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

Difficulty... too high?

Discussion in 'Civ6 - General Discussions' started by Dracul JOSHI, Oct 22, 2016.

  1. Dracul JOSHI

    Dracul JOSHI Chieftain

    Joined:
    May 18, 2015
    Messages:
    59
    That's not as true as it used to be.
    Unity, one of the most versatile engines is using C#, JavaScript and Boo.
    Android is Java Based.
    Console development is mixed, although it does lean heavy on the C++ side, mostly due to tradition among AAA studios, if you want to call it that.

    and now I am going to bed, read you tomorrow... actually, make that monday... I have stuff to do tomorrow.
     
  2. Mad_SawBones

    Mad_SawBones Chieftain

    Joined:
    Oct 12, 2016
    Messages:
    4
    Gender:
    Male
    I had the opposite. My first game the AI is declaring war on me and on their neighbors. The biggest civ is Kongo and they surprised war me in the
    Renaissance.
     
  3. CrimsonEdge

    CrimsonEdge Warlord

    Joined:
    Aug 25, 2006
    Messages:
    256
    I feel like the early game is all about the military and the economy. After that, you can really focus on your victory condition. The Eureka's really help in this as you can focus on getting a few to boost your tech in that direction for super-early bonuses. The early focus on money or trade routes (Trajan is going to be a power house civ) is insanely strong. Units are not that expensive until you hit battleships. With the right civics you can make upgrading your units super cheap, too. Having 8 muskets and crossbowmen in the year 1,000 was too much for my continent to handle.
     
  4. Ulthwithian

    Ulthwithian King

    Joined:
    Apr 25, 2012
    Messages:
    733
    Actually, I know exactly what I'm asking you to do. If you are not willing to sketch out even in the highest level (which I am not saying pseudocode is highest level, naturally), I'm not sure how we can actually talk about the AI.

    Ad hominem gets you nowhere. I know quite a deal about AI.

    And when you run into a game, like CiVI, where you can't brute force it? I mean, really. What can you do? Even the snippet of algorithm you provided above wasn't in that vein.

    You can't play a perfect game of chess against an arbitrary opponent. Game theory tells you that.

    Why are you restricting yourself to MVC architecture? I can think of at least two other architectures that could work just as well, and in neither of them would the AI be considered 'core'.

    Please explain to me how this has any relevance whatsoever to CiVI. First, please explain what a 'perfect' game of CiVI looks like, and then please explain what an AI that can accomplish this 'perfect' game looks like. Do you want an AI that can take any game state and make the decisions that will be the 'best' (for some definition of 'best') at every turn? That changes the system state to the one most favorable to the AI? What does it look like?

    Edit: I would probably use a functional programming language for AI. Scala is a good choice, particularly if the rest of the code base is in Java.

    I will say that I think you have a very good grasp of how to write AI in a certain way. I am not sure how well your way of creating AI translates to a game like CiVI.
     
  5. ackadacka

    ackadacka Chieftain

    Joined:
    May 27, 2016
    Messages:
    13
    Yo dude, im just gonna call you out here because you are talking out your wind pipe and not the one on your head. If you think you can pump out better AI then these guys (dramatic eye roll) at 2k then your dreaming.

    Look man every couple years google has an AI competition you can compete against hundreds of other people who submit their ai into the challenge and there is big prizes and big contracts up for grabs for people who CAN actually write a good AI.

    http://ants.aichallenge.org/

    There is the link.

    Some very renowned people in the AI field regularly submit their own AI's into the competition and its very fierce and heated. If you somehow managed to win this im sure you could easily get a VERY well paying job writing AI code for people like 2k or government agencies...

    EDIT: Please don't reply i do not think you could get a bigger shovel or a ladder tall enough... Your entire i can write more competent AI argument is ABSURD when you fail to take on the current AI at the most basic level. If your grasp of it was such that as you say "i could write a better one in a few mths" then you would be able to run rings around it in Diety and then some.
     
    Last edited: Oct 22, 2016
    Ryika likes this.
  6. Nefelia

    Nefelia Prince

    Joined:
    Aug 12, 2005
    Messages:
    458
    Pretty much. We've all started Civ VI with habits and expectations formed from different Civ games.

    But Civ VI is not an expansion, it is an entirely new game. We're going to have to develope entirely new strategies and - as you aptly worded it - skill sets in order to thrive in Civ VI.
     
  7. Dracul JOSHI

    Dracul JOSHI Chieftain

    Joined:
    May 18, 2015
    Messages:
    59
    I brute-forced it, because it was the more efficient way... not because I couldn't do it differently. This would probably be a mix of both, depending on how the modding would be implemented. For example, if we would include a standardized interface for tech modding which would read out standardized variable like 'based on tech, required for tech(s), affects, etc.' Then a dynamically drawn tech tree would be required (and a good choice) and thus a dynamic handling of the available techs by the AI. On the other hand, for the static tech tree of the base game, a pre-chosen RAW order (pending affection by the AI based on situation) of the techs for each leader-type would probably be most efficient.
    A good way to test the algorithm would be to reverse-engineer the static tree using it.. as in, does the dynamic algorithm for the modded tree when fed with the static tree return the RAW trees?

    As for leader AI types... well, please keep in mind that I haven't done proper research in competitive strategy so I don't know the little tricks yet, but let me make an attempt anyway... also, keep in mind that it is not build on any specific Civ:

    Warmonger (offensive) - Settles 1-3 Cities, builds units until capacity is reached & Raids barbarians, builds military buildings/extorts money from city states/extorts strategical resources>luxury resources>mone from Cultural & Diplomatic civs/Sells expensive 'peace-keeping' treatises to trader Civs, builds fortifications & upgrades units, declares & fights wars against other civs ( first those that refuse their 'offers' starting with City States, followed by those with big militaries, because they are a thread and the other civs are more useful to be extorted), expands mostly through conquered cities if their placement is good and they have enough luxury resources, otherwise razes cities. Gives no gifts to friends, because they should be giving gifts to him. Does not ally with civs with a strong military, does not protect friends or allies in war times, does not agree to foreign 'joint war' offers, but insists on Allies joining their war effort. Builds only wonders that pertain to their Civilization historically or provide a military advantage. Does not surrender & does not accept surrendering. When nobody is left to opposed them, demands 'submission' from the next biggest civ until Domination Victory (+ Religious Victory for certain Civs).

    Sentinel (Deffensive) - Settles 3-5 Cities, builds units until capacity is reached, builds fortifications & tries to make political allies, gives 'protection treatises' to Diplomatic and cultural Civs for open borders and sells them to trader civs for money, pledges to protect military City States and tries to gain their favour, expands when pragmatically viable or during long peace times, Produces military units and gives excess units (particularly outdated ones) to Citiy States, Allies and friends, when ally, friend or self receives a DOW (but not, if an ally or friend sends one) fights against invading units first and then tries to make peace, if refused starts conquering and annexing cities (no razing, if they don't have enough luxury, they either sell it to a trader civ or failing that, give it to an ally or friend). Surrenders when losing a war to fight another day and always accepts surrenders. Tries to unify the world in peace and win the Diplomatic Victory (+ Cultural Victory for certain Civs).Builds only wonders that pertain to their Civilization historically or defensive military ones.

    Trader - Settles 2-4 Cities, builds trade routes, builds monetary buildings followed by either cultural or theological buildings, expands when economically viable, Improves & Sells spare resources and tries to grow to make more and more Trade routes. Vies for Merchantile City states. Buys their army when declared war or buys mercenaries (Units from city States or other Civs, particularly Sentinels). Buys cities from other Civs. Goes for a Cultural Victory (+||Religious for certain civs) Builds only historic or economic wonders. (no cultural or religious), surrenders quickly to minimize losses (but doesn't offer much of worth in their peace treatises). Gifts some (but not much) money to friends every so often. Supports allies in war times financially (pays their military upkeep or similar).

    Diplomat - Settles 1-3 cities, builds any type of building, expands slowly or not at all, has an diplomatic bonus and explores the map to find all city states and civs and gain diplomatic favours. Lets others fight their wars, build their roads and provide their military. Builds a lot of builders and improves their own fields and those of other civs to gain their favour. Tries to unify the world in diplomacy and win a Diplomatic Victory. (+any that might occur earlier), surrenders quickly and tries to do whatever necessary to keep the peace. Refuses to join wars or pick sides in conflicts. Builds no wonders. Vies for all City states.

    Scientist - Settles 2-4 cities, builds scientific buildings and a few advanced defensive military units, scouts to explore and archeologists to research. Tries to stay secular and uninvolved in Diplomacy. Does reject ally or joint war agreements. Occasionally gifts friends techs. Goes for the Scientific Exodus victory. Builds historic or scientific wonders. expands when pragmatically reasonable. Similar speed to Traders. vies for scientific city states.

    Cultural - Settles 4-5 cities. Builds Cultural buildings. Tries to impress other civs but do their own thing. builds historic or cultural wonders. goes for the cultural victory. vies for cultural city states. Expands as much and as quickly as possible. Enjoys friends and allies and gifts monuments, theathers or great works of arts to their friends. Surrenders medium quickly and offers descend stuff.

    Spiritual - Builds 1-3 cities, builds any available wonders and theological buildings as well as fortifications, tries to enlighten the world peacefully (theologic pressure & Missionaries) and gain a Religious Victory (+Diplomatic Victory). Surrenders late and offers little. Likes having friends and allies and tries to aid friends with what they need most if they are capable of it and highly value gifts from others. Expand slowly, but decisively. Vies for religious city states.

    Zealot - Builds 3-5 cities. Builds theological buildings, builds only historic and theological wonders and military buildings. Have a standing army. Are friendly to all that share their religion, neutral to all without a religion and unfriendly towards those of other religions. Build often Inquisitors and aren't beyond a holy war to spread their religion faster. Go for the religious victory. Do not give gifts but appreciate them. Expand as quickly as possible.


    Actually, game theory says that one can't WIN a perfect game of chess, because it would time out according to the 40 turn rule. However, it also tells us, that the next closest winnable game is where White wins... because chess is an unfair game and white has the +1 modifier due to starting.
    One reason I like Tic Tac Toe more, it's basically the same but doesn't take as long.
    That said, I did make up a few different versions of 'fair' chess in chess club back in the day. Nobody really cared, but I had fun with it. ^^
     
  8. Ryika

    Ryika Lazy Wannabe Artista

    Joined:
    Aug 30, 2013
    Messages:
    9,395
    So in a game that is about adapting to your surroundings you propose fixed leader strategies. Not only that, but you also limit the AIs focus to a very narrow window of possibilities. And then you also haven't described any elements that would allow them to adapt to the activity of their neighbors.

    I'm sure you just haven't had the time to talk about these problems yet, because you were busy writing down a system that is basically a much more restrictive version of the flavor-system that Civ 5 used. Overall, that sounds like you know very much what you're talking about. AI bonus yields will soon be a thing of the past, because you will have solved all problems script-AIs have by not addressing them at all.

    You're on a great path to one day win a Darwin Award. Keep up the good work.
     
  9. Dracul JOSHI

    Dracul JOSHI Chieftain

    Joined:
    May 18, 2015
    Messages:
    59
    You're completely missing the point. That's an entirely different kind and most importantly, level of AI... like... MASSIVELY different. If I really tried, I could probably come up with something acceptibly descent for that, but that would be a LOT of work I ain't willing to put in, because I don't see a purpose in that and I don't want a 'good job'... I don't need one. I only talk about this, because I want a better one in Civ and that, I would do... because I always wanted to work on games, that's why I became a programmer in the first place.

    Also, first of, the difficulty is not intrinsically linked to the quality of the AI. You could give it a 'Zerg Rush' AI and if you give them enough unfair advantages, it would be hard to impossible to beat. You are also completely disregarding the fact that I like to play civ games likes sims, not like victory-oriented board games... or that I like to play games for fun and not for challenge.

    In fact, if you had read all my posts, you would know that I HAVE won diety matches in BERT.
     
  10. Ryika

    Ryika Lazy Wannabe Artista

    Joined:
    Aug 30, 2013
    Messages:
    9,395
    Just as a side note:
    "Deity" (Apollo) in Beyond Earth is a joke and not comparable with Civ 5 Deity.
     
  11. Dracul JOSHI

    Dracul JOSHI Chieftain

    Joined:
    May 18, 2015
    Messages:
    59
    No, I ain't... at least not much more than they already are. Those are just outlines of the AI Archetypes they would follow... as in, how the AI algorithm would be weighted for certain archetypes in order to prevent all civs from doing the same thing, because nobody wants that, I think. You are right though that it is basically the same system as the one in Civ 5, because that's actually a very competent way to do this. It's just that the implementation, meaning the underlying AI that uses the variables and the variables in question could be drastically improved if the way the game is build on a more efficient system than XML/LUA.

    again, as in my example, you could create ArrayLists to keep track of who know which Unit and it's location and hand that information at an AI with a difficulty value which would determine the appropriate way to move/act and collect feedback information over the process of the game. The Barbarian mechanic should bolster the AI's early lack until it has sufficient information.

    For example, on diety, it would wait for reinforcement, fight fortified if possible, use hills, woods and rivers to its advantage and try to match unit type and only retreat once the next round would likely end deadly while on settler, it would charge blindly into battle, always attack offensively, leave itself as prone as possible and do the infamous 'shuffling' so that always the unit with the highest health attacks...
    Both require the same determination Algorithm... they just need different decision values.

    And that's just a combat example.. similar types of algorithms would be used for deciding tech order, civics/virtues, settlement locations and times, trading partners, diplomacy etc.

    For example, yet again. Civ 6 China would on Settler build wonders before anything else, but would use a money focus for workers in their cities and not boost the early wonders with builders... they would also build the great wall pieces willy-nilly with spaces in between to not get the money or culture bonus until later. (because they build the wallpieces out-of-order.
    in other words, they would act appropriate to their archetype and weighting, but inefficient and wasteful to give the player an advantage while on diety they would play themselves like a MLG civfanatic would.

    P.S.: What the heck would I want a Darwin award for?
     
  12. Dracul JOSHI

    Dracul JOSHI Chieftain

    Joined:
    May 18, 2015
    Messages:
    59
    Well, so is Civ 6 apparently.
     
  13. TheUndeadFish

    TheUndeadFish Chieftain

    Joined:
    Aug 26, 2012
    Messages:
    18
    Not to be mean or anything, but to write a fully functioning useful AI to play deity level players requires you to have the skill set of a deity level player's decision making skills. Which for the most part is done on the fly in reaction to things that are happening in the game.

    Supposing you knew the end product's release state even from the start of the entire project, which I have no idea when it started, but for the sake of discussion lets say it was when G&K was released for civ 5. That gives just just barely over 4 years to develop an AI that can make good decisions that are constantly learning, and evolving because if you don't make such a thing players will easily be able to rock paper scissors your AI to exploit their rigid decision making skills. Since there is no comeback(you call cheating AI) AI handicap mechanic from the very start of the game they essentially already lost even before they started the game because you already know their exact build order and you know at what army threshold they will declare war at and there is no possible way they can even do simple things like declaring war on you because you have the advantages of evolving your strategies, where the AI only knows what you told it before strategies were even a thing.

    So supposing you can happen to make such an AI that can do all of these things which will evolve its strategies and can counter the deity players to the point where they are actually challenged. Now you have 7 more completely different AI because you need to purposely program the AI to make poor decisions to accommodate for players that can't handle making the perfect ones at the correct time. This also includes those that are to lazy to to make the correct decisions so they wing it most the time. This compounds the problem because you as a deity tier decision maker have absolutely no idea how to make a emperor tier decision because you can't fathom how anyone could possibly make the wrong decision when the correct decision is so easy, and feels like common sense to you. Yet you need to program the AI to recognize this situation and tell it to make the wrong one in these circumstances. So now what you need to do is you need to find 7 more people that play on the other difficulties and have them program their own AI. Which they will play against and test to find out if its at the correct level of difficulty. The problem here is while they do this they will be improving their game play as they go, and in turn be improving their AI which defeats the purpose of making it at inferior difficulties to begin with.

    There is a reason AI in games is done the way it is. No matter how good that AI is it will never be as good as a human until its an evolving AI, which I believe will first be created due to video games, but we are not anywhere near where we need to be for this to happen yet. It is by far much easier to make the best possible AI we can, which I imagine every game does, and after that you decide what difficulty level that AI performs at, and you give it bonuses or penalties from there.

    If you want to play it as a sim I'm sure you could do silly things like make a custom map where everyone is permanently at peace with all victory conditions disabled apart from score, and no barbs. That's not how I imagine most people play though, sure there are TONS of people who play peacefully, my roommate is one of them being a die hard peacemonger, but that's not what I'm talking about. Me and my friends play strategy games for the challenge of developing your strategies, not as a sim games. If I want to play a sim game I will play a sim game.
     
  14. Dracul JOSHI

    Dracul JOSHI Chieftain

    Joined:
    May 18, 2015
    Messages:
    59
    First off, just because a series of actions doesn't change, doesn't change their effectiveness. In fact, most Civ players I know have extremely rigid teching etc. the only time they change something is when the current circumstances require it and that's what even a basic AI would be doing if it actually kept track of what is happening. That's not difficult... like... at all. Honestly.

    The AI doesn't need to learn between games... that would be unneccessarily complicated. It just needs to be able to make halfway competent decisions on the fly. Give it the right variables and a descent algorithm and that's not hard, since the game is extremely specialized and simple due to it's turn-base nature, particularly since the AI isn't pressured by too strict time constraints.

    RPS is a bad example, because you never know which of the 3 will happen... and you can only choose 1.

    also... if they are making decisions it's not rigid... because otherwise you wouldn't need decision... you would simply write a procedure instead... if you have an algorithm that automatically increases adabtability.

    every type and probably every civ has a slightly different build order... it will take time to learn that... and if you do that's basically no different than knowing enemy patterns in an RPG. Also, as I said before, the Algorithm would have influence of that... for example, if any civ is in the negative monetary wise, they will improve their gold gain, even if they aren't traders... and that pretty much applies to every type of issue that can arrive.

    Also, the army threshhold is a VARIABLE... it depends on how many cities the civ has and WHEN their army is ready... which can take different amounts of time based on the production and how many units survived... you may be able to predict their first attack or cause premature war by sending a DOW yourself... but after that it should be relatively unpredictable.

    Well... the algorithm can be updated just like the game itself if somebody finds a better way of doing a certain thing.

    You don't really understand how quantification of decisions within a predictable system works... let me explain it the way you may understand it: Imagine a Tic Tac Toe field and you have the choice of where to put your sign... where do you put it? You have 9 fields and the goal is to put 3 in a row, so the best place is the middle column with the value 4, followed by the corners with 3 and the rest with 2. So if you want to make the best decision, you put it into the middle, naturally... but you should be able to understand that you could simply pick a worse thing to let a small child win... Don't you?
    In order to find the best way you have to first determine the comparison value of every possible decision... changing the oder, priority or value of choice all has influence on the resulting difficulty.
    Properly balancing that is difficult... sure... but that's what testplayers, closed betas and Q&A people like myself are for.

    And no, I don't need somebody else to 'program their own A.I.'... maybe some people who know a thing or to can help me build the A.I. and especially tweak the difficulties... but that's not the same.

    You're next point is entirely correct. Simulated Intelligences would need to be incredibly complex to compete with true intelligence... depending on the environment. However, in a closed system such as a game a simulated intelligence can achieve quite sufficient competing capabilities. To the point that early simulated Intelligences in particularly simple video games were often too good for most human competitors.

    Regretably, you're wrong though. Simulated Intelligence is often an after thought by game companies, since it does not directly impact game stability, visuals or story... in other words, it does not impact that which sells games these days. Thus many games feature only the most rudimentary of Simulated Intelligences. In case of the Civ games this was mostly due to the bad performance of the games based on the XML/LUA structure of the system. In other words, because the Civ games are a Database acting like a game rather than a game using a Database.

    I don't want that everybody is at peace... heck, I enjoy kicking a warmonger's ass in the later part of the game as much as anybody and I have done silly things like that... just now I am playing a duel against Kongo with Religious victory only... but that's besides the point. What I do want is a good game that gives me options to find different solutions to problems than just 'make an army'... because honestly, that's a very close-minded approach to game design. Also... you might not remember this... but Civ was originally a sim game with a good competitive multiplayer... if I remember correct, the advertisement said: "can you make a Civilization that stands the test of time?"... It didn't say "Can you be the best at X?" and the mad amount of customization and modding you could do supported this...
    I remember all the funny/broken things one could do...
    Heck, if you watch 'Civilizations with Lyrics' by Brentalfloss you can see a custom civ with Elvis as it's leader XD
    I also usually don't stop playing after the first victory condition is fulfilled.
     
  15. Dracul JOSHI

    Dracul JOSHI Chieftain

    Joined:
    May 18, 2015
    Messages:
    59
    and I don't remember where it was that somebody critizised my idea of two civ's Religions merging, but I remember the sarcastic tone and yesterday I was too tired to answer it, but today I want to point out real quick that there is Real World presedence for this.

    When the Romans conquered Greece they actually took their pantheon and renamed it, but it was pretty much the same, Jupiter being Zeus, Mars being Ares, Venus being Aphrodite etc. Heck, they even took the epos of heroes and this this, such as Heracles turning into Hercules.

    also, when upper Egypt conquered lower egypt, their pantheons merged.
     
  16. Disgustipated

    Disgustipated Deity

    Joined:
    Nov 14, 2006
    Messages:
    11,230
    Location:
    Las Vegas
    Since this thread was originally about the difficulty of Civ 6 on the easier levels I am bringing it back to that. AI coding doesn't seem to belong in this thread.

    Finally finished my first game which took 15 hours. The game runs slow. I played an easy level (chieftan) to learn the game. Something is seriously wrong with the balance. I'm normally a Prince level player, and while I did win easily, I was surprised at how well India did in the game. They actually had more science per turn (at the end game graphs) than me for much of the game. Granted I only had 2 campus districts, but I simply did not have enough production to build anymore. I couldn't even get enough production to build all the builders I wanted. I feel sorry for novice players, they are going to struggle with this game. I'm probably at the average level since I normally play Prince/King in Civ5, but I don't play optimally, I play to get most of the wonders and things. I could play higher if I wanted to challenge myself. The point is it is strange for India to almost surpass me in science at this easy of a level.

    I did avoid going to war except to steal a settler from Brazil. Poor Brazil, they never settled another city all game, they were stuck with 1 city the entire game. Both France and India were pumping out culture fighting my culture victory, but I was too persistent, and eventually won. It wasn't really a challenge to win, but I did have to direct a lot more effort to it than I thought I would.

    And nearly everyone hated me in game except France and Norway which shared some values with me. Diplomacy and Navy apparently.

    TLDR: novice players need production and science boosts at the easy levels. And they shouldn't have to be fighting extreme housing and amenity problems like I was all game.
     
  17. Staler87

    Staler87 Warlord

    Joined:
    May 28, 2016
    Messages:
    267
    Location:
    Right now? Computer Chair.
    This civ is an actual 4X game from what I've played. You can still play peaceful but slow early expansion is not a good idea. If settler is at all difficult than that may be a problem but you are playing the game in a way that is no longer a competitive strategy at all. Barbarians being stronger, and expansion being better, means civ VI really punishes you if you are a small empire with no military.
     
  18. PhilBowles

    PhilBowles Deity

    Joined:
    Nov 20, 2011
    Messages:
    5,277
    Seems pretty spot on. In my Greek Emperor game I felt fairly far behind, and saw my efforts at cultural victory dwindling; meanwhile my opportunity to move on India appeared to have been lost as the warmonger penalty increased and they became friendly with my Norwegian allies.. Only to hit the mid 1500s having actively cultivated culture policies and buildings, and focused on securing my first Great Writer. I immediately shot from 7th in culture to 2nd. That without ever having left our shared island - a large island, with room for 8 of my cities, 9-10 of India's, one of Norway's and three city states, but this was still the home continent for only two of 12 civs so I presume there's a lot more to find (only just met the first Aztec landing party). If I'd focused on specialising from the start and shooting for a particular victory I'd no doubt be well ahead by now. Never mind if I'd had a proper war, as I've made this progress while barely using Gorgo's ability.

    I've still experienced no AI aggression other than mildly annoying barbarians - after an abortive war with India that I called off amid housing and amenity shortages combined with a barbarian attack, there's been no conflict and little interaction with anyone - the Indians now have walls in all their settlements but they aren't guarding their borders despite being permanently unfriendly.
     
    Miir likes this.

Share This Page