. I bought the game nearly a month ago and after having sunk over a hundred hours into it I can't say I don't love it or that I didn't get more than my money's worth. However quite a few things mentioned in this thread are taking away the challenge and continued enjoyment for me.
In the tactical part of the game, spiders webbing webbed units, units attacking a champion with diamond skin, or enemies casting spells on my magic immune ophidian really make me feel like stealing candy from a baby. Also, in some bigger battles I find that a few units just sit around doing nothing. In my current game I had an encounter with wildlings where most of them swarmed me from the start but a single enemy (warrior?) and one group just sat there while I mopped up the ones that attacked and finished those two off. Another encounter was in the quest with the three dark mages trying to drain the life from the spider queen. Two of the mages cast spells at my group but one of them just sat there waiting to be dispatched.
In the strategic part of the game I find the aptitude of the AI at waging war to be sometimes hopeless. In my last game I shared a border with Yithril who declared war on me after his power rating dwarfed mine. Watching huge stacks of juggernauts and heavily armored infantry in his territory I was about to quit. I decided to play on just to see how bad the slaughter would be and didn't get attacked at all. I brought in a pair of my early-game scouts from the other side of the map and had them running around his territory capturing outposts and destroying resource improvements and he didn't do anything to address this. They were no match for even a single one of his units but he let me keep doing it until one of the scouts ended his turn right next to one of Yithril's units. He was fighting someone else before he declared war on me, so that may have something to do with his lack of aggression, but if he had units doing nothing in his territory, he could at least use them to chase down my two scouts who were doing grievous harm to his infrastructure.
A little off topic but which of the vanilla factions do you consider giving the biggest challenge in the game? I know the answer is "it depends", but I've always found Yithril to be at the top of the power ratings in games. Has anyone ever made a custom map where the player's position is on an island or otherwise cut off from the AI (is that even possible?)? You could put your starting units in your capital, set it to produce research and turn on auto-turn, then watch how the AIs play the game out amongst themselves. It could be a kind of AI arena for finding the best sovereign/faction. I'm also curious what kind of custom faction you guys think would pose the biggest challenge when in the AI's hands. I find that some of the custom sovereigns and factions that I've made are usually magic heavy and aren't used as effectively by the AI as in my hands. Any thoughts?
Lots in this thread, clearly still a hot topic!
In order:
Frogboy, I've said it before and I'll say it again, I'm very impressed by the continuous improvement in this game. I've been playing it fairly constantly for months now, and I can't think of any other game I've played where that has been true; I'm normally distracted by a new game at some point. There are a couple of reasons for this; one is that all the factions play very differently, so completing the game as each of the factions is an absorbing goal in itself. This is vastly different to Civilization, where you might get one slightly better unit and slightly different starting technology. The other reason is the continuous improvement; it's worth playing again to see how the AI has improved and how strategies have to be modified because of balance etc. changes. In twenty or thirty hours playing 1.4 and 1.5, I can't really remember the AI doing something which was outright dumb, and that's vastly improved from Legendary Heroes 1.0. Does it make slightly inferior choices to those I'd make? Yes, probably, but that's how I beat it. I don't expect it to be as good as a human player, but it takes some of the fun away if you think you're beating up an idiot. I don't get that feeling any more, which is great. I'm fairly sure I've said before I appreciate how the AI casts strategic spells, but if you didn't see me say it, I'm saying it now! I provided a lot of feedback in the hope that this could be implemented in a good way, and it seems to work pretty well.
Are there things I occasionally criticise slightly in the hope that they may be improved in the future? Sure, but this is in the hope of getting a better game, it's not meant to be ungrateful for all the work which has gone before. For example with strategic spells it's annoying that the AI is apparently able to cast the same immobilizing spell indefinitely, when the player is not allowed to do this. It would be nice if this was fixed (or alternatively if a counterspell was added for multi turn strategic army spells). It would be great if the AI built more prone resistant units, even if just with Boots of the Spider. In an ideal world if it lost a battle where its troops were getting knocked prone all the time it would build more prone immune troops. But I have no idea how hard that would be to program, and in general I am extremely grateful for all the improvements which have already been made.
In terms of faster travel, personally I use Stables in Outposts, so perhaps if the AI could work out how to upgrade Outposts that would help.
Had you cast Wither on the Wildlings? Units with zero attack will behave as you describe. Perhaps they should try to close and use swarm bonuses, but it's understandable a zero attack unit doesn't do much. In the second case it may well be a problem with the spellcasting, which was heavily modified in 1.4. If you can explain in more detail (maybe provide save game and debug.err) than the developers may be able to fix it.
Re: Yithril, bit hard to say, but it sounds to me like it might be the "war on two fronts" problem, and perhaps the diplomacy could be tweaked so that the AI avoids wars on two fronts if it can, as the strategic AI does seem to have some problems in those situations.
In terms of dangerous AI vanilla factions, Yithril is dangerous, I personally find Capitar normally expands very effectively. Tarth (devastating ranged units) and Resoln (hard to hit high dodge units) both build armies which can be hard to beat.
Your efforts are much appreciated. Even your rants
It's funny I have to tell you this, since you're American and I'm not, but here goes: there's this thing called 'Freedom of expression' this means that everybody is entitled to their own opinion, no matter how silly, and has the right to express that opinion. Even if it makes no sense whatsoever. If they want to, people can write the craziest stuff in forum threads supposedly dedicated to useful player input.
Well, there might be some rules and regulations, because here's the good news: Stardock is a private company. Also, as the CEO, you can freely ignore all those different opinions. Athough it would be great if you did listen to a few really smart people: Derek, you and me
Ok, maybe some others are worth listening to as well. And yeah, it's hard to sift through all the not-so-useful posts to find some good advice or criticisms. I understand how it can get to you, reading a bunch of nonsense. But please remember two things:
- All jokes aside: you are the expert, we are not. So please forgive us if our suggestions/criticisms are unrealistic.
- I, probably this whole forum, heck, all forum-posters on the internet I've come across, love Stardock. We all love the efforts you and your team have put into making FE and LH better products, before and after release. We all love how you are trying to listen to us, and it is very much appreciated!
The custom races I make tend to rise to the top of the power charts when played by the AI.
I just played a game where a custom race I made had 260 power rating at a time when everyone else had less than half of that.
There are definitely abilities that work better in the AI's hands than others.
Some abilities it takes strategic thinking to be able to maximize.
Diamondskin has been mentioned, people just walk up by some AI units and cast diamondskin. There happens to be a rule where AI units must melee anything in melee range if such a thing exists, so they attack uselessly. To me that is a good interaction and I like that this sort of strategic thinking can be used to compensate for the many AI advantages.
There could be an exception case where if that happens, the AI will disengage and pursue other targets if Diamondskin is cast, but where do you draw the line? What if they put a defender with full plate mail and all the defensive traits right next to them? Is it OK then for the AI to disengage? If so what is the purpose of the Defender class? There are whole classes that are based around this interaction.
I think the players should just start following the AI's rules. If there is something in melee with you then you can't move. It raises the stakes for the players without turning the AI code into spaghetti. As a rule, it does make sense.
As for some abilities that the AIs tend to use pretty well, I think Decalon and Master Scouts tend to do pretty well in the AI's hands. Not sure why exactly on Decalon, because I am not real sure how much they are even using it, but for some reason AIs with Decalon tend to do pretty well in the games I play. It could be Pariden or a Decalon race of my own making, but for whatever reason they seem to do well.
Yithril seems to do well too, I think because it can easily maximize the value of all of its abilities. There isn't anything that requires strategic thinking there, it is all about brute force and if there is one thing the AI does know how to do it is use brute force.
Also: Dropbox me some saved games with soecific AI weirdness and I am happy to look at it to find out why. Particularly cases where he AI doesn't do something strategic that should be obvious.
The AI's battle skills seem pretty good to me, ...but I only play on challenging.
What I'd like to see improved is the "War for no apparent reason" and the "I hate you just because you're weak" diplomacy.
I don't mind going to war with the AI, I'd just like a cohesive reason for it.
OT: If we can't get units to move faster in friendly or neutral territory, I'd like to see a spell added that allows the transfer (teleportation) of items between sovereigns and hereos.
I think people just want Brad to invent Skynet.
The AI is a he? Brad what are you up to
P.S. Just being silly
N/A
Your comment is the fallacious argument 'hyperbole' -- twist things to an extreme. No one is calling for skynet, we're addressing 'dumb' things the AI still does (like having 30 or so armies just sitting there unmoving when at war with the player, webbing already webbed units, etc.).
Please don't be a willy, and instead keep the discussion helpful. Thanks!
Back on topic -- yesterday I started a new game 1.5 beta, with the usual insane everything, my new custom dead sov., huge world, but with 4 AI instead of 2. I was testing out the 'distance' thing -- the question "does distance cause the AI problems with armies accumulating and not moving out"? (on the other hand, there's some suggestion that if an AI is already at war with another AI then the first AI isn't as effective in warring us, and more AI in a game means more chance of this, so this may somewhat balance out the 'distance' thing...).
It's a few hours into this game and just after placing my third city I meet an AI (the dwarf standard sov guy), our borders touch. Him vs me:
cities: 9 vs 3
pop: 3300 vs 150
strength: 850 vs 50
armies: 9 units of 6 each vs 6 units (just got) of 3 each (but my non-starting undead have the 'kill a unit and get a member' trait so eventually 6 members. I've seen 11 full armies (no pioneers or scouts!) vs my 1 decent, 1 poor, and 1 pitiful armies
spells: not sure about him but from his development he's got a ton, I just got tornado and have tremor and freeze.
AI is tremoring my units in his territory each turn (I'm also seeing messages regarding the AI casting on me, which is very helpful), and ruthlessly moving his 11 armies at me. He's not at war with anyone else. Somehow he got all 11 armies to hit my third city (I'll be starting a new thread asking for ideas on how he's doing it). I was able to fend off the first few armies but against all 11 in one turn I have no chance. It's looking like I'm toast in this game (starting area has a paucity of settling sites, really just 3 and losing one of them, with the second city also quite vulnerable...)
So -- AI in this game is brutally effective!
One thing this highlights is that the only way I survive insane games is because the AI does 'dumb' things (like having 30 armies just sitting around). I'm only an average player, and if the few 'dumb' things were addressed then I'd be playing 'normal' difficulty. Don't take that as "this game sucks", take it as "with a few small changes this game is cut-throat deadly and kicks my butt regularly!"
There are many great features available to you once you register, including:
Sign in or Create Account
