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

Max level - no AI cheat

Discussion in 'Civ5 - General Discussions' started by YinYang86, Jan 8, 2012.

  1. headcase

    headcase Limit 1 Facepalm Per Turn

    Joined:
    Aug 12, 2011
    Messages:
    1,213
    Location:
    Ontario, Canada
    So to reiterate the answer to the original post, this page says:
    - You get the bonuses in first table.
    - The AI gets the bonuses in the first table's Chieftain row, and the bonuses (or penalties) in the second table for the difficulty you're playing (stacked with the first table for attributes in both tables).

    Warlord is probably the most even, apparently giving AI advantage in some areas (happiness, gold, policies, research), and the player advantage in others (wonder production).
     
  2. Trias

    Trias Donkey with three behinds

    Joined:
    Oct 1, 2008
    Messages:
    594
    You are failing the comprehend how inflexible a script type AI is. A script type AI in civ would for example sit there as a zombie because its predetermined build order requires it to build a warrior but the mod you loaded has a caveman unit instead of a warrior.

    Similar problems would have occured with many of the gameplay patches of civ5. They would have completely broken a scripted AI, requiring a major rewrite of the script.

    Actually, the AI in civ5 does quite a bit of more long term planning than that. For example:
    -It plans out where it wants to expand, and adjust its diplomatic standing with civs that are in the way.
    -It tries to determine which opponents will be the most likely competitors for its chosen VC and adjusts diplomatic standing accordingly.

    There are probably more, which are not as apparent as long as we dont have access to the AI code. But it appears the AI at least has a framework to set mid term goals and such. This a lot more than civ4 had (the lack of any mid/long term planning facility in the civ4 AI framework was a major obstacle to further improvements for the BetterAI team). Of course, (as with many of the aspects of the civ5 AI) it is likely that these facilities are not being used optimally.

    LOL, apparently you dont know what a 4X game is. Empire building games ARE 4X games (at least in almost all cases). FYI http://en.wikipedia.org/wiki/4X_game


    You seem to be confusing ambitious and lazy. One of the major problems with civ5 is that the dev team set design goals that were too ambitious for the rigid time frame imposed by the distributor. The result was game in which many of the features were implemented badly. The areas that suffered most from this were game balance and the AI.
    I think that there quite a few accusations you can make towards the dev team, but them being lazy simply is not one of them. (Sorry to bust the bubbles of foam coming out of your mouth.)
     
  3. headcase

    headcase Limit 1 Facepalm Per Turn

    Joined:
    Aug 12, 2011
    Messages:
    1,213
    Location:
    Ontario, Canada
    Trias: I think the script AI Derpy has in mind is more flexible than the script AI you have in mind.

    Derpy: If Civ V isn't a 4X game, why not? Which one of the Xs is missing, or, if all the Xs are there but it's still not a 4X, what is in the "spirit" of 4X games that is supposedly missing in Civ V? And don't mention the AI; Civ IV could be played with all humans and still be considered a 4X.
     
  4. TheNiceOne

    TheNiceOne Emperor

    Joined:
    Feb 6, 2002
    Messages:
    1,372
    Location:
    Norway
    Well, he said "series of fairly simplistic scripts", that doesn't sound like very flexible scripts. And when you start making the scripts flexible, they soon stop being scripts at all, but rather rules, which is what ciV has today.

    I think we can all agree that the ciV AI rules could and should be improved, but personally (and from a professional view, I'm a software developer) I believe that only people who know nothing about AI programming are the ones who think it would be easy to improve it, or that the faults are due to laziness.
     
  5. headcase

    headcase Limit 1 Facepalm Per Turn

    Joined:
    Aug 12, 2011
    Messages:
    1,213
    Location:
    Ontario, Canada
    Ok, agreed, "simplistic scripts" are not what Civ 5 needs, and improving the AI will take some time (besides fixing the dead simple mistakes they make which frankly I've seen less of in the newer patches, at least in NiGHTS).

    Really I think fans (modders) are better positioned to improve further; the truth is the game is selling well with the current AI so "corporate husk" Firaxis has less and less reason to improve on something that big fans will do for them (after all, they didn't make the better AI in BtS; mods did). Hopefully we see that DLL soon enough.
     
  6. Trias

    Trias Donkey with three behinds

    Joined:
    Oct 1, 2008
    Messages:
    594
    Also, improving the requires many man-hours a lot of which is in gain experience in the game decision making and abstracting that to efficient rules for the AI. This is an area where the fan community has a huge advantage over the developers. They have a huge amount of experience playing the game, and know what works and what doesnt when making decisions. They also have much better experience with the mistakes that the AI makes. (This is especially true for deity level players, since the game at deity level consists of exploiting the errors that the AI makes.)
     
  7. Becephalus

    Becephalus King

    Joined:
    Nov 30, 2005
    Messages:
    725
    Honestly the easiest way to improve the AI is the one fans would complain the most about. That is constraining the number of failure points the AI has to encounter. Instead of having thousands of decisions a turn where the human can outperform the AI 5%, only have hundreds.

    If I were going to improve the AI my first steps would barely touch the AI:

    1. Get rid of ranged and siege units (maybe air units too), all units are "melee". Makes coding the AI much easier and is a system the AI can operate much much more easily.
    2. Get rid of rush buying and other duplicate mechanics. You can already turn gold into production by building production buildings, you don't need a second way to do it.

    You want a game that still has the core of the Civ experience (choosing tech paths, choosing city sites, setting construction queues, assigning workers, commanding armies) while cutting out as many superfluous decision points as possible. That way the game can be a lot more challenging without the AI having to cheat (or at the very least the cheats can be a lot smaller).

    I might even go so far as:
    3. Make the rewards for warring smaller, but make it so the only diplomatic stance is "war". Completely removes all the problems associated with diplo AI (which has been crap in every single version of civ and is only slightly better in this one).

    After you have an actually tight game were every decision matters and the various mechanics aren't at odds with one another, then I might start trying to "improve the AI".

    Gal Civ 2 made this exact same mistake. It had this entirely useless and redundant "planetary focus" mechanism which was way way overpowered, not necessary for core gameplay, and the AI didn't know how to use at all. Using planetary focus was basically an "I win" button. You would point this out to Frog Boy and it was like you were speaking latin. He would be all "But players like the planetary focus option, it allows them more control over their empire?" And you would be all "But the hate the weak AI even more, and this is one way you could drastically strengthen it."

    I wish game designers would think a little more about AI limitations and re-playability when designing their games.

    instead they just pack in the toys and wing dings because gamers all love toys and wing dings and want more options and more power and more interface controls. And then they complain when the AI sucks.

    Every tool you give the player is another club for them to beat the AI with.
     
  8. Derpy Hooves

    Derpy Hooves Grand Inquisitor

    Joined:
    Oct 19, 2011
    Messages:
    325
    Location:
    Muffins!
    Trias, the situation you describe is extremely easy to solve by both the devs or modders by simply changing a single line and should not by any means be a constraint.

    About the long term AI, I'll give you that they change their approach based on where they want to expand, as can be easily seen from the oh so common "Coveting lands" modifier, I would go as far as to say this part of the AI is broken because it often has little understanding of how far the enemy is from their main city cluster and will covet your lands even if your on the other side of the world.
    Not to mention that even though it expands it just tries to occupy as much space as possible with almost no regard for placement, leading to the well known 1pop snow/desert hovels.

    The script that tells it which CiV is a competitor is also broken, from the AI hating you at 500bc for going for the same VC, when you're not even sure yet what you want to do to them just making wrong decisions (Monty, warmonger, yet cultural civs think your going for the same as them because of high culture from killing).
    They also heavily prioritize human players, if another civ gets close to a VC I've never seen the AI gang up on them or hate them because they go for the same VC, if anything they become friends because their army and score ratings are close to eachother.

    And you where the one to start about the 4x stuff, essentially they mean the same, you where the one however to state it was 4x and not an empire builder, implying the terms are mutually exclusive.

    Was the dev team really innocent? Was it the evil distributor that pushed to release an unfinished game? Do you have any statements from Schafer that tell you this? And can you be sure it's not just after the fact rationalizing?
    It seems to me this is just your personal conjecture, heavily founded on your love for a studio that has given you many wonderful days of enjoyment but you're now starting to realise has forsaken you. (Sorry to stop your fanboyish drooling)

    And when I say a series of fairly simplistic scripts, this does not mean they can't be complex, they are just simplistic compared to the grand chess algorithms that you seem to think are the only way to fix this. Not to mention that 5 short scripts can do the same as a single long one, the length or depth or these scripts should not be relevant.

    And NiceOne, the AI as highly defficient, call it lazyness, being rushed or a too difficult task, we should be able to assume a competent studio should overcome these or find ways around them, even if it meant reinventing combat again to be able to create a viable AI, not just the release the steaming heap of crap and hope people are too stupid to notice.
    Not being able to do so makes them into a incompetent studio (a belief strengthened by their facebook game).
     
  9. Trias

    Trias Donkey with three behinds

    Joined:
    Oct 1, 2008
    Messages:
    594
    LOL, NO. You really are a bit clueless when it comes to the reality of AI, aren't you?


    I agree, the current algorithm that determines where the AI wants to expand isn't much good. This is something that needs to be improved.

    No, read back. You were claiming that civ is a simulation, which it is not.

    Where did I say, the dev team is innocent? They quite probably knew the schedule they were on. They seemingly made a big error in judgement in terms of what they could do in the alotted time. And no, this is not rationalizing after the fact, but was my biggest concern the moment the announced the game which was clearly in an early alpha stage, and immediately announced that it would be released the same year. (That did not bode well for the state in which it would be delivered.)

    But as long as we are talking specific no dynamic order scripting it will still be worthless for civ. Civilization needs algorithms that dynamically make decisions. This is hardly chess level AI, but is harder to do, than the relatively simple scripting that the games like SC use to imitate human build orders.
     
  10. Derpy Hooves

    Derpy Hooves Grand Inquisitor

    Joined:
    Oct 19, 2011
    Messages:
    325
    Location:
    Muffins!
    Silly me, overestimating peoples creativity, let me take you by the hand and guide you through some easy solutions to the problem:
    You could leave the cavemen with the Warrior base ID, letting the AI think it's the same unit it's controlling if it's a replacement for the warrior anyway.
    I assume there is a single AI for all melee units, just add the caveman base ID to the list of units using that AI, done.
    Worst case scenario you could even just copy paste the warrior AI and attach it to the caveman.

    I kinda figured these options where so obvious I didn't have to list them, I'll try not to make that mistake when talking to you anymore as the LOL indicates to me you failed to spend any time actually thinking about it.

    Yes I said Civ was a simulation game, you said that couldn't be so as it's a 4X game, as I think I explained last post, the terms are not mutually exclusive.

    You said the dev teams problem that caused this was their ambition, which is usually a good trait, if they where truly ambitious they would have pushed back the release date to truly create an excellent game, like any good studio does, do you think Activision forced Blizzard to set a release date for SC2? Or Bethesda forced Black Isle to rush out New Vegas?
    Yet they did not bother with any of that and just released a broken game.
    Only a lazy studio that is focussed too strongly on making a quick buck and disregards it's fandom would do such a thing, mothereffing lazy.

    Not sure where you got from that the scripts can't be dynamically ordered and I'm not sure what you consider "what it needs" but scripting can immensely improve the AI, again, not to the extent of being a granmaster, but enough to make it a somewhat competent adversary, which is miles beyond the toddler unit spam we get now.
     
  11. Peng Qi

    Peng Qi Emperor

    Joined:
    Aug 19, 2007
    Messages:
    1,431
    Location:
    Irrelevant.
    You can tell CiV is a simulation game because when you adopt the Republic social policy the game sends a senate over to your house who you have to convince to vote in favor of all further actions you take in the game.
     
  12. gozpel

    gozpel Couch-potato (fortified)

    Joined:
    Jan 3, 2002
    Messages:
    4,381
    Location:
    Australia
    It's funny how I've taken the higher (moral) ground lately with Civ5, as in abusing the AI. It's wrong to take money for free by open border one way and selling luxes before attacking.

    I'm a wimpy charlatan and a half demented scumbag! :)

    I played civ3 massively and what did we do? Yep, using every trick in the book and sometimes stuff that wasn't even written down, just to get that upper edge. Having an army of 46 artilleries and 4 swords, you could/ nah WOULD conquer the world on deity and I did many times.

    Your first 5 turns in civ3 deity, you have 1 warrior, you're building another. You meet China near your borders, 8 of their warriors in a stack parading, they already have 2 cities and are most likely building a warrior every 2 turns. Your warrior will take 4-5 turns.

    Smile, be nice, give in for the demands and try to survive. Sometimes that didn't help and you were dead within 40 turns. But I looked back at my 5cc games and wow, did I abuse everything I could? Sure, I did and so did my friends in the SG's.

    I'm off that tall horse now and will do whatever it takes to grab me extra gold or whatever. Doesn't matter what I do, Ill get backstabbed anyways. :)

    The AI isn't better than in civ3 and I'm not as good a player as in civ3, compared to 5, just because I'm a frigging hypocrite!
     
  13. KM

    KM Warlord

    Joined:
    Oct 11, 2010
    Messages:
    284
    deity is the oly balanced level since that s the only level where the AI actually got chanse...
     
  14. KM

    KM Warlord

    Joined:
    Oct 11, 2010
    Messages:
    284
    Max level no AI cheat: Multiplayer ;)
     
  15. gozpel

    gozpel Couch-potato (fortified)

    Joined:
    Jan 3, 2002
    Messages:
    4,381
    Location:
    Australia
    Sure, until the guys wimp out and leave you with nothing to do, great fun! :)
     
  16. Trias

    Trias Donkey with three behinds

    Joined:
    Oct 1, 2008
    Messages:
    594
    Silly me, and I thought a simple example would help you see the core of the problem. I wasnt even talking about controlling the unit since the AI wouldn't know to build one. (Even in heavily scripted AIs such as in SC, unit movement is often based on a more dynamic algorithm such as threat mapping.)

    Of course since this is a silly simple example it would be fairly trivial to fix. The point however is that it would need to be fixed at all. For a scripted AI that goes for every single change. Since the modders do not have access to the AI (yet) this would have precluded pretty much every mod from being made. And even when the full DLL code is available it would impose a huge barrier to modding for the less code savvy.

    And this is just one of the flaws of a scripted AI. The other is the inability to deal with on foreseen in game situations, which given the versatility of civ are fairly common.

    Now, please wrap your O So creative mind around the fact the a more dynamically programmed AI that determines its next move by evaluating all possible moves based on the current game state, is perfectly possible to program. And although requires a bit more investment to get right, it will work better than any scripted AI when properly tuned.

    The only thing which is not very easy to do is to easily adjust the level of play of the AI. (And judging from this prerelease interview, the devs were actually think about deploying a system to adjust actual level of play for the AIs. I am not sure what happend to that. But of course, a system like they describe will only allow the AI's ability to be scaled down, it will never be better than its best.

    I did say it couldn't be. I said it was a 4X game first and foremost. That is not to say that immersion is not important, it is. But, emulation of human history should follow as an emergent result of the game rules, not by the AIs playing by a different set of rules to roleplay for the human. (Remember that the OP was complaining about the AI playing by different rules. That sword cuts both ways.

    Yes, I can only dream of the day that Blizzard (or any other studio with an its finished when its finished policy) buys the civilization franchise. But for now we are stuck with firaxis.


    No, it could not (immensely improve the AI). It might superficially improve it a bit, but not even that much. In fact, it would probably be fairly bad, because the devs clearly had fairly little time to gain enough gameplay experience to build a decent scripted AI before release.
     
  17. Yzen Danek

    Yzen Danek Warlord

    Joined:
    Dec 15, 2005
    Messages:
    182
    You know why this game is too easy? Ask Deep Blue at IBM.

    When the stack of doom was allowed, the AI wasn't wasting its time trying to play what is essentially chess. Now it is. And, as Gary Kasparov and Deep Blue have repeatedly shown, even when he lost, in a game as simple as an 8 x 8 grid with an extremely finite number of decisions to make per turn, a computer is intensely challenged to play well, to the point where it takes the computing power of several late-model Cray supercomputers to make the game remotely interesting. I can't speak for everyone in this community, but my computer, like most gaming rigs built in the last 18 months (my absolute limit before obsolescence), isn't struggling with any demand offered by this software; instead, they (the developers) left the struggle of finding a challenging game to the users; in basic terms they left us to figure out how to limit human aptitude to create a gameplay experience that appears challenging, even when even simple basic human intuition trumps every algorithm contrived on the behalf of the UI.

    This wasn't always true for Civ games. When infinite units could stack, offensive actions by the AI were a simple extension of the hammer-based-economy the AI could muster. The computer does that well, beating humans at simple equations of optimization- hammer production vs. upgraded units vs. gold-purchased units vs. conscription, and the order in which to use that optimization to attack for maximum effect. What the computer doesn't do well, at all, is simultaneously optimize new unit production vs. upgrades vs. positioning when faced with the hex-based, one-unit-per-square spatial backdrop that this game exemplifies. There are somewhere between 8X(ish), and 32X(ish) more squares for the AI to consider tactics than there are in chess, and as human beings we will only acceptably wait 10i-20sh seconds for the AI to make all the decisions we require before we want our turn back. Afterall that's how long we expected to wait in previous Civ titles.

    What that means is that the limits given to the AI in this game very likely have parameters wherein, with certain exceptions, human beings' "gut feelings" would in many cases be superior decisions to those made by the AI under the limits required. No surprise then that Darius, with a 15 city lead over me and 5 cities over 30 population while I have 1, is dicking around with how to best play chess against me on a 256x128-sized board. Some part of the algorithm he was given requires him to make Hospitals and Medical Labs when I'm building the SS parts that are going to kick his ass and win the game.

    This newest iteration of Civ with no unit stacking (like chess) has very obviously become waaaaayyy too much for a simple script AI to handle in an acceptable amount of time while offering an acceptable challenge.

    And thus, the AI frigging sucks, and you should give it the opportunity to have massive advantages to let it try to still beat you. This arises because you are the only player who understand simple game subtleties. To wit: the fact that AI catapults/trebs/cannons/artillery will each use half their turn to step into the range of player artillery, consequently set up, and get fired upon in the turn they use to set up, thereby being half a turn behind in the exchange and losing all units every. single. f'ing. time. And therefore losing.
     
  18. Lone Wolf

    Lone Wolf Deity

    Joined:
    Dec 4, 2006
    Messages:
    9,898
    Well, "challenging Kasparov" and "remotely interesting" are not the same things, and today, engines like Rybka, which you can run on your home PC, play at perfectly Kasparov-like level.

    Civilization UPT is much more difficult for a computer then chess, though.
     
  19. Yzen Danek

    Yzen Danek Warlord

    Joined:
    Dec 15, 2005
    Messages:
    182
    You're correcting me, but we're agreeing.

    Moore's Law, if holding true, makes today's home computers kick the sh!t out of Deep Blue. It was 1997 when Kasparov last played. 14-15 years is an eternity in computing time. Deep Blue, fully optimized by a team of engineers, couldn't have even run Civ 4.

    And what I'm saying is that the Civ Developers completely changed the paradigm with this title. 40X cavalry/cossacks attacking from a single tile, a basic, scripted AI could handle; how to optimally deploy 40x cavalry so they can attack through a 2-space-wide bottleneck, retreat when necessary, avoid low-health units from blocking healthier units from advancing, all in the face of 2x Great General defensive structures in the gap through the mountains, against strong defensive units and backing artillery is a nearly impossible task for the AI. I've never in my entire "Civ life" (don't tell anyone I coined that phrase or I'll find out where you live and murder you) had as easy a time using basic tactics to defeat a strategically superior force as I have had in this version of this game. Was defense a joke in Civ4? No. You weren't merely faced with a turn-by-turn puzzle of how to fend of the front units and stay healed; you had to beat the stack. Barrage damage was the only tool you had against the stack of doom, and a superior stack would overcome it. Now? Cities have ranged attack power. Siege units have to set up until Rocketry, and are in most cases a turn behind by the time they get to fire. The defender's advantage is the strongest it's ever been.

    In this context, a "gut feeling" about how to hold off aggressors is not especially foreign to a human player, and because of that fact, even as a newcomer to this addition to the series, most of us kick the living crap out of the computer if we force the issue to be one of tactics set against the spatial rules of this game. That was my point.
     
  20. Trias

    Trias Donkey with three behinds

    Joined:
    Oct 1, 2008
    Messages:
    594
    You seem to forget that the AI absolutely sucked at stack warfare as well. The only saving grace for the AI in civ4 was that due to its bonuses it could field stack that were several times as large as the human's. This effective counter acted the mistakes it made in choosing the unit composition.

    Is I have mentioned many times. A big flaw in civ5 is that the AI at higher difficulties gets the wrong bonuses. The bonuses it gets are really ineffective at counter acting the mistakes the AI makes (bad unit placement). In fact, giving the AI more units, just increases the mistakes it makes with placement, and causes more traffic jams. It would have been much more effective if at deity the AI got units that are twice as strong, rather than getting twice as many units. (Although, whether if then it would challenging is still doubtful. But at least it would not need as astronomical bonuses for economy as it it getting now.)

    The AI does OK in that department, but not well. And certainly not better than most human players

    Hexes or square hardly makes any difference for the AI.
     

Share This Page