1. We have added a Gift Upgrades feature that allows you to gift an account upgrade to another member, just in time for the holiday season. You can see the gift option when going to the Account Upgrades screen, or on any user profile screen.
    Dismiss Notice

The difference between a System that is geared to fit the AI...

Discussion in 'CivBE - Ideas and Suggestions' started by Ryika, Jun 22, 2015.

  1. Ryika

    Ryika Lazy Wannabe Artista

    Joined:
    Aug 30, 2013
    Messages:
    9,395
    ...and one that is designed around the idea (or just flawed enough) that the player needs sub-fields where he can dominate the AI, even on the highest difficulty. Little safe spaces so to speak, to which he can flee when the AI bullies him too much on the open battlefield.

    This thread contains a lot of mod-talk, but the core of it is really about game-design and my interest in your opinions about the AI, hopefully from a point of view that some haven't really considered yet, so don't shy away just yet!

    Anyway. I'm currently working on a mod that, besides doing a lot of other stuff, changes the way research works to be less about Academies and Flat Bonuses but instead goes hand in hand with general empire development. New techs would unlock new, much stronger Buildings that have a much higher production cost to create a scenario where the player has to balance growth, production and, because of increased Maintenance, Energy all at once, while trying to also push forward in science and culture.

    My idea was really to create an experience where empire-development is at the core of what the player does, so I was rather astonished when I first started to do some serious play-testing. The results on the AI were INSANE. On Apollo I had not the slightest chance of keeping up in science, empire development or really anything else. The AI just steamrolled out of control, later wrecked my with armies that almost made me whimper in fear and I felt completely overwhelmed by what was going on.

    Thinking about it... yeah, it should have been obvious to me that this would happen. AI-Bonuses go straight into stuff that help them develop their empire, so the lack of "intelligence" of doing so is less important. If I then focus the game on direct empire development, then that's obviously a direct buff for the AI. The amount of strength it gained is still astonishing to me though.

    I think this experience could be translated into many other fields of the game. Would wars really be more fun if the terrain was redesigned to not cripple the AI that much and if its "carpet of doom" was scaled down to more reasonable armies to make up for that? Would "strong" mechanics in general still feel "strong" if the AI uses them in the same way the player does?

    This really made me think about the bigger picture of having an "abusable" AI. I pretty much removed or reworked all "Totally OP"-stuff that the AI doesn't know how to use properly and what I was left with was an opponent that was, well... unstoppable - at least on my level of experience with the game and the mod. I wonder if that's really a good thing or a bad thing. Not being able to use "player-cheats" like specialist slots in Civ 5 and Academies in Beyond Earth and instead having only the "direct" playing-field - the one where the AI gets its bonuses - to work with, felt very odd. This experience actually made me reconsider my stance that "An AI should be as good as reasonably possible" - and led me to believe that there's some benefit to be had from having an AI that is not "perfectly capable" in all fields. An AI that has obvious flaws and having mechanics in the game that favor the player over the AI and give him the edge to overcome its bonuses in another area.

    It's probably obvious that the AI in BE is less capable than most of us think it should be (or at least most to whom AI-strength matters). But how far is too far? Are "Player-Cheats" necessary? And - this question is mainly addressing players who like to have a challenge - would it increase the fun you had with the game if you had to play against an AI that gets bonuses in every field that you have access to and the only thing that could make you win against it is playing "flawlessly"?

    Leave your opinions below. And don't forget to like, comment, subscribe.
     
    Guide-on likes this.
  2. XplosiveLun

    XplosiveLun A humble village

    Joined:
    Apr 24, 2015
    Messages:
    234
    Location:
    Hong Kong, China
    I think it depends on "how developers want players to do in their game". From playing Civ, I could tell it is "one that is designed around the idea" and tried hard to gear the AI to play well.

    In my opinion, developers should give themselves a clear vision of how a game should be. Like Wide playstyle should dominate Tall playstyle because Wide players put more effort into their empire. So that it should have more in return. Or Wide should be equal to Tall because everyone is equal in game, it is just a different approach towards victories. It is just personal preference (or marketing?).

    Whether system should be geared to fit AI? I think it depends on how the game should be played and the difficulty to make AI play well. In terms of programming difficulty (I know little about that but it is not as easy as people think) , I think a weak AI with bonuses is more achievable. It can substitute a smart AI and still fulfill the same purpose -- Make the game harder or even unbeatable; giving players a challenge.

    "Would wars really be more fun if the terrain was redesigned to not cripple the AI that much?". Nope, I don't like that. I prefer varieties. A problem should have more than one solution to accomplish. Even AI cannot use solutions well, players should still be given the opportunities. If I want to conquer enemies, I could use military, covert ops or via diplomacy. So that every game is be unique and refreshing to me.

    Although I would like to see a AI playing good, that is impractical and probably resource-hungry. I am okay to stupid AI as long as they process a threat. If you are not satisfied, go playing with real human.
     
  3. GAGA Extrem

    GAGA Extrem Emperor

    Joined:
    Dec 24, 2008
    Messages:
    1,577
    Gender:
    Male
    I strongly believe that the AI should play by the same rules as the player in a strategy game - and vice versa. That is, unless it is a game that deliberately relies on assymetry.

    So whenever I mod for Paradox games, I make sure that the AI will not receive any mechanical advantages over the player. I am willing to give it some minor benefits, but most of my tweaks are (more or less) just things to force the AI to act like a player.

    For example, in Victoria 2 the AI is unable to use National Foci to promote pops (which is important because your income and research points are strongly tied to that). Instead of just giving the AI a big bonus to income and RP to compensate for that inability, I instead modded the promotion logic so that the AI pops will auto-promote to the needed % over time, creating a (mostly) similiar result the player would have if he played that country normally. Once I reach this "parity" of using the mechanics, I start to balance things to add choise to the available options.

    Because the trick is (imho) that mechanics have to offer choise and that this choise cannot just be a binary one between a "more powerful" and a "less powerful" option. It has to either involve risk/reward ("I will be weaker now, but if my units can hold themselves for 50 turns I will get a new model that allows me to take the initiative in this war") OR it has to be a choise between different approaches towards the same goal (like specialist vs. cottage economy in CIV4). In addition to that, mechanics should always have an option for counterplay. If you cannot counter the strategy of your enemy, you can only rely on superior numbers (if your enemy plays on a similiar skill level). Which in BE would probably mean that the side with better terrain / starting conditions auto-wins.

    The question if wars with less units would be more fun depends mostly on how the player wants to win. You can either outwit your enemy and win by countering his strategy and/or tactic, or you overwhelm him with brute force. Both are legimate ways to win - and frankly, while I love luring the enemy into my anti-tank guns in Combat Mission, I have also no problem with winning by just using a swarm of T-34s to break through his defence and rout his core defenders.

    Going back to your problem: In your case I'd try to find a good balance where the benefits of higher difficulty levels compensate for the lack of AI ability instead of just giving it raw number power. If you have an unstoppable AI the balance is obviously tilted way too far towards the AI. Which reminds me that you also have to define the role of randomness for the whole thing. How much should a tundra/wildlands vs. desert floodland/grassland start impact the chances to win?

    ...and I'd like to remind you of my PM in regards to modding that I wrote you a few months ago. :p
     
  4. Gorbles

    Gorbles Load Balanced

    Joined:
    Nov 24, 2014
    Messages:
    3,816
    Location:
    UK
    Welcome to AI development :D

    Fascinating thread, and good work on building the mod base to experiment with this kind of stuff. See, this is the stuff I live for, and find truly interesting to talk about. That said, AI is one of my weakest areas when it comes to games development, so any gaps in my knowledge are sure to crop up here!

    Basically, the AI, like all computers, simulated or otherwise, excel at the simple operations. The linear ones, the ones that perform simple operations in a decent O-time, etc. It can stack these up infinitely. Heck, even if your algorithm for slicing bread is like O(n log n squared) it'll chug through it eventually. Basically, if it is a step of logical steps with a defined end goal, the computer gets there.

    Managing cities? Yup.
    Managing yields per tile? Yup.
    Optimising those yields, assuming the resources have a weight (in BE's case, "flavour") assigned to them? Yup.​

    So the AI in general will crush these kinds of operations with a ruthless efficiency that would make any notable dictator proud.

    However, in any game, there is always a human element. You know this - you know what the AI in BE excels at, you said it yourself. But the thing that it can't account for is this knack of prediction, or gut feeling, or the like. This cannot be factored into programming, despite the massive sums invested in the relative scientific fields each year. The amount of raw computing power required to make a computer make an illogical leap (outside of very well-coded trickery) is insane.

    Can an AI therefore ever make consistently-good judgements like a human player? Nope. This brings us to the primary rule which governs most AI development in video games:

    Given equal terms, a human player will always beat an AI. Always.

    Therefore, some "cheats" (or in the case you've worked out here, simply strengthening the aspects of design that the AI excels at) are necessary in order for a semblance of equal play. It must be noted, that human vs. human is always the end goal here - especially when balancing the game (if necessary). Balancing around the AI generally just involves tweaking the AI to suit given the inherent advantages provided. The most you'll get is buffing player strategies so that they all feel strong / fun / whatever your primary design directive is (and it isn't always "fun").

    But what cheats are best? It somewhat depends on the genre, but you can divide it into basic types:

    1. Resource cheats. This is the simplest, and generally seen as the "easy" way out. But it's reliable enough, especially in genres like RTS (or even 4x). Finetuning is expected mainly around "rush" scenarios, in both not steamrolling the human players or being too weak to prevent being rushed by the human players.

    2. Awareness cheats. If the game has "fog of war", or something similar? The AI ignores it. If build orders are hidden? The AI can see it. Metagame awareness, almost. What humans have to guess at (and normally get quite skilled at predicting), the AI knows from the offset. This has the potential to seem the most "realistic", but also if overused can look the most cheap or "easy" way out (and you'll see player complaints about the AI pulling off impossible manouvres simply by dint of this planned awareness).

    Note: this can come with performance overhead, especially in real-time games. Often doing something as simple as forcing the AI to respect fog of war massively cuts down on search algorithms, pathfinding, etc.

    3. Rule cheats. Possibly the most stupid-sounding, but also used. Actually heavily-used in Single Player scenarios to create an unfamiliar situation for the human player (and, y'know, make it a vague challenge). You can only build three Swordsmen with your three Iron? The AI can build ten thousand, and doesn't care about it. Etc, and so on.​

    Practically, as everyone here already knows or has probably guessed, a lot of games use a combination of all three. My main AI modding experience comes from Dawn of War / Relic Entertainment's RTS games in general (they use a similar language throughout) as well as associated research through my uni years and in the years since. I really don't know much about how Firaxis do it, so I'm looking forward to what you bring, Ryika, in your modding efforts.

    Also, funny stories about how the AI smashed you are always welcome. Unintended consequences of programming is always hella fun to talk about :D
     
  5. Galgus

    Galgus Emperor

    Joined:
    Aug 22, 2012
    Messages:
    1,695
    The most important part of design in a Civ game to me is allowing alternate tactics and paths to victory.

    There just shouldn't be a single tile to spam to pull ahead or any other single optimal path.

    Making a good AI is largely about it pursuing a viable path to victory, which leaves three options.

    1. Make the AI take the one best way to victory, showcasing bad balance and bland gameplay.

    2. Make the AI act in character regardless of what is optimal, requiring heavy cheats to keep up.

    3. Balance multiple paths to victory and make sure that each AI has a viable path.
    ____________________________________

    In all cases, the game design comes first with the AI being slotted in to function with it second.
     
  6. Ryika

    Ryika Lazy Wannabe Artista

    Joined:
    Aug 30, 2013
    Messages:
    9,395
    Alright, I'll spare everyone the scrolling of quoting all three of you. :D

    @XplosiveLun: Yeah, I agree that Developers should follow their vision of the game and this threat is by no means meant as an evaluation of their decisions - it's more a thought-experiment for myself (and hopefully others, who may find it interesting), because I always said to myself: "It would work best if <x was changed to do y>." - very well knowing that a lot of the ideals are unrealistic and inefficient in terms of time- and money invested vs. actual positive effect. The idea of this thread is not so much defining how "things need to be", or "how things should be", but to find out what others think about the advantages and disadvantages of different systems and if the situation I ran into would make other people who may think like I reconsider their position a bit.

    @Gaga: It's funny, you and I seem to be in a totally different mindset when it comes to how playing against an AI should work overall. :D It's interesting though, I'm basically doing the opposite thing in the mod I'm working on and "forcing" the player to play the way the Ai plays by making it the efficient option. I highly doubt there's a way to really get rid of the AI-bonuses in Civ (and if it is, then it's completely out of my league). I'm wondering about the original question though: If the AI got bonuses but the player could use his superior "intelligence" - without having to "abuse other systems, just by playing more efficient to counter the AI-bonuses, to find the best long-term strategies to get ahead, do you think you would like or dislike that? (Not saying that this is any more realistic to achieve, just out of curiosity).

    And sure, I remember that PM. :D The mod is still a huge mess though, with very basic stuff even implemented - and I scrapped a lot of the original ideas and went into a very different direction. If you still want to playtest, feel free to send a PM and I'll upload it somewhere once I've patched up the gaping wounds that currently exist. :)

    @Gorb: Now, that's an interesting post! Not to say that the other posts weren't interesting, but if I had a bonus point to award, then you'd get it. :p Thanks for the input and... yeah, that makes a lot of sense.

    It's interesting how BE and Civ 5 basically goes all-out with the "Resource cheats" to create difficulty settings, only with some "Rule cheats" put in for starting-bonuses and mechanics that didn't seem to work properly otherwise (for example being allowed to move a unit after buying it with gold).

    At the same time there don't seem to be any awareness-cheats as far as I know.

    Still curious about the original question though: Do you think designing a game in a way that forces a player into more symmetric gameplay with the AI is more of a yay or nay?

    No need to look forward for my mod though, it's in no way meant to become an AI-Mod - i more or less stumbled about that while I was playing around with gameplay elements. :D

    Still, yeah... getting pwned by the AI was kind of fun. I actually went back a few times and tried different strategies, but just couldn't manage to even stabilize. Now, to be fair, that start wasn't exactly a "good" one and since then I managed to somewhat stabilize for a while on other maps to find out that the AI stagnates somewhere in the mid-game (which, looking at the situation the AI finds itself in, can probably be addressed with just some edited Flavors/Grand Strategy Settings)... but it's really cool to see how just some changes to fundamental concepts can completely re-shape the core of other sub-systems in the game. ^^

    @Galgus: Yeah, that's probably right, although of course the interesting thing is that the "core" idea of the design I was aiming for met up with what the AI is good at, not because that was my goal, but because the general idea sounded cool to me. It created this interesting situation where suddenly I realized that I had to directly compete with the AI instead of accepting that the AI has its stat bonuses and I would be able to just game it by building a ton of academies.
     
  7. Gorbles

    Gorbles Load Balanced

    Joined:
    Nov 24, 2014
    Messages:
    3,816
    Location:
    UK
    Ahh, sorry, got a bit distracted, heh. Stuff like this always makes me walloftext, and I'm never sure if I'm over-explaining or what. Glad you appreciated it!

    tl;dr: it is a nay.

    If you constrain the player, they will simply reach the limits of design faster. It is much better to design the game for the player, to the greatest extent, and do whatever you have to do to allow the AI to catch up.

    This is in fact another reasoning for giving the AI cheats to compensate, because truly distinguished games design allows players to enjoy games in ways that the developers don't forsee. If you're building an AI to complement, given a lack of applicable quantum physics and / or the ability to write self-learning code that won't evolve into Skynet . . . your AI will have limitations. Your potential players may not. So you build the AI with that in mind.

    Players can break your game, or your mod, nomatter how much effort you put into it. An AI can never break anything that you haven't already broken in some small part of its code somewhere :)
     
  8. Galgus

    Galgus Emperor

    Joined:
    Aug 22, 2012
    Messages:
    1,695
    The ultimate goal is for the AI to have the illusion of being a human player - to compete with them and have their own quirks as players do.

    Personally I think designing various paths to victory for different AIs to take that fit their personality is an easier and less problematic way to do this than hidden bonuses.

    Ultimately if the Arshia Kishk AI is a warmonger, there should be a viable warmonger playstyle with sufficient sub-variations for players that the AI will follow.

    No AIs should be programmed to pick sub-optimal options or pursue dead-end strategies: either the AI or the viability of the option should change.

    If the AI simply knew how to manage their cities properly they would be far more dangerous.
     
  9. GAGA Extrem

    GAGA Extrem Emperor

    Joined:
    Dec 24, 2008
    Messages:
    1,577
    Gender:
    Male
    I am not really sure if that is actually a big difference in the end. The result is that player and AI follow the same rules. I just find it easier to tweak the AI, because players have the tendency trying to find ways to get around limitations set by the game (and they have vastly different playstiles which you have to take into account).
    ...so in the mod I have to use carpets of doom because these mountains and lakes block my path? :confused: :D

    Well, I am sure if you would teach the AI the current basic BE strategies ("rush Academies, then go for Affinity" for a peaceful win) you could cut a lot of them. Thing is (as I understand the way the game handles it) that the AI decision making is based on weighted modifiers (a bit like the AI in Vicky 2 does tech selection), so it seems hard to program an actual fixed tech path. And even then the AI is always at a disadvantage because it won't get AU/Artists/Tectonic Scanner/Laboratory (or the Colonist rush equivalent).

    I gues it depends on the actual numbers. As long as the AI advantage is somewhat realistic I can live with it. But if it's obvious cheating or forces me to completely alter my strategy, I probably won't like it.

    An old RTS game called "Sudden Strike" comes to mind here. It had the tendency to stack the odds incredibly harsh against the player. Like: If you played Stalingrad as the Russians, the Germans had 50 bombers and 50 tanks while you had 8 AA batteries, but if you played it as the Germans you had 12 bombers and 20 tanks vs. 20 AA batteries. That was just silly.

    Or you can take CIV5 as an example: I refuse to play on Immortal, despite having a ~90% winrate on Emperor (those missing 10% being the AI hyper-rushing me early on), because I find the starting bonus for the AI too much. The gap is so big that it more or less destroys the early game content for my playstile (peaceful builder). It tilts the balance way too far in favor of the AI during the early game, thus giving me no chance to grab an early wonder and forcing me to play catchup for half the game. If the AI bonus would scale over time (like AI gets a free tech each era, gets 10% all yields when reaching industrial, etc.) I probably wouldn't mind, because then I wouldn't be forced to completely change my playstile or feel like a worthless underdog for the whole game. I'd just have to play efficiently - which is perfectly fine, because that's what the early game on Emperor feels like: I probably won't be the first to found a Pantheon or Religion, but I can be #2 or #3. I (usually) can't get the Hanging Gardens AND the Oracle, but I will at least grab one of them if I focus on it (the Great Library, though, is already tilted a bit too much in favor of the AI for my taste).
     
  10. Ryika

    Ryika Lazy Wannabe Artista

    Joined:
    Aug 30, 2013
    Messages:
    9,395
    Yes, the Flavor-System is what I'm working with, but it only works semi-efficient and many cases just aren't covered. From what I can see it's for example completely impossible to use the system to tell the AI to favor growth until X population in a city and then favor science - which would be needed to make it work semi-efficient with Academies. Of course, there are many ways around it (the easiest being to just make every Grand Strategy rate Science/Academies extremely highly), but they would still require AI-bonuses to make up for the fact that it would settle a city in tundra and just not get enough pop. :lol:

    Tech path is actually interesting, a few days I was thinking about how to make my Custom Sponsor go down the Harmony-Route consistently. Custom Flavors do unfortunately not seem to work at all and using the existing Affinity-Flavors to create "beelines" would require a full-rework of all flavors and probably limit each sponsor to one affinity - if it at all effects their choices. I later thought about creating a lua-script that just reverts the tech-choices the AI makes and just gives them something I want them to have. But that would be a really stupid idea. :D


    That reminds me of an old Asterix-"Strategy"-Game for PlayStation that I played when I was a kid. It's ... rather similar to the board game risk actually. But the Ai literally begins with like... 10 Soldiers on any field except for your home base, gets more reinforcements and just does random nonsense so it's really easy to win. Obviously, the AI was meant to be stupid and the scenario only makes sense with Rome holding most of the land... but turning it around and thinking that this is how far they had to go to make the AI actually somewhat capable at seeming like it could in theory maybe be a challenge sounds hilarious to me. :D


    Yeah, I actually feel the same. The "front loaded" bonuses the Ai get don't feel too good, but on the other hand on lower difficulties the Ai just falls apart later on. Fortunately modders actually solved many of these issues for Civ 5. Should probably go back and look at their mods to check how much was only possible thanks to dll-stuff and how much I could in theory incorporate if I was to actively tweak the Ai to suit my mod.
     
  11. Acken

    Acken Deity

    Joined:
    Sep 13, 2013
    Messages:
    5,635
    Gender:
    Male
    Location:
    QC, Canada
    I certainly agree that in a single player strategy game the game systems have to be designed with AI programing in mind.

    However one shouldn't fall into the trap of "forcing" (or skewing the flavor too much) the AI down a single optimal playstyle. Variation and different should be kept and if there are options way too strong that someone would feel forced to make the AI take it I suggest to balance it instead. While optimal stuff is probably not completely avoidable this is where the difficulty level get into play. They leverage the AI so that unoptimal diversification can still survive and provide a challenge.
    An example would be rationalism in Civ5. The deity AI can definetely be a challenge if it picks it right away but instead of boringly make the AI always take it a better idea for the game would be to tone it down (or buff the rest).

    If the AI works well enough with your system you may even reach a point where you only need small bonus increments from one difficulty to another. I guess you could view AI strength as a function of h*B where S is the AI ability to perform optimally where S=1 is optimal and B the AI bonuses. To reach a specific value of this function, if S << 1 you need to compensate by a way bigger value of B and vice versa.

    Front loaded bonus are an interesting subject. In my opinion they are necessary when B in the previous point reaches critical values. There is definetely a point where increasing raw bonuses will lead to problems making the AI feal way to far from the norm and frustrating rather than challenging. This is where frontloaded bonuses get into play. They allow to give you an initial bonus. The strength of the AI would become like h*B + B_0. And like a linear function this allows to give the AI more strength without having to increase the slope.
    But if a system already achieves fun challenging levels of difficulty without it then they aren't necessary.
     
  12. Jyrgunkarrd

    Jyrgunkarrd Chieftain

    Joined:
    Jun 22, 2015
    Messages:
    6
    The weakness of the AI since the implementation of 1UPT in the franchise has never been it's use of resources - it's been it's lack of respect for terrain. Honestly, I've largely stopped caring in Civ about my tech level in relation to any AI opponent, because I know that they will death-march every stack they have into choke points 1-by-1, allowing my to block, surround and destroy even units 1-2 tech tiers above my own with impunity. I think I score roughly 100-1 kill ratios these days in both BE and CivV, which renders any sort of economic advantage moot. No amount of bonus gold or beakers is going to help you if you lose that many troops for every stack you manage to kill.

    Fixing this would require granting the AI much more complex pathfinding behavior, but the trade-off for that would probably be bogging down turns as the computer crunches it's various routing options.
     
  13. GAGA Extrem

    GAGA Extrem Emperor

    Joined:
    Dec 24, 2008
    Messages:
    1,577
    Gender:
    Male
    The dirty and easy fix is to just increase the map size (and scale everything like min city distance accordingly). If you ever play the American Civil War scenario in CIV5, you will see that the CIV5 AI is actually semi-competent if it has enough space to maneuver.

    Wouldn't it still be possible to use the flavor system to guide the AI? Iirc every tech, wonder and buildings gets flavor values assigned, so if you remove the science one from all but the essential Academy spam techs (Chemistry, Ecology, Genetics, Cognition), you should be able to guide the AI down a certain tech path. Going further, you could even repurpose some flavor values to teach the AI certain strategies. Like: As it is, FLAVOR_WONDER is pretty useless. Let's use it instead to teach this new sponsor AI about the ideal tech path for Harmony (or whatever else you can think about where it needs guidance).
     
  14. Casworon

    Casworon Prince

    Joined:
    Aug 9, 2014
    Messages:
    415
    From the sounds of it i'm going to be very interested in your mod.

    I did some minor mods myself for Civ V and the areas i modded where the areas the AI seemed to struggle in. I noticed that the AI struggled to take cities, so i weakened cities. I noticed the AI struggled with national wonders, so i removed the requirement for them to be in every city ect ect.

    I think its completely right direction to go in. Making the AI smarter is incredibly difficult, but making the game more AI friendly has a huge effect on difficulty, which for me makes the whole experience much more enjoyable. Because of this i think i will be looking forward to your mod :)
     
  15. XplosiveLun

    XplosiveLun A humble village

    Joined:
    Apr 24, 2015
    Messages:
    234
    Location:
    Hong Kong, China
    How about tile improvement-based affinity yield? Refering to " Affinity As Yield" mod. I think it is an alternative method to the LUA solution to guide AI towards a specific affinity. AI adopts their own flavor-table and constructing specific improvements/buildings to gain affinities.

    By the way, I'm not sure if this "Table affecting individuals AI to build certain thing" exists.
     
  16. Halbbruder

    Halbbruder Prince

    Joined:
    Oct 23, 2014
    Messages:
    387
    For BE specifically, I think there needs to be either reform to victory conditions or obviously, AI knows how to win the other victories. I've never seen AI win purity or supremacy victories and I have kind of seen AI get close to winning contact victories. When I see purity AI build exodus gate, they wait several turns before they start churning out settlers and the settlers just randomly move around, probably because they don't know where to settle as AI tends to squeeze in as many cities as possible.

    For domination victory, I was thinking of a mixed approach between civ 5 and SMAC: random civ characteristics option in civ 5 combined with the erratic nature of factions in SMAC where they sometimes DoW you as soon as they see you because they want to win a domination victory. Then the next issue is unit management and competency, of which I'd love to discuss more but not sure how.

    With civ 5, the AI advantages handled that but there was still much to be desired on unit management.

    I'm no expert on AI, but I'm certain that reviewing AI capacity for victory is needed before you examine the core aspects of diplomacy, tech priorities and unit management. Transcendence victory is so easy that I focus on defeating harmony AI over the other affinity factions. That, I believe, needs to be addressed.
     
  17. Nares

    Nares Emperor

    Joined:
    May 2, 2006
    Messages:
    1,631
    More or less agreed. For my personal CiV AI mod, I tuned the AI to be highly aggressive and completely disregard the safety of its units. This allows it to leverage its sole advantage over the player - brute force.

    Is it capable of presenting a challenge? Well, if I'm unprepared, it will steamroll me, as well.

    I think trying too hard to make a tactically sound AI will push any 4x game too far into becoming a tactical sim. Some AI issues, such as move-and-shoot, would be great to have resolved, but as long as the AI is capable of conquering another AI or an unprepared human, I'm satisfied, and would prefer a more competent diplomatic AI.

    I wouldn't go that far. Yes, there's a cutoff that needs to be made at some point. But, when the result is bland gameplay, I think you begin to invite greater long term issues, particularly in terms of damage to the brand.

    Granted, a great deal of the financial pressure studios seem to be facing these days is that the initial work is, too often, poorly done. This seems to be a problem across the entire industry. Having not studied game theory or software development, I can't say whether it's a problem with how those topics are being taught to aspiring developers, or if it is a more insidious problem relating to overly-PC corporate culture. Most likely, it's a combination of both.
     
  18. Gorbles

    Gorbles Load Balanced

    Joined:
    Nov 24, 2014
    Messages:
    3,816
    Location:
    UK
    Inferring a lack of competence, even initially, doesn't tend to make for great threads. As usual, the answer is generally management. But even they're beholden to stakeholders. This is how the current version of the world works for the most part. Some can behave a bit better or a bit worse than others, but basically projects are funded by investors. Investors only invest to return a profit - that is their sole motivation.

    Which is why things trend to making more money. Which trends to harsher deadlines.

    To thoroughly finish this off-topic (sorry Ryika!), our demand for post-release support as consumers has allowed publishers to be more lax about a state of a product. Back when post-release support wasn't a thing, games had to be as good as they can be on launch. This didn't always mean a better launch, or a perfect game. But it gave publishers less leeway, and less time to fix a poor product.

    Not that I think BE is poor, but that's a topic for another time! :D
     
  19. GAGA Extrem

    GAGA Extrem Emperor

    Joined:
    Dec 24, 2008
    Messages:
    1,577
    Gender:
    Male
    Okay, I gave it a spin. Base mod + your two recommened other ones + some cosmetic stuff (enhanced Biome colors).
    Game Setup: Soyuz, Massive Terran Map, 8 Sponsors, Standard Speed
    Player Setup: PAC, Artists, Specialist Guild, Laboratory

    Am I supposed to start with a free settler? Felt quite odd for the pacing and made the whole trade route stuff pretty OP (but on the flip side seems to equalize starting odds?).

    General feeling was good. As said above pacing has some problems, but I felt like there was more decision making and that ressources around me did (finally!) influence tech choise quite a bit.

    Overall the lack of Academies and the removal of science conversion felt like a good change. Also the additional unlocks for techs finally let me pick techs because of their beefits instead of just rushing all the Affinity techs. At least early on. I guess by making growth/pop the most important factor we are back to a scenario where you have a lot more different ways to achieve the main goal (= max your pop).

    Yields are *a bit* off, though. I finished on turn 237 (Harmony victory) with 6 cities, including a pop 52 capital and 3.8k science per turn. Used a specialist economy like strategy (utilizing the +1 specialist yield loadout) and went for Industry / Knowledge virtues. Energy was always plenty (so I started spamming Terrascapes), food was easy to come by and later on I had insane production values in my cities. Health is pretty borked. I never had big health issues even with the PAC trait (if it even works, it seemed like I only got 6 unhealth for cities?), I was around 0 or slightly positive for most of the early game and then had it skyrocket later on. Finished with 150 positive health, which seems a bit excessive (although, to be fair, I had the Promethian in my capital). Early Trade Routes were pretty OP (I'd recommend to scale them much more over the course of the game).

    In regards to the AI, I didn't notice a big difference compared to vanilla. It was a lot more aggressive early on because of the extra starting settler (and actually killed me in my first two games with a carpet of doom - oh, how much fun that was), but only 2 of the 6 AIs did actually manage to partially start any victory conditions before the end. These two were also the ones that passed me on the leaderboard, but that was mostly due to the fact that they had a whole island / peninsula for themselve. Poor AU was squeezed between me and another AI and had to work with 2 cities. God the map generator is annoying when it comes to player placement.

    Got a lot more detailed feedback, but I think it would be best to discuss that via Skype/ICQ/Teamspeak or the like.

    Final save: http://www.mediafire.com/download/v04bgu5t0qpsxqv/MOD_ENDSAVE.rar

    Addendum:
    Played a second game, OCC as PAC.
    Similiar settings as game #1, lost by one turn to Brazil.

    Final save: http://www.mediafire.com/download/6r24r33r2c4rtkc/MOD_OCC_ENDGAME.rar
     
  20. Ryika

    Ryika Lazy Wannabe Artista

    Joined:
    Aug 30, 2013
    Messages:
    9,395
    Yeah, Free Settler is currently in the game to speed up the start a bit (all Civs get one) as it felt extremely slow without it and my goal was to test how things feel in the midgame. Will of course get removed if I ever get the basic concept done and start actually balancing stuff. :D

    That's good to hear. ^^ In the current version I've actually increased resource yields a bit more, as I found myself ignoring them more and more each time I played. Not quite sure how far that system can be pushed.

    Also good to hear. ^^ Science Conversation was one of the first things that were thrown out, I think it's pretty much the bane of the later third of the game, no matter how much the powercreep gets increased.

    Haha. Well, my fault. It's a mixture of PACs bonus being broken (I have a local version of that mod that reduces their food bonus) and the version I sent you probably still had the 3-health-per-city system that I later deemed unworthy and reverted. (As well as a food-system that was/is not yet adjusted for the increased food rate - the megacities with more pop then you could ever have Specialist Slots are certainly not the design I'm going for).

    Yeah, I realized the same thing happening especially on Sioux. Especially when AIs just start warring in the midgame their game just falls apart. They also still start beelining the way too expensive outer-ring techs way too early, so they miss out on midgame-infrastructure which basically kills their development. A lot of that can probably be fixed by completely re-doing the Tech Flavors, but at the moment I'm observing how the AI handles the changes I've made so far.
    Feel free to add me on steam (but sent me your username via pm if it's not obvious that it's you ^^) - currently don't have a working headset though.

    Haha... yeah, PAC. :D

    Can probably send you the new Version in a day or two if you're still interested. ;) But yeah, let's continue this on steam or somewhere. :) (Or feel free to test a less op sponsor. :p)
     

Share This Page