Great game partly ruined by extremely poor AI

We’ve moved over to the paradox forums. Please come visit us there to discuss:
You can still read the collective wisdom - and lolz - of the community here, but posting is no longer possible.

Home Forums Age of Wonders 3 Discussions Great game partly ruined by extremely poor AI

This topic contains 162 replies, has 14 voices, and was last updated by  Graxx 4 years, 3 months ago.

Viewing 30 posts - 31 through 60 (of 163 total)
  • Author
    Posts
  • #256213

    juhah
    Member

    Thanks @hiliadan! So actually the AI would benefit from a strong starting army and weak defenders or is average optimal? How about game pace? Is normal the best for the AI?

    Biggest map makes sense as you say, I’m just a bit unsure just how long the very large map (with UG, because a fantasy game needs to have caves 🙂 ) would take. I played one large map with UG before and it was long game (because I like to research everything and explore everything rather just win asap).

    Does class/race make a difference, i.e. does the AI play some combinations better than others?

    #256214

    Hiliadan
    Member

    I am not so sure what are the best settings for the AI in terms of game pace and starting army and defenders because I didn’t test it (I did test Strong Defenders and Weak Starting army and the results was bad), but basically the game has been balanced by the devs for default settings so I would think Normal on everything will yield the best results.

    I personally do not like maps bigger than Large and even on Large maps, I feel it’s too big, because after turn 60, you’re just moving big armies over vast distances without much real actions for many turns.

    Yes, the AI does better with some combinations rather than others. As I said, I stopped playing SP a long time ago and now only play PBEM so I’m not the best one to talk about the best combinations. But what I know is that the AI has more difficulties playing Necro (one reason being the lack of economic bonus through happiness, which slows it down) and the gap between what a player can achieve and what the AI achieves is among the biggest for Necro, because Necro has many tactical battle tricks that the AI can’t pull out well.

    JJ is right that mods can improve the game but they don’t make the AI smarter, they just make it harder to beat or make the game more enjoyable overall, independently of AI.
    You have JJ’s mods and you have the PBEM & Single Player balance mods for general balance (+ a live MP balance mod but that’s not your focus).

    #256215

    Jolly Joker
    Member

    Just let me reiterate something so it becomes clear what I mean.
    There are basically two different things that combine to become game difficulty.

    One thing is the amount of EXPLOITATION the system allows the good human (as opposed to AI) player. What do you need to beat defenders? How long does it take you to level heroes up to significant levels – in short, INDEPENDENTLY of every AI player, what can a human player accomplish.

    The second thing is, how apt is an AI player in making it difficult for you?

    The first thing is QUITE important because it defines how long it takes you to really get rolling. If you play standard unmodded Necro, for example, a lot will depend on Leader (and first hero) development. Control Undead isn’t that good – but Greater Reanimate is, and once you reanimated a Titan that Titan in turn will enable you to roll faster. Then comes Inflict Ghoul Curse, and then you waltz into Dungeons, Ghoul Manticore Riders and Phalanxes and Warbreeds and at that point it snowballs. The only chance to be stopped at that point is if there is a powerful PRODUCING opponent, like a Warlord or Dreadnought.

    So it’s all about timing. The AI handles Sorcerer quite well, who come off early, and the main point is, to give the game a chance to build up opponents before you start snowballing.

    This is only possible by making it as difficult for the human player to get going. With the unmodded game this is more or less possible, depending how good players are, but the main problem is, that unmodded game allows heroes to develop too fast, since the system of XP gain is too exploitable.
    Even if you do NOT know the technical details, once you grasp that the more you do in a battle, the more XP you get, you’ll try and extend battles even more as you already do it in order to make use of all healing abilities in battle – just web the last guy standing and cast yet another Awaken Spirit or something.
    And the more difficult the settings you pick, the more you work yourself into exploitation paths, because it’s the way to overcome everything.

    So if you want to play a fair game as a GOOD player, you have to make sure you cannot exploit the game – without taking the fun out of it, because obviously finding ways to beat the system is part of the fun.

    #256216

    Gloweye
    Member

    @gloweye jolly joker on the other hand said the opposite, so I wanted to give it a shot.

    Likewise weak army, as it can produce units much faster. How would I benefit more from these than the AI?

    The point is, AI tries to clear with it’s starting army and build up it’s city a little first. But then it can’t fight any site it can guaranteed clear, so it’ll try one it thinks it has a decent chance of success. Then it loses units, and even the decent chance gets pretty small, thereby showing it’s “feeding units to treasure sites piecemeal”

    You, on the other hand, will pay good attention to tactical combat and prevent losses, even if the odds are against you. So early game looks harder for you, when you’re against independents, and late game will be much easier, when you’re against the gimped AI’s.

    How is slow game my advantage if the AI is faster at research and economy?

    Because you will compensate by focusing on research and storehouses, while the AI won’t.

    As a closing note, it would be perfect if we could set AI starting army independent from our own – to have us start with a weak starting army and the AI extra strong against strong defenders would make for a completely different picture here, and much stronger AI’s when you find them.

    • This reply was modified 4 years, 9 months ago by  Gloweye.
    #256218

    Hiliadan
    Member

    Just let me reiterate something so it becomes clear what I mean.
    There are basically two different things that combine to become game difficulty.

    […]

    For me at least, I had already understood your point. But you’re off topic, sorry. 😛

    The fact that some setup is hard or even almost impossible to win doesn’t make the AI smart. My original complaint was not that the game/AI is too easy as it is always possible to make it very hard with settings or own house rules. My complaint was that the AI is poor. These are two very different things.

    It’s not about game difficulty, it’s about AI not doing dumb things. The AI can do dumb things and win. It can do smart things and lose. Having trouble winning and facing a smart AI are two different things, even if they are related obviously. You’re only talking about having trouble winning.

    #256220

    Jolly Joker
    Member

    Is that worth a discussion? I mean, OF COURSE the AI does dumb things in a game which forces you to make difficult decisions.

    Anyone who plays a reasonably complex game and expects a competent non-blundering AI is expecting too much, because, simply spoken, it would mean, you have to explain to a person who ONLY does what you have told them, not more, how to play the game. There will always be situations when the AI is simply lost, because there are no proper instructions how to proceed.
    It’s worse – you wouldn’t actually WANT to play a game that allows something like that.

    The AI’s purpose isn’t to play the game like a human player would. The AI is just supposed to create the impression of resistance to overcome.

    #256221

    juhah
    Member

    I think it is worth talking about because AIs don’t need to be very stupid. AoW has an AI that seems dumber than many others, so it could well be improved, likely easily even. That’s one thing. Also instead of writing simple if then else algorithms games could likely use neural networks and let the AI learn from playing with humans. That of course is not easy to implement likely, but could offer a much more human-like opposition. I think it’s pretty astonishing that AI in strategy games hasn’t developed much in the past 20 years.

    #256222

    Hiliadan
    Member

    I think it’s pretty astonishing that AI in strategy games hasn’t developed much in the past 20 years.

    Not that much.
    If you do an analysis of all the posts in the forum, I would expect that of those talking about the difficulty of the game, 90% are about it being too difficult and 10% about being too easy.
    Then consider that maybe only 20% of players ever post on the forum about difficulty (and I think I’m very optimistic), it means that about 80% are happy enough about the difficulty, 20%*90% = 18% thinks it’s too hard and 2% thinks it’s too easy. And then among these 2%, most will conclude that they want to play against human players and start PBEM and live MP. Personally, I’m not asking Triumph to improve the AI but rather improve the PBEM experience.

    EDIT: and overall the AI is not that stupid

    Even if my figures are wrong, I think the broad picture is still that only a very tiny majority of players want better AI. So why invest in it when the rewards will be very low and most people are happy with the current situation?

    • This reply was modified 4 years, 9 months ago by  Hiliadan.
    #256224

    Jolly Joker
    Member

    I also think that tbe AI is QUITE good compared with others.

    #256226

    Motasa
    Member

    only a very tiny majority of players want better AI

    *cough* *cough* Civilization V and VI *cough* *cough*

    I think this is not the case. Maybe it’s just a vocal minority that speaks about bad AI, but I find it hard to believe the silent majority is OK with the state of most TBS AI.

    Luckily AoW III hasn’t arbitrary systems to make the game more strategically “deep” or “complex”, because most of the time it makes the game just easier against AI, who don’t seem to cope around those systems. Take the one unit per hex rule in Civilization V and VI, the AI is so atrocious in handling it, waging war isn’t more challenging or more tactical than in Civilization IV, I’d argue it broke these games.

    The main problem I have with the AI in this game, aside from the occasional dumb actions
    (attacking a treasure site with half a stack of units) and frequent silly quirks (like dispelling Blessed from my unit’s every other turn), is that its behaviour becomes stagnant sooner or later during play. In many of my plays in Age of Wonders III the roamers and neutrals seem to do a better job in harassing me than my AI opponents. It baffles me when an AI player becomes overly passive and inactive; turtling around their capital with their best troops and heroes, and not daring to go for a major assault.

    #256227

    Gloweye
    Member

    In many of my plays in Age of Wonders III the roamers and neutrals seem to do a better job in harassing me than my AI opponents.

    And it’s the same code that makes their decisions.

    It baffles me when an AI player becomes overly passive and inactive; turtling around their capital with their best troops and heroes, and not daring to go for a major assault.

    Because it can calculate, and thinks it has a snowball’s chance in hell of actually beating your armies in the field, or even in offensive siege.

    Really, I think the AoW3 AI, primarily combat AI, but also strategic, is one of the best I’ve played around. I didn’t play Civ 6 or EL myself, but from people who play both a lot, I hear AoW3 is the better AI out there.

    #256230

    Motasa
    Member

    I hear AoW3 is the better AI out there.

    It definitely is, in my humble opinion. Compared to Civ and EL, this AI actually “knows” how to play the game (aside from being bafflingly stupid at casting spells, but that is besides the point). My only complaint is that for a game focused on waging war and moving your armies around a strategic map with the goal of conquering all other players, the AI can be predictably passive or be really lousy in doing so adequately. But even if conquering isn’t the only solution in winning, with a Seals Victory the AI seems to trigger their focus to them very late or tries to counter me only once I have 2/3 of the necessary Seal charges for a win.

    For me, a more active and aggressive AI, even if it ultimately fails in its assaults and pushes, would be far more engaging than seeing it sit there, waiting till I’ve mustered enough forces to eradicate him/her from existence. Or see it approaching, only to hold of at the last moment. Earlier skirmishes, more armies and army movements and opening up multiple fronts would make the game much more enjoyable. Now it’s roughly 1/3 roaming and clearing the independent cities, structures and treasure sites in the area; 1/3 mustering forces to start encroaching on nearby enemy territory; 1/3 finding the Capital, ignore everything else, because the leader is waiting conveniently for me there for one/two large battle(s).

    a snowball’s chance in hell

    I like this expression. I like it a lot. Since I feel that the game often shifts in a power race, where I or another AI at the opposite side of the map gets too little interference from neighbouring AI-players and grows out to an unstoppable force that then steamrolls around the map. This snowball effect is something I see too often and more aggressive AI behaviour would be a step in preventing this from happening. Must say I enjoy it sometimes when such clashes happen on bigger maps, where the AI gets really good in keeping up Age of Deception and I really have to work to get things going in my favor again. 😀

    • This reply was modified 4 years, 9 months ago by  Motasa.
    #256232

    Fogcrow
    Member

    regarding AI, I have to agree that you can´t take anything but emperor seriously, thats because the lesser AI levels not only have less morale buff, they´re also not allowed to use some high end spells, and squires can´t even attack with multiple stacks at the same time.(there might be more limitations in place I don´t know about)

    The AIs skill at playing classes differs greatly, my estimation would be that its strongest/best as warlord, does a decent job as theo or dread, a little weaker as druid and sorc, plays rogue like a warlord, instead of bringing out the classes true strengh of wrecking an opponents economy by spamming him/her with city debuffs, and completely fails at necro, for various reasons.

    I´m not a modder myself(nor a mod user) so I´m not sure, but iirc it has some vital research and casting priorities wrong, but that can be modded…has anyone who really understands the game done/attempted a smart AI mod?

    well, on a different note: I´ve played chess against that AI that can´t be beaten, the best one can do against it is reach a draw(not me, a guy who´s far better xD) well…it just wasn´t fun…My guess is the AI isn´t designed to be as good as it can possibly get, its designed to be fun to play against. And since its always the same AI, they took into account the campaign scenarios, which are absolutely awsome btw, where it starts with a big advantage.

    #256233

    Fogcrow
    Member

    The toughest fight ever for me against the AI was a certain campaigns final misson, where I had to face it with just my leader, no heroes…so to have a more challenging, but not nessesarily more fun game against the AI, I´d lower the hero number.

    • This reply was modified 4 years, 9 months ago by  Fogcrow.
    #256235

    Motasa
    Member

    has anyone who really understands the game done/attempted a smart AI mod?

    I have meddled with AI priorities for spells, skills, city buildings, unit’s active and passive abilities and racial governance. Increased their sneaky healing and defense-versus-neutrals bonuses. But it is such a subtle change, that I doubt it will perform miracles.

    But I overhauled many things in the game and stumbled on things to help the AI play better, albeit marginal improvements. I added a hero skill for all classes that gives a passive experience boost for itself and units in its army, which I myself won’t pick, but the AI has “very high” priority for. So now I see heroes around midgame that can keep up in levels to my heroes (and I’m a player that doesn’t bother about min-maxing and exploiting experience gain in tactical combat).

    Concerning spells: I, for example, delayed research of the Disjunction spell a bit and its cost on every spell to prevent the AI spamming it like crazy, yet increased its effectiveness a little to counter the ease of reinforcing spells. I also removed the Dispel spell from the spellbook (heroes with this hero upgrade and Support units with this ability are aplenty). I reduced the obstacle damage of the fireball spell, since the AI calculated more often than not that a wooden gate at full health is a more viable target than my trebuchet.

    Another example: I also boosted the effect of expensive buildings like the Hospital, giving +5 hp to produced units. The AI almost always builds all possible buildings in its Metropolises, so this buffs the AI more, since I can’t afford it most of the time.

    Possible, there are more ways to affect the behaviour of the AI and alter its calculations that influence its actions even more, but to be honest, I doubt it.

    My guess is the AI isn´t designed to be as good as it can possibly get

    Could be. My guess is that the game in question, be it AoW III, Civ, EL, Fallen Enchantress, Total War or Europa Universalis, has just too many moving parts and variables that influence certain parameters and outcomes, that this might interfere with the goal the AI is programmed to achieved. Maybe it is that difficult to make the AI aware of all the things affecting the game, or maybe it’s aware and this makes it passive.

    I’m not saying the AI should hinder a game design from the get go, but for some games I get the suspicion the AI might have been an afterthought in the design process of a game; apparent by a total lack of an AI capable of understanding basic game functions or some of it’s more advanced systems. For me that’s a deal breaker concerning games in this particular genre.

    #256236

    Jolly Joker
    Member

    I had requested an option to add a multiplicator for AI XP gain in the game settings – until we get that, modding in hero upgrades that do this is a pretty good idea! Thanks for that, I’ll explore this.

    #256237

    Hiliadan
    Member

    I have meddled with AI priorities for spells, skills, city buildings, unit’s active and passive abilities and racial governance. Increased their sneaky healing and defense-versus-neutrals bonuses. But it is such a subtle change, that I doubt it will perform miracles.

    That’s very interesting! Could we discuss the details? It’d be interesting to include some of your changes in the PBEM & Single Player balance mod to improve the experience for everyone, and maybe analyze with you what else we could modify to improve tactical AI for auto-combat in PBEM and for single player. Are you on Steam? Then you can add me there, or else we can exchange by email or on the Battlefield maybe?

    #256238

    Motasa
    Member

    Could we discuss the details? It’d be interesting to include some of your changes in the PBEM & Single Player balance mod to improve the experience for everyone

    Sure, I’ll add you on Steam. But as I said, since the changes I’ve been able to make are just minor (changing AI research and building Priority between very high, high, medium or low and BuildFrequency of units), I can’t say I’ve noticed a real improvement in the sensibility of AI behaviour and decision making. So don’t get your hopes up too quick. It is not to say that altering the priorities of buildings, units, research and hero upgrades is a waste of time or has zero effect, but as far as I’m aware it will only shift the order in which the AI prioritizes to build things and research things. What I’m trying to say is: it is hard to verify how much influence these small adjustments really have.

    modding in hero upgrades that do this is a pretty good idea! Thanks for that, I’ll explore this

    In the end, for a more noticeable difference in AI behaviour, I think exploring ideas of adding new hero upgrades or empire upgrades (which you can set to “Free for AI Players”, so the AI gets it right from the start) and messing with Racial Governance (the AI happiness boost gives it an edge in gaining higher tiers earlier) is really worth the effort. To come back to my earlier example: I see in the late game many AI units with +5 hp compared to my units, just by adding a health boost City Property to the Hospital, a building the AI builds far more often than a human player would. So by modding this simple addition to an existing building I use the quirk of the AI (to build every building and use it’s extra production and resources) to actually be beneficial for the AI, and it also gives me more incentive to build the required building chain to get a Hospital in my cities.

    #256241

    This is promising stuff.

    I do find it slightly ironic that this topic is on the same page as “King AI is kicking my butt…”

    Has anyone seen a way to alter the income variables of the AI?

    I think a good experiment would be to double the AI’s starting army and resources, but half it’s current income bonus.

    That would make the early game more interesting, as the AI would be more powerful, but would hopefully reduce the late game slog of smashing through dozens of AI armies.

    I wonder if an AI fight priority could be set the beeline for the player.

    It wouldn’t be a remotely fair AI but it could help to redistribute the challenge temporally (harder earlier game).

    #256244

    Motasa
    Member

    I think a good experiment would be to double the AI’s starting army and resources, but half it’s current income bonus.

    That would make the early game more interesting, as the AI would be more powerful, but would hopefully reduce the late game slog of smashing through dozens of AI armies.

    I’ve looked at this a long time ago, since it would indeed counter the early problems with certain settings of the AI players losing too many units against roamers and neutrals defending structures, especially with strong defenders. And it would alleviate late game (and really late game, for those playing beyond turn 100) problems where the AI has no ceiling to its progression. But as far as I can tell, there isn’t a way to exclusively give AI players a higher starting army value and/or more resources.

    I wonder if an AI fight priority could be set the beeline for the player.

    Although I’ve seen ways to alter the priority values in it’s calculations for tactical combat targets — which I haven’t messed with since the AI is fine in that regard as far as I’m concerned — I have yet to find settings in one of the many core rpk files that involves strategic map movement.

    #256246

    Bouh
    Member

    A very interesting topic !

    I’d like to add some things about the state of the AI that I think should be emphasized.

    First, it’s not a real AI but a program. Which means it won’t have any inovations in strategies or whatever. So after you played it for a time, you know all of its tricks and you will never be surprised, which is a big drawback in a war game against a good oponent. It also means that if the AI was able to beat you, it would beat you all the time, which would not be enjoyable for most players. Then you can see the problem if you take the place of the AI : you can only beat the player if you bring more stuff than he has. Another important point is about the sites : the AI cannot gain a leverage with them because it fights an equal oponent there unlike the player. Additional resources the AI gets are merely a compensation for the site clearing the player can do. Then the strategic level : as the AI, you must realize that you will never beat the player unless you can corner it with a lot more resources. And once the player gets an endgame army, you’ve basicaly lost, because he will only engage it when he is sure to win.

    So in practice an AI, to offer a decent challenge, should play like an excellent rogue player. There probably are some things that could be done to partialy achieve this, but its not easy. And either the algorithm risks to be to basic, and the player would find a counter, or would have some randomness and be unreliable.

    So the current strategy the AI use is the best available to her I think, eventhough it works best with production classes.

    Second, the neural network : these things are currently good at finding patterns, not creating strategies. I don’t think this kind of technology can do anything for a complex strategy game currently. I don’t even think they are the way to go for this problem.

    I have meddled with AI priorities for spells, skills, city buildings, unit’s active and passive abilities and racial governance. Increased their sneaky healing and defense-versus-neutrals bonuses. But it is such a subtle change, that I doubt it will perform miracles.

    I think this is a brilliant way to improve the AI, even if only slightly so. Probably the most ellegant way without access to AI intelligence routines directly.

    Although I’ve seen ways to alter the priority values in it’s calculations for tactical combat targets — which I haven’t messed with since the AI is fine in that regard as far as I’m concerned — I have yet to find settings in one of the many core rpk files that involves strategic map movement.

    Do you think the AI strategic decisions are accessible ? I’d love to work on that.

    #256248

    Motasa
    Member

    Do you think the AI strategic decisions are accessible ? I’d love to work on that.

    It is quite possible there are certain parameters and values hidden in the rpk files that are used by the AI to calculate their goals on the strategic map, but alas I have missed or overlooked them till this day. But even if such values are found, like those used by the AI in tactical combat (Title.rpk, settings, AoW global settings, tactical AI settings (way down the list)), it is very difficult to test the effects of slight changes in these values.

    I see a lot of obstacles in modding values that seem to affect the behaviour of the AI, but seem quite arbitrary. How would you change the existing Die to Unit Score Penalty of -30? According to the description the game devs gave this variable, it is a value the AI adds to target evaluation scores if the actor will die attacking a target unit. Will -35 have an improved effect in preventing kamikaze attacks or would -25 prevent the AI from ignoring your low health units? More importantly: how often would the AI behave differently from the standard -30? And how would this affect the game positively or adversely? The problem might be you’d trick yourself in thinking the computer controlled players behave differently, whilst the numerical change is ignorable; it has zero effect. You may attribute certain AI behaviour to the changes you’ve made, whilst something else is responsible for the behaviour. At worst, your efforts might turn against you and you’d end up with a poorer AI or even a broken one.

    In the hundreds of hours I dabbled with the modding tools I focused primarily on adding new stuff (buildings, skills, unit abilities and spells) and overhauling existing things (like racial governance, dwellings etc.). Changes I’ve made that I believe accommodate better AI have been just a minor part of my tinkering with this game. I don’t consider myself that good a modder and don’t have that much patience to test and experiment with values I barely understand the meaning of, even if the modding tools allowed for even more options in messing up the game, I think I’d leave it as it is 😀

    #256253

    Another important point is about the sites : the AI cannot gain a leverage with them because it fights an equal oponent there unlike the player.

    Actually Bouh that’s not true.

    On Emperor the AI gets a ‘secret’ defence boost against independent units. I can’t recall the number but I think it gets +2 defence and resistance.

    #256254

    Jolly Joker
    Member

    It’s 3 def and 2 res – moddable! Actually, a lot more of what has been said here isn’t true either.
    I’ve been dabbling a bit with things, and it is NOT true that the AI is handicapped, when you pick weak starting forces, simply because the AI isn’t getting weak starting forces, but instead average/normal – at least this is so on Emperor difficulty (I didn’t check the others).

    Meanwhile I had a pretty cool idea how to boost the AI a bit, and reliably so, and I will try to mod that next week or something.

    #256255

    Hiliadan
    Member

    http://age-of-wonders-3.wikia.com/wiki/AI

    when you pick weak starting forces, simply because the AI isn’t getting weak starting forces, but instead average/normal – at least this is so on Emperor difficulty (I didn’t check the others).

    Would it be a boost to Emperor’s starting army over the board? So Average would be Strong, Strong would be Very Strong, and Very Strong even more than that? That’s not listed above but maybe that’s the case.

    #256258

    Jolly Joker
    Member

    I checked King. and the effect is there as well. Also, the AI gets more with the other starting force options, too.
    I suppose, though, that it’s not “a better starting force option”, but a general bonus (like, with Necro).

    #256260

    Hiliadan
    Member

    Ok, would be good to see how much more it gets and if it works also for lower difficulty, I guess it doesn’t, and then update the Wikia. 🙂

    I suppose, though, that it’s not “a better starting force option”, but a general bonus (like, with Necro).

    What do you mean? You mean it’s not Weak -> Normal, Normal -> Strong, but rather a “flat” bonus like x1.3 of whatever value (Weak, Normal, etc.)?

    #256280

    Motasa
    Member

    , but rather a “flat” bonus like x1.3 of whatever value (Weak, Normal, etc.)

    The way I’ve gathered it is that weak, normal, strong and battle constitute a certain total unit value, and each unit in the stack you get at the start of a map has a value that fills up to that total unit value. If you choose to start a game with a “weak” starting army (which has a value of 160), you most likely end up with three irregulars or a combination of tier 1 units and a cavalry unit to explore. The AI seems to exceed this army value at higher difficulties, so there must be a multiplier value hidden somewhere, as it consistently gains not only more units than a human player, but also more often tier 2 or even a tier 3 unit. Frostlings are screwed in that matter, since their cavalry unit is such a unique case, the AI controller Frostling players may start with just a Mammoth Rider and a single Ice Scaper on lower difficulties and on higher difficulties they sometimes are still left with less units to work with. Something is messed with those Frostlings… still searching for the error though.

    it is NOT true that the AI is handicapped

    Strong defenders and Adventure mode (which I only played a couple of times) has me always lose a couple of opponents against structure defenders, roaming neutrals and independent cities within 10 turns.

    Meanwhile I had a pretty cool idea how to boost the AI a bit, and reliably so, and I will try to mod that next week or something.

    Maybe you want to share your findings once you’ve tested your changes out? Eager to hear about any way to improve the game.

    #256285

    Jolly Joker
    Member

    Okay, I spent the evening working on it, and, well, I got it working… 🙂

    I had modded in your – Motasa’s – idea of a hero upgrade (I made a 100% XP gain stack upgrade for 1 UP with very high AI interest) and disappointedly registered that the AI wouldn’t pick it on the first level-up. You see, I play with a mod that makes it a lot more difficult to level HEROES (not units) up, and that means, the AI has a problem to develop the heroes, which means, I needed a way to make sure they start with it.

    And that is actually very easy. I have now an “AI-Mod” that currently does 2 things:
    1) I changed the cheat bonus from +3 def and +2 Res against neutrals to +5/+3 and

    2) I made 3 AI-specs, called AI-Explorer, AI-Expander, AI-Partisan; when you set up a game, make sure your AI-players pick ONE of those, because each one is like the original, but in addition will hand out a second free Empire Upgrade (like Folk Hero, for example) which doubles general XP gain (doubling it may seem a bit harsh, but for my mods it’s ok). I got it working, but it will take some time to find out what the AI makes of it.

    3 and following: clearly this is the way to go: AI specialisations. Obviously, you can do everything you like, put it in a spec and give it to the AI. This includes, for example an Empire Upgrade that will give +5 HP to units produced in a town with a Hospital (if you don’t want the human player to be able to get that as well) – but also doesn’t EXCLUDE the option to tie it to the Hospital, so that humans may be tempted (or forced) to build them as well.

    You can also make class-otiented specs. For example, Necromancer is clearly a class the AI is playing sub-optimal, and a “Necromancer-AI-Boost specialisation” may help them.

    A LAST NOTE: I remember people complaining about the AI swarming them with waves of Manticore Riders and stuff like that. Well, actually, that’s technically impossible. There is a percentage value that forces the AI to stop building a unit, when that percentage (of the whole army) is esceeded. The default value is 20% (.2), which means, as soon as the AI hits the 20% mark for a specific unit it stops building. If the AI had 100 units, it couldn’t have more than 20 Manticore Riders. This value is of course moddable, but actually, I think that .2 is fairly low and shouldn’t be set lower.

    #256286

    juhah
    Member

    I’ve played my emperor necromancer gsme now to turn 80. 3 AIs destroyed and I’m in the lead now. It’s been fun and a bit more challenging compared to king. I haven’t been able to get all the empire quests done myself being the biggest difference maybe.

    Still as said here before by someone the AI is very passive, or should I say passive aggressive. 🙂 The three AIs I destroyed all declared war on me almost as soon as I found them and they were at that stage stronger than me, especially one of them was the biggest of all AIs. I still wonder why they declare war but then do nothing but sit and wait for me to attack. Doesn’t make any sense, it’s as if declaring or not declaring war is a completely separate algorithm from what to actually do with the troops on the map. If they are stronger the algorithm declares war but then there’s a huge disconnect to what they do with the troops (which is sit and wait). All of these had superior forces and could have at least taken a few cities from me but instead they did nothing. They also sent me messages saying “we’re coming for you” which further ruins immersion as they most definitely are not coming for me. 🙂

    I can’t agree that AoW AI would be among the best as it is so passive. Civ and others at least send out troops and attack if they are stronger and declare war even if they are not very smart otherwise.

Viewing 30 posts - 31 through 60 (of 163 total)

You must be logged in to reply to this topic.