Discussion in 'Civ6 - General Discussions' started by Steve Eric Jordan, May 18, 2019.
I had no idea Vox Populi was for Civ 6.
Nobody said that. It's been only said that the AI in Civ VI is simply ridiculous, below any decent standard. Everywhere. Just look at Steam reviews, look at multiple threads here or the net etc. It's been also often said that in contrast to other companies, Firaxis does very little about it. Unfortunately! So here it is:
"A call to the developers"
I wish it were.
I'm all a bit tired from seeing this attitude in the community.
Most of us agree in one thing, we enjoy the game and we wish we could have an even greater game. But all these threads are everything but constructive critizism.
I have not seen a thread where the community puts together behaviors that the IA should do but doesn't or shouldn't do but does, bugs they want to be corrected or specific ideas for FXS to built.
Instead i have seen dozens of threads with people mostly gathering to say how garbage it is, how unplayable it is and so on. Two expansions and seven or eight patches, where the AI has been always touched have happened (maybe only a bit, but incrementally i think that is significantly better now than upon release). However not a word has changed in the way we talk about it. This is all but constructive criticism. And i think this makes more difficult for FXS to address the issue.
Also I think the VP mod appears in these conversations constantly in a way I don't like. Yes two guys created a mod with a good IA for Civ-V. I know, we should also ack that it took them several years of job with a stable and finished game to do so. I don't think FXS is using more manpower than two guys to work in the IA, probably not full time and in a constant change of game rules environment.
Yes they should be praised for their work and FXS should be criticised for the lack of resources dedicated to the AI. But in these point that should be granted and we should move to a more reasonable way of discussion.
Cause i think there are a lot of positives in FXS and Civ. I'm happy that they keep working on the game, and happy that they provided in the past pretty much the entire code for players to improve upon. I don't think any other big studio does that. And i hope they keep doing the same. That's the reason i enjoy so many of FXS games and actually think there are good reasons to defend the company, even if the game still has some bugs (that are being constantly but slowly fixed) a poor interface (that is being constantly improved) and a lackluster AI (that is also being constantly tweaked, but is still disappointing).
All in all, i just wanted to say that I would like to see constructive conversations for these issue. I would like to see suggestions and ideas for improvement. Talk more about the features that already nice AI mods for Civ-VI like Real Strategy bring to the table and how could be expanded upon by FXS... And so on
I actually think, that the combat AI can be greatly improved, without changing much, and without affecting other core systems. With a more cappable AI on wars, and an improved pace on the late game ala 8 ages of peace, the game would be already much better.
Improve pathfinding and make the AI send more units together for attack and defense opperations
Make AI more aggressive in the middle-late game.
Make AI engage more actively in air defensive and offensive operations in late game. Also increase AI ability to support terrestrial operations with naval and aerial forces.
Make AI commit more aggressively with the wars they declare. They should pursue peace when they are involved in an undesired war, also they should not declare a war they will or cannot actively engage in. But they should commit longer to a war they declared. Maybe till they achieve their objective, or till they understand they are being defeated or are unable to achieve the objective. Another reason to stop a war would be a long stale situation.
Improve AI ability to coordinate defensively when they are attacked. The AI should be focused on actively build and buy units to support the front in a coordinated effort from all nearby cities.
Make AI commit more to the victory type they pursue. They should pay less attention to victory paths that are disabled and pursue only victory paths compatible with agendas and personalities, but also the status of the game.
Decrease general preference for religious buildings and increase preference for science buildings.
@o Siyeza. Try to read some forum posts before posting yours and making assumptions. This is the one i posted a long time ago in https://forums.civfanatics.com/thre...ith-regard-to-war.643233/page-5#post-15407536
People here are extremely passionate about the Civ game and a lot of them have posted really good ideas for improvement in the game. All MY complaints are made with a view that someone from Firaxis might one day have a look at one of these post and things might change.
The AI might have improved "incrementally" in every area BUT WAR. Is the AI able to use land units properly let alone naval or air units? NO. Does the AI run circles around your cities and gets its units killed even with numerical advantage? YES. Does the AI makes settlers and wonders even when you are at the city's doorsteps? CHECK. So do enlighten me how has the AI improved "incrementally" in this respect?
Is the AI able to use land units properly let alone naval or air units? Sometimes YES.
Does the AI run circles around your cities and gets its units killed even with numerical advantage? NO (I didn’t experience this when given numeric advantage).
Does the AI makes settlers and wonders even when you are at the city's doorsteps? This actually can be easily fixed. Anyone posted this on Bugfix forum?
Human player attack when AI’s army is already depleted by previous battles with other AI. So they have no chance against you. It doesn’t really matter what they do tactically. So what to do here? I mean how could this be fixed? Perhaps AI should peace out with all other neigbours before invading a chosen one? Did anyone try attacking an otherwise non-depleted previously peacefull AI which just amassed its army in the middle of its territory? I’ve seen plenty of such huge passive army in mid or late game as well, but never attacked them.
To be awfully clear, I did not say Nobody has ever made a constructive claim about AI. Just expressed the general feel of the threads, that quickly become a "this is the worst thing i have ever seen" contest.
To be even more fair I have read hundreds of threads on the forum but not every single one of them, I hope that is not a requirement here. I was making an observation on the way this particular topic usually goes without pulling any names. I obviously did not pretend to claim that valuable suggestions have not been made by you or a lot of people. But rarely conversations appear around these suggestions. That i have to clarify this point in order to navigate this conversation is quite telling, I have to say.
Is quite baffling that the tone of these posts escalate the way they do. Given that we don't disagree in what we want or what areas of the game we want to be improved.
I will give you some examples on how the AI is better now than upon release on war: It builds planes and ships, it uses ships, it sometimes uses planes, mostly escorts settlers and other units, understands that sometimes it has to prioritize attack over healing, uses appropriated units to garrison in cities, uses ranged units mostly well, builds and tries to use siege and bombard units in order to attack cities with walls.
I'm not saying the AI is good now, it is better.
A civ6 AI would need to know how to:
- know human is the biggest threat to victory and be less diplomatically favourable to human
- fight in the 3UPT environment, not just 1UPT
- create unit clusters that rely on melee units and siege vehicles rather than having 58 catapults and one spearman
- move, replenish & regroup unit clusters
- recognize hexes important for either conquest or defense that need to be held and by which unit type
- prioritize city walls and roads to border settlements
- attach great generals & siege units, and to which unit
- scout with great admirals and scout lands held by neighbors
- scout with spies by moving them from city to city
- make predictions on unit losses and take them into calculation for unit production
- what districts to prioritize and in which city
- focus on production when it comes to terrain improvements
- how it really wants to win
- when and what to trade for
- manage city state investments, befriend and levy city states offensively/defensively
- properly recognize threat from human attack based on known factors
- properly recognize known beelines and conquest pushes when human reaches certain techs
- aggressively pursue districts & great people for chosen victory type
- aggressively beeline military techs regardless of victory type
- have a pool of money as backup for levies and unit upgrades
- bribe other AIs into it's wars
- prioritize science, income and culture over crappy districts like overblown amounts of encampments and holy sites
Yes, 1UPT is just a fraction of the problem and no, Civ6 at least where AI is concerned is far from greatness
Can’t help but wonder how many human players actually tick all of these consistently (or ever). Personally, I can’t recall ever building walls....
"create unit clusters that rely on melee units and siege vehicles rather than having 58 catapults and one spearman"
It does it. The 58/1 is when it’s army got depleted during defense (because catapults dont participate in defense). So instead of deleting these units it sends them to attack, since it doesnt loose anything with it anyway. So this is one way how this can be a no bug.
Yes, but unlike (good) human players, AI needs walls. I personally love walls myself, especially if enemy units are dripping in one by one. Enables me to send my unit clusters elsewhere with no worry.
Encountering 10 catapults and 1 courser doesn't mean I depleted/killed 40 of its other units, but rather that it overbuilds catapults. The ratio of melee/mounted vs siege/useless is ridiculous. And no rams or towers ever. In fact, I have yet to see any siege past catapults, as AI doesn't know "how to niter".
"doesn't mean I depleted/killed 40"
You sure? I encountered this also. But in all cases other units were depleted by me or another AI player.
To me you hit the bullseye when you mention about handling units as a group, anticipate loses and being able to regroup replace and position is a very basic feature the AI needs to improve on.
I actually think that biggest flaw of the combat AI, is being narrow minded. In a war, production and grouping of units is a key element the AI does badly.
You need to replace your losses and attack and move units in groups.
When attacking, consider how many units you would need and start the attack once you have produced and deployed the army you need, while keep producing more units in a couple of nearby in anticipation for losses.
In deffense all your production should focuss on building units, all units shoud come in reinforcement of your lines. And u have to bring your units in groups to positions where u are strong to hold the attack.
The AI needs to do the following:
* Build walls earlier
* Be aware of players more than 50% done with a victory condition and act accordingly. Now this actually existed but was hated because it boinked diplomacy, so maybe only if the AI isn't friendly.
* Be aware if itself is going to win. If it can say go for a science victory, the path is more or less set on what to get once it gets a spaceport. If it is 50% @ domination win, it should focus on war and even backstab people.
* Be very stingy in trade. The higher the level, the more it should ripoff the player. In Civ 4, you have to give the AI a fortune to make it do anything. Don't tell me this is impossible. But in more concrete terms, no trading the AI luxuries right away at the start. Don't value Great Works that much.
These numbers are arbitrary and can be adjusted. Everything else (tactical ability, district placement) is not that important-- it's more about the grand strategy and Civ AI has never needed to be able to outmicro the player... is not esport. Also the Civ 6 AI already makes tons of units and with the exception of siege units doesn't really use them that badly..
The game itself lso needs to still further adjust (slow down) tech/civc development. ATM, you are forcing the AI to keep up with a player that will win in 150-250 turns and that will always be skewed. People are going to break the game anyways, but the less problems, the better. Personally, I would nerf chop scaling by at least half or more and greatly increase the cost of mid-late game civcs/techs.
To me, the K-mod for Civ 4 had pretty much the perfect AI that still more or less acts like the default AI but acutally plays the game. It didn't need to massively outwit the player to win the game or micromanagement (though there was still a lot of that) but it could be pretty nasty even on prince.
Part of the big issue from a developer perspective is that we actually disagree in what we want and which areas should get priorities, as we disagree on the definition of a "good AI".
Some people want an AI that is able to win the game at all cost, following the game's rules the best it can. Some people want an AI that is able to provide a challenge without causing frustration by abusing the game's rules. Some people want an AI that is able to RP a Civilization with more logical diplomacy. Not even talking of the game's mechanisms that would have to be changed to fit each type of AI (we can't have a good RP AI with the current diplomatic rules and victory conditions IMO)
Which means for an AI to please everyone, it would require a lot of options (and as many different code sections)
And by trying to make an AI that won't displease anyone (larger possible audience), the developers end up with a middle ground AI that units all players in wanting a "better" one... until we start talking specifics of said "better" AI, that is.
For example, I would not want to play against this AI:
That's why IMO it would be much easier for them and to directly release the source code so that any modder can make it's own "perfect" AI.
I am not sure what is wrong with the trade part. We know the ai will never be smart enough to avoid being conned so it should err on the side of caution. In civ 5 for example you could sent the ai to war for a few gold per turn. I am not really sure what would be wrong with an ai that doesn't do that or buy things it doesn't need.
Maybe I worded it badly? When I mean it should be stingy against players, I mean other AIs too; currently the AI does treat other AIs like other players.
This is like one of the easiest fixes and far more reasonable than expecting the ai to outsmart a player in war.
not you, my bad, I assumed "humans players".
and I agree, but I'd still prefer to overhaul the diplo first, so that the logical solution to a runaway civs is to allow smaller civs to create logical/balanced alliances and actually be able to win as allied instead of everyone dogpiling the leader.
for example starting the game with 10-20 civilizations, ending it with the victory of one of 3-4 superpowers, some still being single nations/players, some being alliances of multiple nations/players
Ah, sorry. I actually strongly object to a "screw the human" bias that existed in say, Civ 4 and before. That reminds me too much of games like Mario Kart where you get punished for doing too well. One of the GOOD things about Civ 6's AI is that it actually plays by the same rules you are, eg. it suffers from unhappiness and also gets angry at other AIs for going against agendas and warmongering. Seeing the AI in past civ games ignore these mechanics is why I did not have much respect. If there is a bias it should strictly be against the leader, and not just against you. I can't speak for everyone, but I imagine many players don't want an opponent that doesn't try to win but is just there to grief them.
Whatever malus one may give, at least diplomacy in Civ 6 is "fair". Now balanced is another issue.
We do have this case with emergencies-- shared vision is actually really nice to deal with this runaways. I'm of mixed opinion on this and don't mind either way, but do you feel that emergencies lack something?
The return of the Permanent Alliance is something I'd like to see, but I suppose some will complain about it making things easier- unless other AIs would do the same thing. I remember that happening in the past.
Proper AIs should be able to beat each other up (if they want). And they sometimes can, even now. But given the opportunity, yes, AI should beat up the human player as well, not be eternally locked in an alliance, especially if the human is close to winning.
Enable a long and tedious way of getting such an alliance with a human player that the AI wins as well (and is thus incentivized to be a part of the team).
I, as a human, would never keep such an alliance, so why should the AI? That's human bias.
Separate names with a comma.