[GS] Devs discuss AI, World Congress and other things

[Victoria]

I'd like a proof of that, my memory disagree:

... so you are saying the dev may be incorrect?
Don’t get me wrong, they are as human as anyone, it was however quite a succinct statement, did you have visibility of IV code?

 

[Gedemon]

... so you are saying the dev may be incorrect?
Don’t get me wrong, they are as human as anyone, it was however quite a succinct statement, did you have visibility of VI code?

Well I may be wrong, it's a long time ago. For him too.

We don't need civ6 code in the context of this discussion (but hey, Firaxis, we REALLY need it for modding, TIA), we're not looking for free stuff mid game there, we're looking for it in civ4, and I don't remember such things in its code.

 

[AsH2]

To me it sems like some don't (can't) differ Civ III and Civ IV when it comes to AI matters and "stack of doom".
I recommend http://www.sullla.com/Civ4/civ4survivor4.html to study Civ IV AI in action.

 

[Jaybe]

To me it sems like some don't (can't) differ Civ III and Civ IV when it comes to AI matters and "stack of doom".
I recommend http://www.sullla.com/Civ4/civ4survivor4.html to study Civ IV AI in action.

Ooh! Sulla is ... controversial.

 

[AsH2]

Ooh! Sulla is ... controversial.

Grab some popcorn and enjoy!

 

[darkace77450]

Incidentally, that last point is part of why the oft-requested "let me barter with other players to get them to vote the way I want" isn't being implemented (and is instead covered conceptually by trading Favor).

Forgive me for bumping a two-week old topic, but I wanted to chime in on this. Trading for a Civ's diplomatic favor, in practice, doesn't have the same impact as convincing them to vote with you on a resolution. Say me and Civ A both have 60 Diplomatic Favor. If I'm allowed to negotiate for Civ A to vote with me on a resolution - and for simplicity's sake let's say we both sink all our votes into said resolution - I'm securing 8 votes (1 free vote + 1 for 10 + 1 for 20 + 1 for 30 = 4 votes * 2 = 8 votes). Under the current system, if I trade for all of Civ A's Diplomatic Favor and sink all my votes into that same resolution I'm securing 5 votes (1 free vote + 1 for 10 + 1 for 20 + 1 for 30 + 1 for 40 = 5 votes). In this scenario, should Civ A use their free vote against me in the resolution, they effectively wipe out the 1 extra vote I secured via trade, which in turn renders everything I gave for their Diplomatic Favor a waste.

 

[Ziad]

Not rly. They would have sunk 2 votes instead

 

[Dotsworthy]

My general experience is that the AI refuse to trade away their diplo favor in any circumstance. Unless they ask for a promise and it tends to only be forward settling.

 

[Pfeffersack]

My general experience is that the AI refuse to trade away their diplo favor in any circumstance. Unless they ask for a promise and it tends to only be forward settling.

The general refusal begins in the moment you acquire the first diplomatic VP (before they will trade up to 20 favor per turn and AI to you) The fast shutdown of any favor trade is still balanced around the old 10-point-goal for diplomatic victory; allowing the AI to sell you as long as you are below two VP at least would be resonable, IMO.

 

[ThunderLizard2]

The Civ IV BTS AI didn't work well because it "cheated." It worked well because it was well programmed and the game mechanics were not so overly complicated like in Civ VI. Also the 1 UPT is a dead end and they should drop it with Civ VII.

 

[PhilBowles]

The Civ IV BTS AI didn't work well because it "cheated." It worked well because it was well programmed and the game mechanics were not so overly complicated like in Civ VI.

It worked well in general because the game mechanics were simpler, certainly - Civ V and Civ VI have more sophisticated AIs struggling because they have to do more complicated things. But 'cheating' was absolutely the core of the difficulty level system - the AI was no better on Emperor than Chieftain, the difficulty was entirely a function of AI bonuses and player handicaps (the handicap system being missing in Civs V and VI - in Civ IV and predecessors a multiplier was applied to production of all kinds, so that building and science took longer for the player at higher difficulties).

 

[PhilBowles]

There’s a thread on Reddit where the FXS Devs discuss the AI and other issues. Link here. You can find some other quotes searching the Dev’s profile.

Some quotes below:

***​

“Both of these changes would fall into a camp of feature that I've tried to avoid with all of my work on Civ 6: they would be features that do absolutely nothing when used on a human player.

This is something I avoid for two reasons: 1, the obvious multiplayer implications of the game containing features that simply don't apply in that context. 2, the fact that it gives the player more tools than the AI has.

Your comment about Civ being singleplayer, with multiplayer slapped on top, is sadly kind of true. But it's something that we're trying to shy away from: we want multiplayer Civ to be as well-supported as singleplayer Civ. This is why more concrete effects have been added to systems that in base game were only for manipulating the AI: diplomatic visibility now affecting combat, alliances now having strong in-game effects beyond locking in your diplomatic state with the AI, etc.

As long as they stick to things like those mentioned that don't disrupt the single-player experience, but Civ will always fundamentally be a single-player game - few multiplayer groups would commit the needed time even without Civ's historical connectivity problems.

And, for point 2, we're very aware of the feedback that we're constantly receiving about the level of play of the AI. A lot of people look fondly back on the difficulty of the AI in games prior to Civ 5, but with 5 and 6, we made a change to have the AI actually play the same game as the human player. Previously, the game would just magically grant units and techs and such to the AI players in order to keep them competitive. Now, we may sneak the AIs extra yields or have them start on better footing, but they're not then cheating with those resources.

Which is fundamentally misconceived unless the AI can make effective use of those starting advantages to leverage an ongoing advantage throughout the game. It can't, and as a result the game has a very flat difficulty curve - it takes progressively longer to overtake the AI on higher difficulties thanks to the greater starting bonuses, but once overcome the AI represents no greater challenge at Deity than at Prince. A design paradigm based on giving AI more starting advantages and then leaving it to do its own thing relies on an unrealistic assessment of AI capabilities, since it demands - indeed - that it actually be able to play the full game. A good follow-up question to the devs would be why they implemented this change, and in particular why they kept this model for Civ VI after it generated a lot of dissatisfaction with Civ V.

Also, all the comments of this nature lumping Civ VI and Civ V together don't do anything to explain or justify why Civ VI AI is some way below Civ V's and the game considerably easier as a result.

This whole point answers a HUGE amount.
All those civ IV player posts whinging about a bad AI while all the time their version was cheating it’s pants off
It does bring up a very important fact, that 2 players in civ VI playing the same game could have quite different levels of cheating used against them which in a way invalidates their competition games. Intriguing.
This one clarification really limits their argument about a bad AI compared to VI.

It's also worth noting that the vast majority of players who seem dreamy-eyed about Civ IV in the community apparently didn't play the version of the game Firaxis released, at least in the version they look back on so fondly - they're talking about Rhyse and Fall, which was essentially a rebuild. This is also something this dev comment may ignore - I don't know how the AI mods scaled difficulty or whether it took the same approach as the base game, but it sounds as though the dev is commenting from a perspective that takes no account of modders' input and hasn't incorporated lessons learned from them into Firaxis' process.

 

[orasis]

Has anyone ever seen the AI use aircraft to bomb (not WMDs) or pillage anything?

 

[Gedemon]

It's also worth noting that the vast majority of players who seem dreamy-eyed about Civ IV in the community apparently didn't play the version of the game Firaxis released, at least in the version they look back on so fondly - they're talking about Rhyse and Fall, which was essentially a rebuild. This is also something this dev comment may ignore - I don't know how the AI mods scaled difficulty or whether it took the same approach as the base game, but it sounds as though the dev is commenting from a perspective that takes no account of modders' input and hasn't incorporated lessons learned from them into Firaxis' process.

I can admit that it's possible that my memory of civ4 make it look better than what it really was, but I didn't play it with AI mods and never used Rhyes and Fall (the latter being a shame as I surely would have loved it)

And I'm still waiting for someone to point me to where this was in civ4 code, because I've not found it:

Previously, the game would just magically grant units and techs and such to the AI players in order to keep them competitive. Now, we may sneak the AIs extra yields or have them start on better footing, but they're not then cheating with those resources.

And yes, even if it was not, the question remains as why not using it, at least as an option, IMO "dynamic difficulty" sounds better than "huge bonus at start and fixed yields bonus the rest of the game". Both are cheating AI, but the first at least solve the "early game too hard/late game too easy" issue.

 

[Casworon]

The modder Fearsun made a great adaptive difficulty mod that is worth checking out. I would love to see more modders try to tackle the adaptive difficulty idea and see their different approaches. There are more mods now that have progressive difficulty that give the AI per Era bonuses. To try to spread out the AI power level though even they struggle to deal with player snowballing. Having difficulty be adaptive to player progress seems like the perfect solution to having the game be challenging all the way through.

There was a great podcast i listened to recently with the game designer of Civ IV talking about the new 4x game he is making. He made the point that the game has to provide some push back, and that with no challenge the experience just feels kinda pointless. To me this rings completely true and is i feel why so many people quit civ games half way through once they get to the snowball point. They know that they are eventually going to win without any challenge and the game becomes a 'next turn' button presser simulator.

I've been having more fun with the game now I've changed my fundamental approach to difficulty. Turning off time victory and lowering the amount of turns avaliable so defeat becomes not achieving any of the other victories in time rather than the AI beating me too them, which never happens even on the highest difficulties.

I doubt firaxis would change their approach to game difficulty to make it adaptive and dynamic when every game in the series has used flat yield bonuses. But I'm surprised more modders haven't tackled this.

Its possible with scripts and doable without the DLL and is probably the biggest thing that will have a positive impact on the game experience outside of getting DLL access to change the underlying AI code. Perhaps we should try to get together a community 'Dynamic Difficulty' team and bounce around ideas and test things to try and find the perfect balance. Having scripts to provide the AI with extra units depending on the difference in power or perhaps player victory progress when they declare war on you for a start would make AI invasions much more challenging than they are now.

 

[PhilBowles]

I can admit that it's possible that my memory of civ4 make it look better than what it really was, but I didn't play it with AI mods and never used Rhyes and Fall (the latter being a shame as I surely would have loved it)

And I'm still waiting for someone to point me to where this was in civ4 code, because I've not found it:

Not the code, but a summary of the AI 'cheats' in Civ IV, which were mostly production modifiers that favoured the AI and/or handicapped the player:

https://civilization.fandom.com/wiki/Difficulty_level_(Civ4)

The comment about free techs etc. seems to have been either a case of the dev not recalling the specific bonuses the AI received, or talking about prior versions of the game generically (some of which may have done just that). That doesn't invalidate his central point, was that the AI was given bonuses to help it compete because it was poor at playing the game, and certainly didn't vary its play by difficulty.

 

[Gedemon]

Not the code, but a summary of the AI 'cheats' in Civ IV, which were mostly production modifiers that favoured the AI and/or handicapped the player:

https://civilization.fandom.com/wiki/Difficulty_level_(Civ4)

The comment about free techs etc. seems to have been either a case of the dev not recalling the specific bonuses the AI received, or talking about prior versions of the game generically (some of which may have done just that). That doesn't invalidate his central point, was that the AI was given bonuses to help it compete because it was poor at playing the game, and certainly didn't vary its play by difficulty.

Yes, I've posted a similar link already for civ4 on the previous page, you'll find the same for civ3 there.

But if you want to directly compare from one source, just use:

https://civilization.fandom.com/wiki/Difficulty_level_(Civ6)
https://civilization.fandom.com/wiki/Difficulty_level_(Civ5)
https://civilization.fandom.com/wiki/Difficulty_level_(Civ4)
https://civilization.fandom.com/wiki/Difficulty_level_(Civ3)
https://civilization.fandom.com/wiki/Difficulty_level_(Civ2)
https://civilization.fandom.com/wiki/Difficulty_level_(Civ1)

If his central point were "that the AI was given bonuses to help it compete because it was poor at playing the game"* in civ1-4 that would be false too, because it's still the case for civ5-6, and it still certainly doesn't "vary its play by difficulty".

*I don't think it is, the central point is "previously, the game would just magically grant units and techs and such to the AI players in order to keep them competitive" in civ1-4. But that never existed AFAIK, and that's what I'd like to be pointed to in civ4 code.

 

[Infixo]

Has anyone ever seen the AI use aircraft to bomb (not WMDs) or pillage anything?

Here

 

[PhilBowles]

Yes, I've posted a similar link already for civ4 on the previous page, you'll find the same for civ3 there.

But if you want to directly compare from one source, just use:

https://civilization.fandom.com/wiki/Difficulty_level_(Civ6)
https://civilization.fandom.com/wiki/Difficulty_level_(Civ5)
https://civilization.fandom.com/wiki/Difficulty_level_(Civ4)
https://civilization.fandom.com/wiki/Difficulty_level_(Civ3)
https://civilization.fandom.com/wiki/Difficulty_level_(Civ2)
https://civilization.fandom.com/wiki/Difficulty_level_(Civ1)

If his central point were "that the AI was given bonuses to help it compete because it was poor at playing the game"* in civ1-4 that would be false too, because it's still the case for civ5-6,

He didn't claim it was good at playing the game in Civ V or VI - he was using the change in design paradigm to explain why the AI in Civ V and VI is worse than that in the older games.

and it still certainly doesn't "vary its play by difficulty".

No, the difference is in where the bonuses are applied. In Civ VI the bonuses are entirely front-loaded - the AI gets a bonus at the start (mainly extra settlers) that let it get an advantage early. In Civ IV and earlier they scaled through the game. Civ V was a hybrid of both systems (which is likely partly why Civ VI is so much worse than Civ V AI-wise despite the similar design philosophy).

The problem I have with the dev's explanation is that he doesn't explain why they made this change, if not deliberately to make the game easier. The Civ VI system is obviously worse than the older ones both in theory and practice, which proved (yes, even in Civ V despite the grumbling - it was a step down from Civ IV, but still a challenge if you didn't just go on a killing spree. I don't tend to play in optimised ways and won only about half the time on Immortal and rarely on Deity in Civ V). I only lose in Civ VI - which I play exclusively on Deity - if I get Zerg-rushed very early either by excessive barbarians, barbarians plus a civ, or two civs in a joint war, get an especially poor start, or am focused on winning a particular victory or meeting a specific objective rather than winning the game per se - as with my recent Khmer game where I was going for the religious victory achievement having avoided that victory condition in the past

*I don't think it is, the central point is "previously, the game would just magically grant units and techs and such to the AI players in order to keep them competitive" in civ1-4. But that never existed AFAIK, and that's what I'd like to be pointed to in civ4 code.

It was true in at least one game, but he seems to misremember which one: Civ V gave the AIs starting techs, and it spawned extra units for the AI throughout the game. If anything this was the source of some of the complaints - the AI wasn't doing anything more 'unfair' than the bonuses in the previous games, but because it was getting units that were visible on the map (and in some cases could be seen spawning), the 'cheating' that the AI did 'under the hood' in earlier games was more visible to the player.

 

[Gedemon]

He didn't claim it was good at playing the game in Civ V or VI - he was using the change in design paradigm to explain why the AI in Civ V and VI is worse than that in the older games.

if by "change in design paradigm" you mean going from "AI that play the game" to "AI that try to win the game", I would agree, but I've some difficulties to read that from the developer's post.

No, the difference is in where the bonuses are applied. In Civ VI the bonuses are entirely front-loaded - the AI gets a bonus at the start (mainly extra settlers) that let it get an advantage early.

And +32% science/culture/faith, +80% production/gold for the whole game. I don't call that "entirely front-loaded"...

In Civ IV and earlier they scaled through the game.

how exactly ?

It was true in at least one game, but he seems to misremember which one: Civ V gave the AIs starting techs, and it spawned extra units for the AI throughout the game. If anything this was the source of some of the complaints - the AI wasn't doing anything more 'unfair' than the bonuses in the previous games, but because it was getting units that were visible on the map (and in some cases could be seen spawning), the 'cheating' that the AI did 'under the hood' in earlier games was more visible to the player.

Civ6 give the AI starting techs/civics boost.

And again, I don't remember any code about spawning extra units in civ5, those complaints looks like incorrect conclusions based on individual observations.