From Easy to Deity. A new way to code the Ai difficulty?

[Lazy sweeper]

Will we see a new way in how Ai is coded in the difficulty levels in Civ VII?

Young bloods doesnt accept loosing. They will complain, it is too hard, they killed my archer, they took my capital...
sells will drop, devs will panic, and within one week of launch they will nerf the Ai to Oblivion if it was not already..
We all know this, it is how business goes, new customers is the target audience, move on...

So what I propose is try to determine an alternative method for coding the difficulty level.
Instead of cheating AI, they will boost the player stats. And nerf the Ai, on Easy or Settler level.

Chieftain will see no boosts nor nerfs. Ai and players are on equal ground.

From Chieftain on, Ai should improve, without cheating. Ai will stay put, get no boosts untill Deity. Only then, Ai will receive the same boost player got at Easy.
At each level of difficulty, Ai should receive one upper level of aggressiveness, and or expansionst/economical/diplomat buff
something like, on Easy, it will never use an Army to conquer a city, nor deploy artillery for both defensive and attacking, but just use some melee skirmishes into
player territory, and only a complete undefended city might be taken.

Chieftain, they will use everything at their disposal, but not consider the weakest points, and not prepare an invading army in advance with odds in mind...
From chieftain on, Ai will produce enough troops, to alter the odds of a battle in their favor, and use guerrilla tactics exponentially more and more advanced.
At Deity, they will make sure they got 100% win odds in their favor, in both attack and defence.
Military alliances should be one core aspect of tactics, that's why they should be really easy to obtain...
At one point, only alliances can alter the odds...

The only cheat, at this point, is how they calculate the odds... players should use spies to do that...
If Ai is not able to do that, in my view, that means that basic math should be adjusted for everyone, untill the system works.
As opposed to alter the underlying math, in favor of Ai, in order to mantain superiority over players.
The Civ VI system failed at providing a consistent feeling of difficulty progression in the eyes of many.

What will the future holds?
Also, could probability be introduced instead of just true or not in any code argument??
What to do modders say about this?
I haven't even seen a config file of Civ ever since Civ III...
Happy new Year 2025 folks!

PS: If all Leaders have hidden Agendas and start dislike each other, it is very unlikely they will try to make Military Alliances...
Getting the core math right is essential. The no warmongering mod in Civ VI didn't solve the problem of incompetent Ai.
I don't know the nuances to further inquire on the issue but I suspect it has smg to do with unorthodox Leaders complexity...

 

[Krikkit1]

I mean that would all be nice, but there’s no way they have the resources to make any differences in AI behavior by difficulty levels , other than maybe a little less aggression v the player on the easiest levels.

 

[Taefin]

A two parameter difficulty (based on cheats) would be nice, where you can set the difficulty at the start of the game and the rate of increasing difficulty, akin to Stellaris and smooth difficulty mod in VI, would be nice. Presumably easily moddable, and I would probably use mods to fine tune it even if they included such a system in vanilla.

The AI might not need as aggressive of cheating in VII, since the adjacency bonuses seem easier to optimize and they won’t need builders.

Based on the livestreams, it looks like the generals deploy-and-attack-in-same-turn upgrade will be abusable, letting the player take out several enemy units before they counterattack. I think human players will like this feeling, and I think the AI will need extra production to offset this with larger armies, especially to stay competitive. I’d also imagine that an AI willing to rush in without full numbers (and with the cheats to resupply) will give players a sense of fending off waves of attacks before striking back.

For a non-cheating AI to be competitive, I imagine they would need to be empowered to snowball on the opposite side of the map as the player, and I have seen more aggressive AI in VI (when I lowered the threshold to attack, and weakened cities down to where VII looks like it will be), actually conquer each other and expand.

 

[CivKirk64]

I traditionally play on Prince manly because I'm not competitive, don't want it to be hard, but I want the computer to be as challenging as it can be without cheating (higher yields or reduced costs)
If the AI ca be more difficult on a higher tier without having to give them boons that would be a more fun challenge.

 

[SupremacyKing2]

The game Phantom Brigade (a turn based mech combat game) let's you customize the difficulty by tweaking specific aspects of the game. See below:



I've played the game and I really like that level of customization in the difficulty setting because it let's you tweak the aspect that you need help with or that you want to make more challenging while leaving other aspects of the game the same.

Civ7 could do something similar. You could have sliders for unit strength, unit cost, science output, culture output, influence output, gold output, upgrade cost from towns to cities, build cost for buildings and wonders, aggression of independent peoples, mastery techs on or off etc... This would let the player choose what aspects of the game to make easier or harder. I think this would make difficulty levels a lot more interesting. Maybe a newbie player struggles keeping up with the AI in science, they could play a normal game but give themselves a science boost. Maybe another player struggles with combat so they play a normal game but just make unit strength higher to make combat easier for them.

 

[GoneFishingX]

That's a nice wishlist but basically impossible to achieve. Starcraft 2, a game with far less complexity and variance took about a decade to finally get really solid professionally skilled AI opponents as a difficulty.

Civ will never have AI that can challenge a skilled player mechanically. The only real tool is buffs (cheats). The game is just too complex. I imagine a bug hurdle in game AI (i.e. chess robots) will be a game like Civ. You would prolly need a massive amount of processing power to make a machine play as well as a human Civfanatic.

Also, there is no market for it. Few players actually want an AI opponent who can play at their caliber. Playing against a human player can often be obnoxious enough as they can exploit systems or engage in cheese.

 

[Krikkit1]

That's a nice wishlist but basically impossible to achieve. Starcraft 2, a game with far less complexity and variance took about a decade to finally get really solid professionally skilled AI opponents as a difficulty.

Civ will never have AI that can challenge a skilled player mechanically. The only real tool is buffs (cheats). The game is just too complex. I imagine a bug hurdle in game AI (i.e. chess robots) will be a game like Civ. You would prolly need a massive amount of processing power to make a machine play as well as a human Civfanatic.

Also, there is no market for it. Few players actually want an AI opponent who can play at their caliber. Playing against a human player can often be obnoxious enough as they can exploit systems or engage in cheese.

I think the additional complexity of Civ v. Chess or Starcraft 2 isn't the biggest issue. The issue is
1, the instability of the game, Chess didn't stop getting made because a Chess 2 came out, Chess also hasn't receive updates to the way it is commonly played for the last few centuries
2. as mentioned the market (number of Chess players >> Starcraft 2 players >Civ players)

 

[GoneFishingX]

I think the additional complexity of Civ v. Chess or Starcraft 2 isn't the biggest issue. The issue is
1, the instability of the game, Chess didn't stop getting made because a Chess 2 came out, Chess also hasn't receive updates to the way it is commonly played for the last few centuries
2. as mentioned the market (number of Chess players >> Starcraft 2 players >Civ players)

Both solid points. Plus the lack of a competitive environment to Civ compared to SC2 and Chess.

 

[stealth_nsk]

Honestly, I'm all hands for a single unified victory scale. All other settings shouldn't be tied to difficulty:

1. Difficulty level is the measure of player skill if you play a competitive game. Someone who plays of Emperor difficulty is usually stronger player than the one, who plays on Chieftan. That's kind of additional level of competitiveness and understanding your strength. Multiple difficulty settings don't give clarity.
2. The concept of improving AI is kind of hell of a programming. Writing one AI is a lot of work work, writing several of them is too huge. Remember - Civilization is not top-level chess, AI doesn't just play to win, it also roleplay. It also has to look smart for human player (because there's no such thing as actually being "smart" for AI), etc. This means each AI needs to be manually tested by humans if you have many of them.
3. Each additional setting multiply the amount of testing you need as you need to test them against all other settings. And difficulty really depends on ALL other setting, so you'll need to see how each combination of difficulty settings interacts with each combination of map size (because some AI version may settle to slow or too fast or not far enough), map types and so on. So, only the settings which truly improve game variety should be there. And additional difficulty settings are not part of this.

P.S. Side note, when talking about low-level chess AI, it's programmed to make human-like mistakes, so coding it has more similarities for AI in a game like Civ, although still a lot of differences.

 

[King of Prussia]

Adding/changing modifiers is the easiest and fastest way to make AI more or less challenging. This is what the vast majority of games do. It can be however not as fun to play against. The are generally easy to expose as options tho, giving you more customization.

Adjusting the weight to certain AI logic where possible can lead to a more challenging AI, but also a lot of weird decisions.

Locking off certain mechanics based on difficulty for the AI is also possible. This can lead to the game playing very differently which may negatively impact newer players going up in difficulty. It could also make it easier for new people to gradually learn the game. With that said, spending a ton of time trying to make the AI use a mechanic just to have it be not used does kinda feel like a waste of dev time.

Of course, how much should the AI roleplay and how much should it try to win. If there is one really good unit, the AI should spam the crap out of it to win. On the other hand, is it fun as the player killing the same unit over and over? Doomstacking in total war comes to mind.

Anyways, AI is super complicated. Hours and hours to make it just work. Hours and hours and hours are spent fixing random issues that crop up. Many more hours trying to make them use the mechanics well. Adding even more things that can impact it will take even more time.

 

[Lazy sweeper]

akin to Stellaris and smooth difficulty mod in VI, would be nice. Presumably easily moddable, and I would probably use mods to fine tune it even if they included such a system in vanilla.

AAhhh... the slavery system of Stellaris... please don't add salt to the open wound...

Civilization only ref to slavery is in the civilopedia under 'Plantations'...
Stellaris has different systems of slavery...

 

[Lazy sweeper]

That's a nice wishlist but basically impossible to achieve. Starcraft 2, a game with far less complexity and variance took about a decade to finally get really solid professionally skilled AI opponents as a difficulty.

Civ will never have AI that can challenge a skilled player mechanically. The only real tool is buffs (cheats). The game is just too complex. I imagine a bug hurdle in game AI (i.e. chess robots) will be a game like Civ. You would prolly need a massive amount of processing power to make a machine play as well as a human Civfanatic.

Also, there is no market for it. Few players actually want an AI opponent who can play at their caliber. Playing against a human player can often be obnoxious enough as they can exploit systems or engage in cheese.

It's not impossible.
It is a big problem, and it needs to be addressed at the source if they want to eliminate it.
There is nothing more important than opening the config file for Ai, and make it adjustable, leave modders the possibility to alter the functions.
It will make devs work lighter, player base happier, and stand the test of time.
Civ VI will not stand the test of time.

Also your assumptions there is no market for it is irrilevant. Devs wants to make a good game, always.
Only publishers that don't play their games, and hire external studios to just make a product for the market, follows the money blindly.
Firaxis has devs that plays their games.

 

[Lazy sweeper]

Honestly, I'm all hands for a single unified victory scale. All other settings shouldn't be tied to difficulty:

1. Difficulty level is the measure of player skill if you play a competitive game. Someone who plays of Emperor difficulty is usually stronger player than the one, who plays on Chieftan. That's kind of additional level of competitiveness and understanding your strength. Multiple difficulty settings don't give clarity.
2. The concept of improving AI is kind of hell of a programming. Writing one AI is a lot of work work, writing several of them is too huge. Remember - Civilization is not top-level chess, AI doesn't just play to win, it also roleplay. It also has to look smart for human player (because there's no such thing as actually being "smart" for AI), etc. This means each AI needs to be manually tested by humans if you have many of them.
3. Each additional setting multiply the amount of testing you need as you need to test them against all other settings. And difficulty really depends on ALL other setting, so you'll need to see how each combination of difficulty settings interacts with each combination of map size (because some AI version may settle to slow or too fast or not far enough), map types and so on. So, only the settings which truly improve game variety should be there. And additional difficulty settings are not part of this.

I re-iterate the concept, the only thing devs should do, is to open moddability for the Ai difficulty settings, adding as many variables as possible, including adding probability.
Players will do the testing for free.
Modders will do adjustments for free.

Unless Firaxis will not turn into Bethesda it should be fine.

 

[evonannoredars]

I mean that would all be nice, but there’s no way they have the resources to make any differences in AI behavior by difficulty levels , other than maybe a little less aggression v the player on the easiest levels.

Didn't they talk about in one of the streams how much bigger their AI-focused department was this game than in any previous game? Can't remember where tho, might've been with a youtuber or I might be misremembering.

 

[GoneFishingX]

There is nothing more important than opening the config file for Ai, and make it adjustable, leave modders the possibility to alter the functions.

Also your assumptions there is no market for it is irrilevant. Devs wants to make a good game, always.
Only publishers that don't play their games, and hire external studios to just make a product for the market, follows the money blindly.
Firaxis has devs that plays their gagames.

So, if you're talking about letting modders have access to the cpu script that runs their AI then thats doable and a totally reasonable request.

Again, to reiterate. There is little to no market or reason for an AI oppnent that can compete with a civfanatic. It isn't even good game design or reasonable allocation of resources.

As for having a CPU opponent that scales in skill based on difficulty. It can be done, but it just isn't worth it most if the time and also not an easy thing to build. I would rather them spend the money and resources on a well built CPU opponent that just gets more buffs. That way they are still capable at lower levels, but just get more stuff to send at higher levels.

You comment about how publishers push things out the door before they are ready. It's very true, but keep in mind firaxis can only spend so much time without income before they shut down. Even if they were self published with a lot of money they've still gotta ship something. Self publishing might even have made them push the game out even sooner cause they'd be taking out loans.

 

[Core Imposter]

The reason I basically don't play the latter iterations of Civ or a lot of other games is the bad AI. In my mind, you can continue to Civ26 or Civ103 and you will never actually have a new game until someone puts together a good AI. Without that, it's all window dressing.

There is little to no market or reason for an AI oppnent that can compete with a civfanatic. It isn't even good game design or reasonable allocation of resources.

It wasn't good design, either, for American manufacturers to engineer and build refrigerators and freezers, many of which are still in service 40 years later. The planned obsolescence model in the game industry today is all about not building the necessary artificial intelligence that would allow for a game to repayable and become a staple of a generation's entertainment regimen.

"It's not personal Sonny. It's strictly business".

The reasonable expectation is that developers provide an AI that can use the features of the game at a level that gamers will find challenging to beat and some randomness to keep it interesting. But they don't, it's not a matter of falling short, it is purposely creating illusions. Which act like a taste of crack, hooking us chumps and bringing us back to be sheared over and over. It's not the wolf we sheep fear, but the sheep dogs that are always nipping at our heels. Yes, we are stupid creatures.

 

[Core Imposter]

Didn't they talk about in one of the streams how much bigger their AI-focused department was this game than in any previous game? Can't remember where tho, might've been with a youtuber or I might be misremembering.

Are they really building a better AI or is it just a case of simplifying the game down to a level that makes the AI seem competent? I guess it will be seen.

 

[Gorbles]

Young bloods doesnt accept loosing. They will complain, it is too hard, they killed my archer, they took my capital...
sells will drop, devs will panic, and within one week of launch they will nerf the Ai to Oblivion if it was not already..
We all know this, it is how business goes, new customers is the target audience, move on...

When has this . . . ever happened before? What evidence do we have between low-end AI difficulty has any impact on volume of sales (which, historically, as a Chieftan enjoyer, has been pretty trivial)?

This seems pretty core to your OP and I simply don't get it. Everyone was young once. Everyone hated losing games once. Some folks still do, nomatter how old they get.

 

[GoneFishingX]

The reason I basically don't play the latter iterations of Civ or a lot of other games is the bad AI. In my mind, you can continue to Civ26 or Civ103 and you will never actually have a new game until someone puts together a good AI. Without that, it's all window dressing.

It wasn't good design, either, for American manufacturers to engineer and build refrigerators and freezers, many of which are still in service 40 years later. The planned obsolescence model in the game industry today is all about not building the necessary artificial intelligence that would allow for a game to repayable and become a staple of a generation's entertainment regimen.

The AI in civ has never been good. They've always lied, cheated, and stealed their way into feeling competative. I won't deny civ 5+6 had more issues than other versions, but keep in mind that 1-upt made the game vastly more complicated. I think it's a trade off between 1-upt and doomstacks. Neither is the better system.

I think it's intensely tin foil to think firaxis makes bad computer opponents so you buy the next game. That doesn't even make sense. People play games until they get tired and a new game gives them a fresh experience. Also they have never cut support or shutdown an older civ game. That would be actually toxic.

 

[gdr_willter]

Are they really building a better AI or is it just a case of simplifying the game down to a level that makes the AI seem competent? I guess it will be seen.

When the game is simplified enough for AI and enjoyable enough for human at the same time, I'll call it as the "well-designed game with better AI" - even if the AI itself is not so different from the previous titles.