4X Make Civ(likes) Great Again! Brainstorming design thread

[IncendiaryDevice]

Technology does have a lot of effects that are disruptive. For example, Gunpowder resulting in the end of the Landed Nobility as the premier armed class, and eventually bringing in the age of mass conflicts, where vast armies of plebs with guns can beat all but the best armies of mercs and retainers.

I like the idea of introducing disasters, and them being weakened or placated by other moves.

I do agree that things like dark ages and disasters do need a change in design philosophy so that its not only possible, but fun, to come back from them. We need mechanics that make it boring to save-scum, and more fun to go with blows. "Losing is fun" and all that.

There's different kinds of realities involved in a civ-like game. There's realities which make sense, such as ensuring that a tank is more likely to kill a spearman than visa-versa, but then there's realities that make no sense in a civ-like game, such as having different rulers with different skills/agendas every 10 or 20 years. Putting in too much destructive stuff is in a similar vein to the latter, in that the whole point is not to mimic reality, but to create an idealised game where all of your decisions are accountable only to you. For example, in Civ3, your city could get destroyed by a random volcanic eruption, however, it would still have been your decision to build a city near a volcano. Having lagre amounts of completely random events designed purely "for the lols", for want of a better description, goes completely against the Civ concept at its very heart, unless you can, in some way, prepare for, build towards, mitigating said destructive random elements. At which point your game becomes more boring because players don't want to build towards preventing % chance random outcomes, they want to build towards "Muh global empire of power and glory".

I can't say I ever really save-scum any civ games and I have no idea why anyone would. The only time I reload is if I've forgotten to do something that I intended to do when my turn started but by the end of the turn I'd briefly forgotten before clicking end-turn. And this would be, like, once every other game. I think civ games are already "fun to lose" games and most people are ok with seeing their city die to an AI onslaught, because making a new map is so quick, easy and the start of the game is the most enjoyable anyway.

 

[Cael]

Technology does have a lot of effects that are disruptive. For example, Gunpowder resulting in the end of the Landed Nobility as the premier armed class, and eventually bringing in the age of mass conflicts, where vast armies of plebs with guns can beat all but the best armies of mercs and retainers.

I like the idea of introducing disasters, and them being weakened or placated by other moves.

I do agree that things like dark ages and disasters do need a change in design philosophy so that its not only possible, but fun, to come back from them. We need mechanics that make it boring to save-scum, and more fun to go with blows. "Losing is fun" and all that.

There's different kinds of realities involved in a civ-like game. There's realities which make sense, such as ensuring that a tank is more likely to kill a spearman than visa-versa, but then there's realities that make no sense in a civ-like game, such as having different rulers with different skills/agendas every 10 or 20 years. Putting in too much destructive stuff is in a similar vein to the latter, in that the whole point is not to mimic reality, but to create an idealised game where all of your decisions are accountable only to you. For example, in Civ3, your city could get destroyed by a random volcanic eruption, however, it would still have been your decision to build a city near a volcano. Having lagre amounts of completely random events designed purely "for the lols", for want of a better description, goes completely against the Civ concept at its very heart, unless you can, in some way, prepare for, build towards, mitigating said destructive random elements. At which point your game becomes more boring because players don't want to build towards preventing % chance random outcomes, they want to build towards "Muh global empire of power and glory".

I can't say I ever really save-scum any civ games and I have no idea why anyone would. The only time I reload is if I've forgotten to do something that I intended to do when my turn started but by the end of the turn I'd briefly forgotten before clicking end-turn. And this would be, like, once every other game. I think civ games are already "fun to lose" games and most people are ok with seeing their city die to an AI onslaught, because making a new map is so quick, easy and the start of the game is the most enjoyable anyway.

I have to agree with this.

The whole lure of Civ is right there from the very first one: To build an empire to stand to the test of time.

What is the point of building an empire that the computer can destroy with a few bad random numbers and there is nothing you can do about it because they are all random events? Yes, it is realistic. Sooner or later, a comet is going to come and laminate the lot of us into the bedrock, but to put that into a Civ game is basically randomised rocks fall, everyone dies.

 

[Accolades]

4x is one of the most technologically constrained genres of video games, because the core component of a 4x game (the AI) is something just making a smarter AI isn't a viable option.

Mate come on we live in a world where an AI that taught itself to play Dota 2 1v1 mid beat 99.999% percentile players, another one beat the #1 GO player a few years ago, and they're on the road driving vehicles. Even fairly pedestrian machine learning should more than up to the task of playing 4x games. All it takes is a decent score system, AI knows high score at the end of the game is what it's aiming for, either plays itself a million times or analyses human replays, learns to play the game at high level without cheating. Even sans machine learning the AI in the last few civ games should have been way better than it was, the fact that it was improved by mods proves it's totally doable, Firaxis is just incompetent and there's no excuses for releasing a game with AI as broken as Civ VI's was

 

[Mr. Pink]

4x is one of the most technologically constrained genres of video games, because the core component of a 4x game (the AI) is something just making a smarter AI isn't a viable option.

Mate come on we live in a world where an AI that taught itself to play Dota 2 1v1 mid beat 99.999% percentile players, another one beat the #1 GO player a few years ago, and they're on the road driving vehicles. Even fairly pedestrian machine learning should more than up to the task of playing 4x games. All it takes is a decent score system, AI knows high score at the end of the game is what it's aiming for, either plays itself a million times or analyses human replays, learns to play the game at high level without cheating. Even sans machine learning the AI in the last few civ games should have been way better than it was, the fact that it was improved by mods proves it's totally doable, Firaxis is just incompetent and there's no excuses for releasing a game with AI as broken as Civ VI's was

You are severely underestimating the cost and expertise required to operate and debug learning AI. Game developers are not neural network experts. Just the machine that alphaGo runs on costs 25 million dollars. Google spent at least 400 million dollars on DeepMind. None of this will be coming to a personal computer for at least a decade. (although AI rental might be a thing soonish)

Expecting your average game developer to implement machine learning in a strategy game magnitudes more complex than go is like expecting your average civil engineer to build an atom bomb from scratch.

Francis is incompetent for too many reasons to list.

 

[Bad Jim]

4x is one of the most technologically constrained genres of video games, because the core component of a 4x game (the AI) is something just making a smarter AI isn't a viable option.

Mate come on we live in a world where an AI that taught itself to play Dota 2 1v1 mid beat 99.999% percentile players, another one beat the #1 GO player a few years ago, and they're on the road driving vehicles. Even fairly pedestrian machine learning should more than up to the task of playing 4x games. All it takes is a decent score system, AI knows high score at the end of the game is what it's aiming for, either plays itself a million times or analyses human replays, learns to play the game at high level without cheating. Even sans machine learning the AI in the last few civ games should have been way better than it was, the fact that it was improved by mods proves it's totally doable, Firaxis is just incompetent and there's no excuses for releasing a game with AI as broken as Civ VI's was

You are severely underestimating the cost and expertise required to operate and debug learning AI. Game developers are not neural network experts. Just the machine that alphaGo runs on costs 25 million dollars. Google spent at least 400 million dollars on DeepMind. None of this will be coming to a personal computer for at least a decade. (although AI rental might be a thing soonish)

Expecting your average game developer to implement machine learning in a strategy game magnitudes more complex than go is like expecting your average civil engineer to build an atom bomb from scratch.

Francis is incompetent for too many reasons to list.

That said, I think 4X AI could be a lot better. The reality of AI in games:

• Marketing a game on the strength of it's AI is difficult because it's almost impossible to prove that your AI is good yet it's easy for someone to post a video on Youtube of your AI doing something stupid. Most people are very bad at deciding whether AI is smart, often praising it when random movement resembles tactical maneuvering and lambasting it for rational decisions. Game journalists are even worse.
  
• The games industry pays well below the market rate wherever it can get away with it, including programmers. Many firms such as stock market traders pay top dollar for machine learning experts. Machine learning experts are thus pretty rare in the games industry.
  
• Games are often designed with no thought to how the AI is supposed to play them. The obvious example is 1 unit per tile in recent Civ games. A doomstacking AI can beat a good player simply by sending enough doomstacks, but a 1UPT AI needs a tiny bit of tactical savvy. However there is a trade off between AI competence and features, and if I were to run a "GalCiv 2 vs MOO 2" poll on here, the result would not be a compelling argument in favour of AI competence.

One thing I'd like to see is some sort of chess style AI to avoid tactical blunders. Of course, combinatorial explosion would be an issue when large numbers of units are involved, but I think if you took a random sample, say 1000 random moves, and looked at 1000 random responses to each move, the result would look a lot less stupid than what you usually see.

 

[Average Manatee]

You certainly could implement some simple machine learning. Just limiting yourself to figuring out a good tech path, city construction list, or unit composition would be fairly simple and improve most 4x games tremendously.

 

[Bad Jim]

You certainly could implement some simple machine learning. Just limiting yourself to figuring out a good tech path, city construction list, or unit composition would be fairly simple and improve most 4x games tremendously.

If the AI knows what it is teching towards, what it is trying to do with it's cities, and what enemy units it will face, that's easy. Unfortunately, this requires the AI to decide what it is teching towards, what it needs to do with it's cities, and who it will need to fight, which of course requires the AI to actually understand the game and have a plan, which is where this turns into a massive expensive AI project. Otherwise, the AI will only be slightly better than if you just gave it a few canned build orders.

 

[Average Manatee]

You certainly could implement some simple machine learning. Just limiting yourself to figuring out a good tech path, city construction list, or unit composition would be fairly simple and improve most 4x games tremendously.

If the AI knows what it is teching towards, what it is trying to do with it's cities, and what enemy units it will face, that's easy. Unfortunately, this requires the AI to decide what it is teching towards, what it needs to do with it's cities, and who it will need to fight, which of course requires the AI to actually understand the game and have a plan, which is where this turns into a massive expensive AI project. Otherwise, the AI will only be slightly better than if you just gave it a few canned build orders.

Figuring out build orders is like 90% of 4x games. MoO2, for example, could see vastly improved AI if it did the same research and build plan every player does (i.e. rush auto facts and so on).

 

[Bad Jim]

Figuring out build orders is like 90% of 4x games. MoO2, for example, could see vastly improved AI if it did the same research and build plan every player does (i.e. rush auto facts and so on).

Building what every player does can be done with canned build orders. It might even be better that way, because when the AI thinks for itself it usually shits the bed, then performs every household chore except for washing bed sheets. I think AI programmers might actually be trying too hard on this one, and that simply following a build order would get superior results most of the time.

 

[Average Manatee]

Building what every player does can be done with canned build orders. It might even be better that way, because when the AI thinks for itself it usually shits the bed, then performs every household chore except for washing bed sheets. I think AI programmers might actually be trying too hard on this one, and that simply following a build order would get superior results most of the time.

You are assuming that developers know how to play their own game enough to figure out the ideal build orders. This has almost never been the case. Instead just tabulate results from real games played against real people and have your learning AI figure out a few optimal aggressive or economic strategies as necessary. Bonus: You can effortlessly adjust the AI's intelligence downward for easier difficulty levels rather than just giving the player cheats or handicapping the AI.

 

[Bad Jim]

Building what every player does can be done with canned build orders. It might even be better that way, because when the AI thinks for itself it usually shits the bed, then performs every household chore except for washing bed sheets. I think AI programmers might actually be trying too hard on this one, and that simply following a build order would get superior results most of the time.

You are assuming that developers know how to play their own game enough to figure out the ideal build orders. This has almost never been the case. Instead just tabulate results from real games played against real people and have your learning AI figure out a few optimal aggressive or economic strategies as necessary. Bonus: You can effortlessly adjust the AI's intelligence downward for easier difficulty levels rather than just giving the player cheats or handicapping the AI.

Or they could just put the build orders in plain text files so it's easy for players to mod them once they figure the game out.

I think its a bit harsh to say that devs can't figure out decent build orders. Sure, the community will find better build orders, but even Firaxis could write build orders that were somewhat sensible. Inefficient is okay, because a production advantage can make up for that, and the player isn't paying close attention. What you are trying to avoid is the AI building stuff that makes no sense, yet it can still pump out units, because that breaks the illusion that it is playing by the same rules as you.

Also, I don't think it's a good idea to make an AI easier by making it more derpy. Sometimes the derpy AI will make better choices than the regular AI, and then you have a difficulty inversion. Cheats and handicaps are more consistent.

 

[The Brazilian Slaughter]

They could do build orders according to situation.
So the AI files changing depending on what it needs. Say, a bellicose AI getting ready for war would only build things that make it able to support and build a proper army, and then it starts spamming units until it thinks it has enough to start fighting. But, when its full-war, it just throws all the production in making more and more units.

A builder AI would have a pre-done build order fitting peaceful builders - buildings that multiply its riches and internal power, and only a few units for internal defense.

Paradox Games have editable AI files for every single nation, and they can easily have multiple files, set up by time or AI switches.

 

[rezaf]

Along with just throwing massive amounts of resources at it and excempt it from game rules (such as fog of war or range limitations) one of the easiest and thus most popular approaches at AI design is exploiting its far superior abilities when it comes to multitasking, number crunching and resisting boredom.

Unfortunately, this leads to the effect that, in order to play effectively so you can beat the AI, you also have to numbercrunch or engage in boring repetitive actions, which - at least for me - can suck out all the fun.

I have to admit, though, that the 1UPT decision was maybe a stroke of genius seen from this angle, as it robbed the AI of the ability to properly benefits from ANY advantage or bonus it was granted.
Not by accident even fairly average players tend to play the late Civ games on the highest difficulties...

 

[The Brazilian Slaughter]

but then there's realities that make no sense in a civ-like game, such as having different rulers with different skills/agendas every 10 or 20 years.

Why not? How long the average ruler lasted? I like the idea of having multiple rotating leaders, so from time to time a nation's behavior and bonuses change. Say, some whatever dudes get a epic leader that goes on a conquest spree, or a huge empire gets a retard ruler that runs the thing into the ground and in the end its a mess of civil wars and foreign invasions. Maybe the option to get rid of bad leaders by somesort of "Coup" button, but you have do it right or you run the risk of a civil war.

(this could be better if all civilizations were more interconnected, so you could generate civilization collapse dominoes)

Sci-Fi and Fantasy mods/games really work better with the immortal ruler thing - you can just say they use tech or magic to live forever. The faction leaders in SMAC doing constant rejuvenation treatments is a good example, because it ties directly to lore and gameplay as these treatments put them into trances that lead to contact with the Planetmind. Later on, you get things like the Immortality Vaccine.

For example, in Civ3, your city could get destroyed by a random volcanic eruption, however, it would still have been your decision to build a city near a volcano. Having lagre amounts of completely random events designed purely "for the lols", for want of a better description, goes completely against the Civ concept at its very heart, unless you can, in some way, prepare for, build towards, mitigating said destructive random elements. At which point your game becomes more boring because players don't want to build towards preventing % chance random outcomes, they want to build towards "Muh global empire of power and glory".

I don't think they should be "random events for the lulz" but rather the sort of things that an empire should have to deal with - like natural disasters, or climate change.

What is the point of building an empire that the computer can destroy with a few bad random numbers and there is nothing you can do about it because they are all random events? Yes, it is realistic. Sooner or later, a comet is going to come and laminate the lot of us into the bedrock, but to put that into a Civ game is basically randomised rocks fall, everyone dies.

Let's try a good example from RL, the Romans - who are the closest thing to an Empire out of a 4X game in RL.

Why the Romans stop conquering? Because they reached their maximum extension - at best, they could have maybe took over Germania, but Arminius kind of fucked up that chance. It was as big as it could be, pretty much.

One example of how "random events" would work, is how the barbarian invasions were set up by climate change causing adversity in the Steppes, which lead to the Huns migrating westward, which lead to proto-slavs and germans being pushed westwards into the Roman Empire. The climate worsened and became colder at the time, there was starvation at the time.

 

[Cael]

Let's try a good example from RL, the Romans - who are the closest thing to an Empire out of a 4X game in RL.

Why the Romans stop conquering? Because they reached their maximum extension - at best, they could have maybe took over Germania, but Arminius kind of fucked up that chance. It was as big as it could be, pretty much.

One example of how "random events" would work, is how the barbarian invasions were set up by climate change causing adversity in the Steppes, which lead to the Huns migrating westward, which lead to proto-slavs and germans being pushed westwards into the Roman Empire. The climate worsened and became colder at the time, there was starvation at the time.

What happens in RL doesn't mean it automatically becomes a GOOD thing to put into a game. As I said before, randomised crap that severely affect your empire without you being able to do a single thing about it is basically randomised "rocks fall, everyone dies". It is NOT FUN. For a GAME, realism is not the goal. Fun is.

 

[The Brazilian Slaughter]

Let's try a good example from RL, the Romans - who are the closest thing to an Empire out of a 4X game in RL.

Why the Romans stop conquering? Because they reached their maximum extension - at best, they could have maybe took over Germania, but Arminius kind of fucked up that chance. It was as big as it could be, pretty much.

One example of how "random events" would work, is how the barbarian invasions were set up by climate change causing adversity in the Steppes, which lead to the Huns migrating westward, which lead to proto-slavs and germans being pushed westwards into the Roman Empire. The climate worsened and became colder at the time, there was starvation at the time.

What happens in RL doesn't mean it automatically becomes a GOOD thing to put into a game. As I said before, randomised crap that severely affect your empire without you being able to do a single thing about it is basically randomised "rocks fall, everyone dies". It is NOT FUN. For a GAME, realism is not the goal. Fun is.

"The climate worsens and barbarians invade your empire fleeing from a bigger barbarian horde" Where you see "cannot do anything about it"?

 

[Cael]

"The climate worsens and barbarians invade your empire fleeing from a bigger barbarian horde" Where you see "cannot do anything about it"?

Climate worsen in Civ means that land somewhere is being changed to less productive lands. Not something you can do anything about.

Barbarian horde suddenly invading your land when you are busy with other civs is tantamount to "good game". In Civ4, I have seen event spawn barbarians wipe out entire AI civs. That is BAD.

 

[Bad Jim]

In Civ4, I have seen event spawn barbarians wipe out entire AI civs. That is BAD.

That's part of the fun. As long as the player gets sufficient warning, big world events make the game interesting. Of course they are not everyone's cup of tea, but it is usually possible to turn them off.

 

[Cael]

In Civ4, I have seen event spawn barbarians wipe out entire AI civs. That is BAD.

That's part of the fun. As long as the player gets sufficient warning, big world events make the game interesting. Of course they are not everyone's cup of tea, but it is usually possible to turn them off.

You don't. You get the event and AT MOST, two turns later, you lost a city. They appear at your civ's border, which in the early game is pretty darned small.

 

[Bad Jim]

In Civ4, I have seen event spawn barbarians wipe out entire AI civs. That is BAD.

That's part of the fun. As long as the player gets sufficient warning, big world events make the game interesting. Of course they are not everyone's cup of tea, but it is usually possible to turn them off.

You don't. You get the event and AT MOST, two turns later, you lost a city. They appear at your civ's border, which in the early game is pretty darned small.

I think The Brazilian Slaughter is really talking about grander events. Not "a boat lands near your city and three horses pop out" but something like "barbarians will land somewhere on the map every turn for 50 turns". In that sense, you are warned about the barbs because they spawn every game and you know you should be building some military to deal with them.

 

[Cael]

In Civ4, I have seen event spawn barbarians wipe out entire AI civs. That is BAD.

That's part of the fun. As long as the player gets sufficient warning, big world events make the game interesting. Of course they are not everyone's cup of tea, but it is usually possible to turn them off.

You don't. You get the event and AT MOST, two turns later, you lost a city. They appear at your civ's border, which in the early game is pretty darned small.

I think The Brazilian Slaughter is really talking about grander events. Not "a boat lands near your city and three horses pop out" but something like "barbarians will land somewhere on the map every turn for 50 turns". In that sense, you are warned about the barbs because they spawn every game and you know you should be building some military to deal with them.

The problem is that if it only affects one or two civs, then you have just basically removed them from the game. This is something that happens in the Civ4 Fall From Heaven mod, where a powerful early game unit spawns 75 turns into the game (strength 5 when the best you are likely to have is probably Strength 3, 4 if you are lucky). That one scripted event is known to end games, and it has certainly made a lot of my games easier when it is the next door civ that copped him in the teeth.

 

[IncendiaryDevice]

This whole desire to add random to a civ-like game goes completely against both the thread title and the concept of a civ game. Way to make a mockery of your own thread "make civ-likes great again", implying that they were once great and have declined, to then suggest 'improvements' that were never meant to be a part of a civ-like experience. You don't make civ-likes "great again" by gradually transforming them into roguelike strategy games just because you personally like roguelike mechanics. Perhaps instead The Brazilian Slaughter you made this thread because you don't like civ-like games rather than because you thought they were once great. The fundamental concept of a civ game is to simply manage an empire, a realism based fantasy empire, for 6,000 years, where its success or failure depend entirely upon the decisions you make and for there to be a myriad of choices regarding those decisions on every turn. Civs and civ-like don't need random, are not designed with random in mind (beyond starting map), and for whom random is merely a light spice rather than an actual game mechanic. If your suggestion on how to improve them is just to add more and more 'light spice', then that is exactly the problem civ 6 is having...

 

[Bad Jim]

This whole desire to add random to a civ-like game goes completely against both the thread title and the concept of a civ game. Way to make a mockery of your own thread "make civ-likes great again", implying that they were once great and have declined, to then suggest 'improvements' that were never meant to be a part of a civ-like experience. You don't make civ-likes "great again" by gradually transforming them into roguelike strategy games just because you personally like roguelike mechanics. Perhaps instead The Brazilian Slaughter you made this thread because you don't like civ-like games rather than because you thought they were once great. The fundamental concept of a civ game is to simply manage an empire, a realism based fantasy empire, for 6,000 years, where its success or failure depend entirely upon the decisions you make and for there to be a myriad of choices regarding those decisions on every turn. Civs and civ-like don't need random, are not designed with random in mind (beyond starting map), and for whom random is merely a light spice rather than an actual game mechanic. If your suggestion on how to improve them is just to add more and more 'light spice', then that is exactly the problem civ 6 is having...

The exact problem Civ 6 is having is the AI. Nothing to do with spice.

Ideally, a Civ game is the sort of thing you want to sink 1000+ hours into, and that won't happen if every game plays exactly the same. It needs to be mixed up a little.

To say that Civ games are not random is ridiculous. The map is random, your starting location is random, what you get from villages is random, the AI goes to war at random, barbarians appear at random, and in the earlier games it got wonders at random. They are about doing your best with whatever you get. They have far more in common with roguelikes than you appreciate.

 

[IncendiaryDevice]

This whole desire to add random to a civ-like game goes completely against both the thread title and the concept of a civ game. Way to make a mockery of your own thread "make civ-likes great again", implying that they were once great and have declined, to then suggest 'improvements' that were never meant to be a part of a civ-like experience. You don't make civ-likes "great again" by gradually transforming them into roguelike strategy games just because you personally like roguelike mechanics. Perhaps instead The Brazilian Slaughter you made this thread because you don't like civ-like games rather than because you thought they were once great. The fundamental concept of a civ game is to simply manage an empire, a realism based fantasy empire, for 6,000 years, where its success or failure depend entirely upon the decisions you make and for there to be a myriad of choices regarding those decisions on every turn. Civs and civ-like don't need random, are not designed with random in mind (beyond starting map), and for whom random is merely a light spice rather than an actual game mechanic. If your suggestion on how to improve them is just to add more and more 'light spice', then that is exactly the problem civ 6 is having...

The exact problem Civ 6 is having is the AI. Nothing to do with spice.

Ideally, a Civ game is the sort of thing you want to sink 1000+ hours into, and that won't happen if every game plays exactly the same. It needs to be mixed up a little.

To say that Civ games are not random is ridiculous. The map is random, your starting location is random, what you get from villages is random, the AI goes to war at random, barbarians appear at random, and in the earlier games it got wonders at random. They are about doing your best with whatever you get. They have far more in common with roguelikes than you appreciate.

I actually talked about that in the post you quoted...

 

[Bad Jim]

I actually talked about that in the post you quoted...

That's why I quoted it.