idlih10 wrote:
But all that is beside the point. Why have a quest victory for the AI when it's not much different from a turn victory? It's highly unfair that the AI would play by different rules to achieve the same victory type as the human player.

I think you're not understanding what the Dev said.



The AI gets to automatically complete their faction quest steps, so as not to make them miss out on the benefits (the various rewards) that you get at each step. However, because allowing the AI to achieve faction-quest victory in this manner would be cheap, they can never actually finish the very last step.



So they'll "progress" their faction quest to the last step, so as to gain the rewards from it, but they'll never finish the quest that way, thus, they will never achieve the faction quest victory condition.

Interesting enough, in SimulationDescriptors xml for game difficulty, Normal difficulty is the only difficulty that is not having Interact_OddsTriggeringAQuest set to 0.



Not sure it is bug. Bug all other difficulty have it set to 0 with following comment:





While for Normal difficulty it looks like this:





With section that does this commented out, so it does not apply.



I'm not sure what to think about this...





Also, Serious and all difficulties below it (exempt Normal, which has it commented out), have Interact_OddsLooting, with substraction of 40 (penalty?), while Impossible and Endless have it as addition of 40 (bonus?). This also does not seem right (why penalize difficulties above normal, and why again is normal ignored).

Slashman wrote:
I was wondering what quests the AI CAN complete. Because that is a quest that requires you to go to a specific place and have a specific amount of a specific resource. Is it the only one?

I´m pretty sure the AI only "cannot" complete faction quests, for both wonder and quest victory. Any quest that´s triggered by the player to affect the world (like Tetikes quest, A plague of savagery, the second mouse) or by the World (from dust to rust, defeat guardian X to get Y) can be completed by the AI just fine.

Grzemek wrote:
Reading with comprehension too hard for me (-_-) Disregard my question, Slashman, I read your post again and got it Sorry for starting a misunderstanding.

No need to apologize. I may not have been as clear as I could have been anyway.

As for other quests, AI can also complete "Plague of Savagery" quest, the one where you need to search for 2 ruins in a Tetike-infested region for a hefty pile of strategic resources, ~defeat either Hurnas or Tetike army~, the first step at least (AI DOES spawn these armies on their own so I guess they trigger the quest form ruins).



All above quests along with FDtR don't really require AI that specific and intentional actions - ruins/villages involved in the quests may simply be the ones on their exploration/expansion route. While I don't mind if Plague of Savagery is quickly resolved this way (I frigging HATE that Cultist units from converted villages are empowered by PoS), there are some quests that I despise being failable by AI, especially when T3 Weapon/Armor tech is the reward.



I have no idea if they would complete the "Lack of Luxury"- a competitive quest which requires 10 double-cost booster activations of Luxuries from specific Era - this quest procs very rarely and very late into a game for me.

Interesting. I wonder if future quests added to the game will be doable by the AI or if they decide that after they put them in and test them.

Slashman wrote:
I was wondering what quests the AI CAN complete. Because that is a quest that requires you to go to a specific place and have a specific amount of a specific resource. Is it the only one?

Reading with comprehension too hard for me (-_-) Disregard my question, Slashman, I read your post again and got it Sorry for starting a misunderstanding.



As for other quests, AI can also complete "Plague of Savagery" quest, the one where you need to search for 2 ruins in a Tetike-infested region for a hefty pile of strategic resources, ~defeat either Hurnas or Tetike army~, the first step at least (AI DOES spawn these armies on their own so I guess they trigger the quest form ruins).



All above quests along with FDtR don't really require AI that specific and intentional actions - ruins/villages involved in the quests may simply be the ones on their exploration/expansion route. While I don't mind if Plague of Savagery is quickly resolved this way (I frigging HATE that Cultist units from converted villages are empowered by PoS), there are some quests that I despise being failable by AI, especially when T3 Weapon/Armor tech is the reward.



I have no idea if they would complete the "Lack of Luxury"- a competitive quest which requires 10 double-cost booster activations of Luxuries from specific Era - this quest procs very rarely and very late into a game for me.

Grzemek wrote:








I am confused now :P Did you mean that in your games AI has not contributed to FDtR previously and has started doing it recently?

No it hasn't for me. I was playing most of my previous games at normal difficulty. Maybe that made the most difference. But in my current one (hard) I contributed absolutely nothing to that quest. The Ai did.



I was wondering what quests the AI CAN complete. Because that is a quest that requires you to go to a specific place and have a specific amount of a specific resource. Is it the only one?

Slashman wrote:
One thing to note is that the AI does seem to complete the co-op quest (Dust to Rust I believe it is called) where you have to contribute blocks of 20 Palladium.

From Hard difficulty and higher AI does contribute to From Dust to Rust. Happened to me numerous times, with different resources needed to donate, I play solely single player.



Keep in mind that AI has bonuses to production and Dust on higher levels (don't remember about strats and luxuries) and it might be beneficial to keep the production penalty because AI has more Dust to purchase resources and "thinks" it will help the opposition more than them.

Slashman wrote:


In my current hard game, the AI completed this quite handily. What is at work there I wonder?

I am confused now :P Did you mean that in your games AI has not contributed to FDtR previously and has started doing it recently?

One thing to note is that the AI does seem to complete the co-op quest (Dust to Rust I believe it is called) where you have to contribute blocks of 20 Palladium.



In my current hard game, the AI completed this quite handily. What is at work there I wonder?

I think there is some confusion regarding faction/victory quest and AI involvment.



AI auto-resolver will only resolve faction quests, not victory quest (aka spaceship quest).

Also, it will resolve some other side-quests.



Now one thing that is not completely clear from developer responses is how capable is AI to go for actual victory quest, since it is not resolved for them automatically.



1) Does AI have actual algorithm to try to win victory quest manually?

2) If not, can it actually "stumble" upon winning it (by randomly going into ruins)?

3) Or is AI incapable of winning victory quest and, thus warning that some faction has started victory quest is a "fake" warning and that there is no danger of AI winning this way.

I think calling the Victory Quest "broken" just because the AI plays by different rules is an unnecessarily negative way to describe the situation.

It also ignores the very real constraints on the amount of work that can be put into a single game.



The way I see it, the timers are more of a safe-guard than the primary quest solver, anyway. There are many faction quest steps the AI will finish on its own, for example searching four ruins, leveling up a district, defeating the armies besieging its city, etc. On the other hand, some quest steps are incredibly specific, such as equipping a hero with a specific item and then searching a specific ruin while leading an army of specific size or composition, or the aforementioned quest to eradicate a city at the other end of the continent, and unless there was some fixed scripting or overwhelming weighting of the AI's priorities, they would never finish the quests. Both scripting the AI to pursue the quests or skewing their priorities runs the risk of hampering them in the long run, because AIs are not as adaptable or capable of seeing the big picture as a human player is.



Would it be nice to have an AI able to decide when to pursue quests and fully commit to them? Definitely.

Can I understand why we don't and accept that the AI plays by a different set of rules? Certainly.

ThorTillas wrote:
Hi there!



I guess there is another post somewhere in the history of this forum which talk about the same subject... but I have not the heart to look for it ^^.



In short:

Yes, AI quest solving is based on a turn trigger.

And yes, you may modify them in xml.

But they will never win by quest because the last step is never solved by an AI... Because... they are cheating... so we do not want them to win by cheating...

Moreover, they do not parley to avoid abusing the system. And they get no quest from the ruins either for the same reason.



In not so short:

We tried to think about a way of dealing with quest in EL. But at the time (and still today... snif...), we were experiencing a lot of major AI bug on the battle, economy, diplomacy, etc... And when we come up with some basic direction and started to estimate it, we decided that spending 2 or 3 month to build a real quest solver was not the best way to improve the gamer experience.



So at first we just said: Well, then we do nothing and hope that by doing "stuff" the AI will trigger the quest! (yes, programmer's proposition try to always be as "do nothing" as possible)

But the designers pointed us that Quest give you real benefits which you do not want to let on the side... So we ended up building a "cheaty" quest solver based on turn timer for each quest step.



The goal the quest solver is not to challenge the player, it's to be sure that the AI get the quest benefit as much as the player do.



I hope it answer the question... and that you won't hate our AI for that...



Cheers,

flo.

Thanks for the clarification

ShadowDragon8685 wrote:
It's difficulty-agnostic. You get various unique buildings from the faction quests, which the AI would probably be unable to complete, since some of them require some very outside-the-box thinking. For example, the Roving Clans quest requires that you take control of a territory owned by another faction, which is really problematic for the Clans since they cannot declare war. You'd either have to let an AI-controlled Clans ignore the major drawback of their faction, either generally or specifically only in this circumstance, hard-code the quest so that if the Clans are AI, that objective will never appear in a human player's territory and will guarantee that that other AI decides then and there to declare war on the Clans (regardless of how insane or suicidal that is; also, this breaks down if it's a lone AI Clans vs. an arbitrary number of exclusively human opponents,) thus allowing the Clans to wage war on them to claim the territory (this also breaks down if the other players are exclusively human, unless you want to force the player's empire to declare war regardless of the player's desire to do so and hardcode the clans not to accept a truce until they have that territory; and also disable the Drakkenling's ability to Force a truce on the Clans whilst doing this quest), hardcode the Clans to know about and pursue the Privateers strat specifically to take that one territory, or hardcode that quest itself to skip just that one particular step.



And that's one step for one faction. I really, really hope you see the problem here.

You may target for a particular victory path but things may not always turn out the way you want to favor that path. Which is why the quest victory (like all other victory types) is meant to be opportunistic where if you find the situation to favor a particular victory path as the game develops, you go for it. Granted this is a complex game where the AI may make sub-optimal moves if it ONLY pursues a quest victory to the exclusion of other victory types and different factions have different strengths. However, all these do not negate the fact that a victory condition is a victory condition meant to be achievable, not something that simply grants rewards after a certain number of turns. The dilemma and challenges faced by an AI seeing an opportunity for a quest victory would obviously be the same as a human player in this position. A "smart" AI would be able to decide whether the cost of pursuing such a victory to the very end outweighs the benefits, just as it would a human player. So back to the question: why have a broken victory condition?



Edit: That said, if the quest victory is proven to be unbalanced in favor of the human player, then it should be fixed or removed altogether. Not left as a broken victory condition.

And I always thought the quest AI was really good

idlih10 wrote:
If so, that still doesn't make sense. Doesn't the AI already get bonuses for harder difficulties? Why would the victory quest condition be made broken intentionally? It's just baffling.

It's difficulty-agnostic. You get various unique buildings from the faction quests, which the AI would probably be unable to complete, since some of them require some very outside-the-box thinking. For example, the Roving Clans quest requires that you take control of a territory owned by another faction, which is really problematic for the Clans since they cannot declare war. You'd either have to let an AI-controlled Clans ignore the major drawback of their faction, either generally or specifically only in this circumstance, hard-code the quest so that if the Clans are AI, that objective will never appear in a human player's territory and will guarantee that that other AI decides then and there to declare war on the Clans (regardless of how insane or suicidal that is; also, this breaks down if it's a lone AI Clans vs. an arbitrary number of exclusively human opponents,) thus allowing the Clans to wage war on them to claim the territory (this also breaks down if the other players are exclusively human, unless you want to force the player's empire to declare war regardless of the player's desire to do so and hardcode the clans not to accept a truce until they have that territory; and also disable the Drakkenling's ability to Force a truce on the Clans whilst doing this quest), hardcode the Clans to know about and pursue the Privateers strat specifically to take that one territory, or hardcode that quest itself to skip just that one particular step.



And that's one step for one faction. I really, really hope you see the problem here.

idlih10 wrote:
But all that is beside the point. Why have a quest victory for the AI when it's not much different from a turn victory? It's highly unfair that the AI would play by different rules to achieve the same victory type as the human player.

The way I see it, requiring the AI to solve the quest "manually" would rarely ever "increase the difficulty" for the AI, unless there were no specific algorithms to solve each quest - which could lead to the AI performing subpar since it can´t really rationalize its objectives.



I might be wrong here, I´m certainly not the most qualified to be discussing that, but I think there are three big options:

A- the developers simply provide the parameters required to solve each quest and let the AI heuristics choose to pursuit those conditions or not, which is not a lot of work from the developers but which also might lead to a weaker challenge for the player.

B- the developers provide the AI with a specific algorithm for each quest to allow it to always perform the quests to their best possibility, which not only is a lot of work but also might lead to more AIs playing "quest" than it would be desirable from an RPG point of view since they have very specific (and efficient) routines to follow.

C- the way it is right now.

ShadowDragon8685 wrote:
I think you're not understanding what the Dev said.



The AI gets to automatically complete their faction quest steps, so as not to make them miss out on the benefits (the various rewards) that you get at each step. However, because allowing the AI to achieve faction-quest victory in this manner would be cheap, they can never actually finish the very last step.



So they'll "progress" their faction quest to the last step, so as to gain the rewards from it, but they'll never finish the quest that way, thus, they will never achieve the faction quest victory condition.

If so, that still doesn't make sense. Doesn't the AI already get bonuses for harder difficulties? Why would the victory quest condition be made broken intentionally? It's just baffling.

I'm let to believe that there is some cheating involved regarding faction quests for AI factions.

While there is evidence that some quests are triggered for the AI players (yellow faction armies), some of the faction quests have very difficult objectives, like the one for the Valuters where you need to conquer city in some region, where I can be 100% sure that specific AI could not have completed such quest using regular play. Yet, their quest counter has moved forward.

So if there is a cheating, what kind of cheating is used?

It would be nice to get some feedback from developers regarding this.

Keilun wrote:
So then as I understand it, they don't get any ruin/minor faction quests which means they can never attain the high end armor/weapon techs? That's quite an advantage for human players if that's correct - even quest weapons are quite strong now.



I guess then there's the question of whether or not they'll even equip it since by all accounts I rarely find AI equipping their units particularly well.

Depends on difficulty level and starting resources. In a game as complex as this, I'll be very cautious about saying who has the advantage apart from hard bonuses the AI gets on harder difficulties. I don't know if the AI has improved in terms of equipping its units, but it knows how to put together decent armies to wear yours down by attrition even if your weapons are superior.



But all that is beside the point. Why have a quest victory for the AI when it's not much different from a turn victory? It's highly unfair that the AI would play by different rules to achieve the same victory type as the human player.

So then as I understand it, they don't get any ruin/minor faction quests which means they can never attain the high end armor/weapon techs? That's quite an advantage for human players if that's correct - even quest weapons are quite strong now.



I guess then there's the question of whether or not they'll even equip it since by all accounts I rarely find AI equipping their units particularly well.

Victory quest is not part of the automatic solver.

Guardian quest are not part of the automatic solver.

Nor are the search/pacification quest either.



Only the faction quest is part of this. (as far as I am aware of it at least).



As for balancing, I guess not having the right to use quest for pacification or to have quest instead of looting while searching ruins or temples do them more damage than having their faction quest going forward... but that my personal view... I will let WillBeFast answering from design view...



Cheers,

flo.