[R&F] Difficulty

[DWilson]

I'll say that the AI is programmed pretty oddly, and the game difficulty changes don't always reflect your difficulty.

For example, I legitimately lost (at least enough to give up) a leve 1 or 2 game due to barbs swarming worse than they ever had on me in a deity/immortal game. I've had a prince game where an early rush by the AI made the game way more difficult than my typical emperor game in which I have peaceful neighbors. In contrast, a peaceful game on deity allowed me to build up and expand rapidly to the point that I quickly matched, then surpassed the AI's starting bonuses.

It seems like the AI is only good at an early rush. If they do it, my game is going to be vastly harder. If they do and I am prepared enough to over come, I'll win every game. If they do it and I am not prepared, the game can be a toss-up. If they don't, even without trying to be optimal I will at worst match the AI's science and cultural development, and generally far outstrip them.

Possibly counter-intutitively, this has led to me under-preparing intentionally in games I want to be harder, to invite the sort of attack that makes the game challenging. If anything, to program the higher difficulties I would just make it more likely that the AI will attack early on.

What that would unfortunately do is only re-emphasize the unfortunate truth that the AI will never be able to maintain it's lead, and simply surviving that stretch is generally enough to win.

This is why I strongly advocate, in a system that relies on AI bonuses for difficulty, that the bonuses are not stacked up to start the game (as they currently are), but rather applied throughout the game to help the AI keep up. Applying all the bonuses at once means your options are limited until you catch up, at which point you can do whatever you want. Spreading them out means that you can play as you like more easily (without the do or die early set up of specific strategies), but still be challenged consistently throughout the game.

Problem is, spread out bonuses will probably feel more gamey as the AI suddenly gets boosts mid game. Well, you can't please everyone.

 

[Trav'ling Canuck]

Expanding should come at a high cost, being able to defend your entire empire with 4 units should be unwinnable, having no culture, or faith, or science should hurt your chances a lot. The issue is you can just pick a victory and beeline for it before the AI can do anything.

Very well said. I still feel the game suffers from having a variety of separate victory conditions instead of different paths to a consistent victory objective (being acknowledged by all other civs as the global leader).

However, some of my issue with this would go away if the AI could pursue those different victory conditions more efficiently than it does now. Then on higher difficulty levels you might have to make an effort to delay the religious or cultural victory of the AI at the same time as pursuing a science victory. That would help the game to feel like a more cohesive whole.

 

[ShakaKhan]

I don't know much about AI design; I know you can't make an AI think which is what gives the player such a big advantage. But I think there are a few things that should be pretty simple to make happen that would make the AI more of a threat. Foremost, observing what the AI has when I stroll through them, they're obviously just sandboxing directionless for the whole game. Play a game as Macedon- you start to pay attention to which districts are where and where the wonders are. All the civs, regardless of their bonuses, have holy sites everywhere yet they can't organize a way of achieving religious victory. They have numerous encampments which is stupid because the human player will approach them from the direction where they don't have to engage them. They have wonders, particularly the early game wonders, whose bonuses aren't doing anything significant for them- the quintessential example is Petra in a city with no other desert tiles, but also civs building the Oracle, Hanging Gardens, and Stonehenge and not doing anything with them except locking the human player out of getting it. News flash AI: it's not difficult to beat you without the Hanging Gardens.

Instead, it should be pretty simple to code the AI so that it's working towards beating the human player to a victory condition. When you come across a specialist civilization, they should be working towards their bonuses. When you encounter the Kongo, they should have a theater square in every city. Similarly, when you come across Russia, they should have a Lavra in every city and enough theater squares to house the great works they provide- one thing an AI can do (and better than any human) is math- have them calculate how many turns until they produce a great writer and start constructing a theater square and amphitheater so that it's finished in time to house the great works. When you come across Scythia, they should plant their original 3 cities, then build another 2 or 3 before they reach horseback riding, and then have them make all their cities shift production to saka, and then when they have 2 horses, switch to horseman and then swarm the opposition.

For the civs that are more well-rounded and not geared towards a specific victory condition, randomly assign them a "victory coding" to work towards one of the four victory conditions.

 

[acluewithout]

@ShakaKhan All good points. But as an aside, I don’t think the AI actually needs to be programmed to beat the player or achieve victory (although that is one approach). The AI just needs to be programmed to create a challenge. So, it might make sense for the AI to build encampments or Holy Sites, not to help the AI win, but to make it harder for the player to conquer those cities (with military or religion). Equally, if the AI wars with the player, it doesn’t have to because it’s going for a Dom victory, but could just be to make your life hell.

Indeed, if the AI was solely focused on winning, then I don’t think there would be any room for Diplomacy. The AI would never cooperate or form alliances, and would probably not even play the game in an interesting way - every AI would just settle 8 cities and chop out spaceships.

The keys things are (1) the AI must have some logic to its actions (winning, agenda, whatever ), otherwise there’s not much strategy for the player and certainly no diplomacy and (2) whatever it does, the AI should do it well.

That said, some AI probably should drive towards some victory type, if only to force the player to have to stop that AI, and to create some urgency to victory. And either way, the AI does do some stupid stuff, particularly around wonders. So, mostly I’m agreeing with you.

@Trav'ling Canuck I’m more and more sold on the AI being the real issue, which I think has been your point for a while. The more I look at it, the more I think so much of the mechanics in the game rely on or a geared to the AI creating the needed challenge. Part of the problem I think is that the AI needs a leg up sometimes, if only to counter the whole snowballing thing. But increasingly I agree the bigger problem is that the AI doesn’t know how to play the game (or, more accurately, doesn’t know how to play the game in a way that creates fun and challenge for the player).

 

[ezzlar]

What I admire about the Rhyes mod in Civ 4 is how it balances all aspects of the game creating a unique challenge. Rhye didnt depend on an uber smart AI to create a challenge. Most of the challenges were within the own empire, in a way that the player would create most of its own problems. A large empire with a huge military wasn´t guaranteed a win.

The situation were every city had to be carefully evaluated wether the net benefits were positive is balance at its best. No ICS. Late era spawns takes care of frontloading of difficulty. Science penalty for huge empires is also a good balance. City costs increasing exponentially is also great for the economic game.

The dependence of a smart AI would lessen if some smart mechanics were introduced.

 

[acluewithout]

@ezzlar Agreed.

That said, I think the current mechanics are very much geared to only two things providing challenge - the Map and the AI. You’re right - the game could introduce challenges from other sources (eg governance). But it’s not geared to do that at the moment.

 

[Trystan CT]

Fearsun has some fantastic mods that create challenges for the payer that the AI doesn't face. It really changes the feel of the game away from a PvP with badly programmed AI opponents towards a more PvE experience. Which i feel is the direction the 4x genre should move in. Since getting effective AI opponents is very difficult to program. With Fearsuns mods the AI difficulty is also adaptive to your actions rather than getting flat yield bonuses. Though they still get the bonus to production

 

[Starwars]

I find that there are also problems with how the AI conducts wars later on in the game. Everyone knows about the early rushes which can be very dangerous, especially after the game has improved after patches. The AI can be decisive and deadly.

But, the other day I had a game where I had not gotten to the "catch up" point with my neighbour and I was really slow in getting my empire going. Had no defense to speak of. So, neighbour declares war, and in swoops a terrifying force of Cavalry armies, the very first turn they surround my city that was closest to their borders. "Uh oh" I thought, "this is basically a game over". But nope, on the following turns the AI broke up its troops, sent them all across my empire. Sure, pillaging a bit here and there but made it possible for me to slowly pick them off one by one and I could actually end the war in my favor.
Now, I was so far behind that the Cavalry armies could've completely wrecked my city, walls and all in a turn or two. I was *completely* unprepared in military and behind in science and yet the AI didn't manage to accomplish a single noteworthy thing, and like I said, even ended the war with me getting a bunch of gold per turn. And, if it had been even slightly... decisive, it could've game-overed me *easily*.

It's just so disappointing when stuff like that happens.

 

[c4c6]

Problem is, spread out bonuses will probably feel more gamey as the AI suddenly gets boosts mid game. Well, you can't please everyone.

Yes! 'frontloading' isn't sufficient. But it is better digested by people, who like the illusion, that they are playing the same game as the computer

So Boni in the beginning are better tolerated as the starting conditions aka difficulties. Mid game Boni are more easily denounced as "cheating" ... maybe making developers more hesitant.

As a modder&player I have no problem at all about "cheating" myself

 

[Trav'ling Canuck]

@Trav'ling Canuck I’m more and more sold on the AI being the real issue, which I think has been your point for a while. The more I look at it, the more I think so much of the mechanics in the game rely on or a geared to the AI creating the needed challenge. Part of the problem I think is that the AI needs a leg up sometimes, if only to counter the whole snowballing thing. But increasingly I agree the bigger problem is that the AI doesn’t know how to play the game (or, more accurately, doesn’t know how to play the game in a way that creates fun and challenge for the player).

And to that, I also always come back to the game still being in an interim state of development. If after the community has figured out the best ways to play the game and the final rules are set post the last expansion, the AI isn't tweaked at that time to create a more fun and challenging environment for the player, one where there's a reason to worry about the diplomacy game, then I'll truly be disappointed.

What I admire about the Rhyes mod in Civ 4 is how it balances all aspects of the game creating a unique challenge. Rhye didnt depend on an uber smart AI to create a challenge. Most of the challenges were within the own empire, in a way that the player would create most of its own problems. A large empire with a huge military wasn´t guaranteed a win.

I'd love to see the game go in that direction, too, as that's both more historically accurate, in my opinion, and creates game play that would be more interesting to me. There's some indication, between the Loyalty and Happiness system, that the development team may be inclined to take some steps in this direction, but for the most part, their focus seems to be on making the game fun by adding lots of variety. Also, there seems to be a large contingent of fans who want actions to be uniformly good, not to come with consequences.

 

[Acken]

I find the obvious bad balance much more a problem than the AI. First because it makes player choices less interesting and I have also always argued that your AI would improve as a side effect if you eliminate noob traps it always fall into by making it actually valuable. You cannot have perfect balance of course but some things are obvious.

 

[acluewithout]

And to that, I also always come back to the game still being in an interim state of development. If after the community has figured out the best ways to play the game and the final rules are set post the last expansion, the AI isn't tweaked at that time to create a more fun and challenging environment for the player, one where there's a reason to worry about the diplomacy game, then I'll truly be disappointed.

Yes, that’s what I’ve assumed too. I’ve also assumed the explanation for the delay in fixing balance issues is the same. I guess let’s see what happens.

 

[Victoria]

I believe that giving the AI +100% production when building units would compensate to a degree for the lack of unit upgrade and re-arming.
The game is better than it was with regard to intelligence, the game is a lot easier purely due governors and golden era’s and emergencies that was not balanced by loyalty as it messes with the AI and so sadly the early gains on deity get caught up way too fast.
The AI probably struggles a lot with timeouts to keep the game fast. Lots to consider.
New toys is keeping the game easy.
You need to be careful what you ask for in this world although one suspects the toys would be thrown into the cot regardless.

 

[acluewithout]

New toys is keeping the game easy.

This. So much this.

Loyalty was a good move. It does potentially make things harder for the player, and it does add at least a patina of “governance”. But the AI does isn’t equipped for it - either in terms of knowing how to deal with loyalty, or having the right bonuses.

The other changes though just doubled down on the players ability to focus on a few key yields without consequences. I do increasingly think improving the AI would make the game more challenging - but so long as you can just bear down on a few key yields or spamming strategies without consequence, the game is never actually going to be challenging regardless of the AI (ie a better AI would likely just result in the AI rushing science victories as well).

I’m not totally opposed to “new toys”. The game probably does need a few more more before it feels “complete” (a very subjective thing). But what the game desperately needs is a thorough, er, “review”, I guess. FXS need to give the balancing, AI, and gameplay a hard look, and really make sure everything “works together”.

A long time ago I used to play tabletop rpgs. Good times. Anyway. One thing I noticed about the various games was (1) randomness tends to work against players , ie the more things you dice roll, the more the player is likely to eventually “lose” no matter how good they are, and (2) the more options you give players, even if those options aren’t all that powerful individually, the more OP players become because they stack them in unexpected ways.

You see this in the game now. There really is only randomness is the early game, and that’s partly why the early game is challenging but the late game is not. Equally though, there are now so many options, you can craft some very specific strategies which let you ignore most of the game and romp home. I mean, spamming campuses and chopping space ports was powerful before, but you can basically do that so much more effectively now.

I think @Trav'ling Canuck is right - a lot of work needs to be done on the AI (and I think balancing) and that can probably be only sensibly done once the base rules are settled. I think @Sostratus likewise made a good point about how hard the game must be to balance giving all the moving parts now. So, when the next expansion comes out, or a few patches after that, I do hope things get better.

But I also still remain concerned that’s it just too high a mountain for FXS to climb, and the current design philosophy of the game (nothing random beyond the map and AI agendas, open ended strategies for victory are supported not punished) and so many toys will make it impossible to really nail the right challenge level without also ditching or cutting across things people like about the game. It’s a tough task FXS have. Given how good Vanilla was, I’m not saying it can’t be done. But yeah, tough.

For that reason, I really do hope FXS keep supporting the game past the next expansion. I just think it’s unlikely they can completely nail it with one more expansion, no matter how good it is. Games like EU4 have got as good* as they are because they’ve had such long development periods, with designers being willing to significantly rework existing mechanics to get things to a good place. I think FXS need to take the same approach - if FXS just get the next expansion out, and start working on BE 2 or God forbid Civ 7, then it’ll be a huge shame.

(Well, I think it’s good. Haven’t got round to buying it yet.)

 

[Leathaface]

I don't think Firaxis will improve the AI enough for the players who win on deity regularly and who want more of a challenge. That challenge would need better AI and not just giving more huge bonuses to the AI from turn 1.

Only modders can do that, and they need the DLL source to improve the AI to that extent.

EDIT: I just looked at Firaxis' steam post for the "Australian Summer 2017" update and under the MODDING TOOLS section they state "ModBuddy will also receive additional updates in the future, as part of a modding SDK update later. These tools do not include DLL source for Civilization VI at this time."

That sounds a bit promising, they didn't rule out the possibility of releasing the DLL source in the future. Though the Steam post was from the 21st of February 2017.

 

[Disgustipated]

A good place for this link. This video has been posted here before, but it is relevant to this thread.

AI's aren't built to win for a reason. See the above video. Also keep in mind the more fun mechanics they give the player to mess around with (production policy cards for example), the worse the AI will be since a human player can always handle game mechanics better than the AI. You could simplify the rules to the point the AI can usually beat a human player, but is that fun to a majority of players? Without simplifying mechanics I see no way to make the AI as competitive as civ5. Policy cards would probably have to go for that to happen.

Also keep in mind we players expect a degree of role playing from the AI. For example, many players would think it silly if Alexander the Great built a bunch of Holy sites and pursued a religious victory. Though I think I have seen him get a religion before. Or have Gandhi push for conquest victory. We want AI civs to act somewhat in character with some variation of course so the games don't get stale.

Edit: Now I am curious as what effect removing the cards may have on the ai. Because the AI makes horrible policy card choices. Probably can't mod them out because of the way governments are set up, but I may be able to set their bonuses to 0. I'll have to poke around in the game files. Or I could play a game and purposely choose policies that don't help me at all. These cards are a big reason the AI struggles. Although they will still have the problem of lack of focus and inability to concentrate efforts on a victory path.

 

[Archon_Wing]

For example, I legitimately lost (at least enough to give up) a leve 1 or 2 game due to barbs swarming worse than they ever had on me in a deity/immortal game. I've had a prince game where an early rush by the AI made the game way more difficult than my typical emperor game in which I have peaceful neighbors. In contrast, a peaceful game on deity allowed me to build up and expand rapidly to the point that I quickly matched, then surpassed the AI's starting bonuses.

That really reflects another problem in the game and it's basically the snowball time table. Basically, you want to hit various landmarks at a certain time and should anything happen that deviates from it, you're going to have a much harder time. This limits the options you have and punishes you heavily for not doing the "correct" thing.

For example, if you are not able to block the AI from getting two of the first 3 Great Writers, it's going to be much harder since there will be some point where the AI's culture will surge and the timing window for a culture victory vanishes. If stuff like Russia or Greece is in the game, it'll be harder by default. On the other hand if you do succeed in doing that, chances are you will probably win the game smoothly regardless of mostly everything else

If you spawn on a map where you can't boost Political Philosophy, then you're just going to have a slow game because those few turns really do matter. It always annoyed me to read about "strategy" when it seemed to be all about getting really lucky with City States. .

All and all, I think this makes the difficulty widely inconsistent more than it needs to be, and as a result it's very hard for me to tell whether it was a game where I actually played better or I was just being carried by my spawn. This isn't unique to this game, but with stuff like Harvest Goddess/Defender of the Faith/Crusade providing a huge early advantage to you (or a huge disadvantage against you), it makes it much harder for a player to get proper feedback.

 

[Aristos]

That really reflects another problem in the game and it's basically the snowball time table. Basically, you want to hit various landmarks at a certain time and should anything happen that deviates from it, you're going to have a much harder time. This limits the options you have and punishes you heavily for not doing the "correct" thing.

Huh??? If anything, civ 6 is the most forgiving game of the series bar none. You can easily recover from anything even on Diety with the only "cost" being the finishing time (which I couldn't care less about). I am pretty sure that is a design philosophy as it caters to the newer generations very well. (If you have kids and live in North America, and know what they are being thought in school, you will likely understand what I mean... among many things, "nothing has negative consequences").

 

[Archon_Wing]

Huh??? If anything, civ 6 is the most forgiving game of the series bar none. You can easily recover from anything even on Diety with the only "cost" being the finishing time (

That is not mutually exclusive with my statement. The ability to come back is based on relying on the same overtuned mechanics. Sure, you can play bad or pick a wacky strategy by Magnus chopping and use spies to sabotage spaceports, but in the end it's about using thse dominant mechanics and the AI's i8nability to win rather than anything particularly meaningful.


You can of course get away with mostly anything. But if you do stuff like found religions, play off bigger cities instead of more cities, play peacefully, it feels like you are handicapping yourself-- there is no reason to do these things other than challenge.

There is a significant difference in a game with many options where it allows multiple approaches to victory and a game where you have many options because your opponents aren't playing the game and none of it means anything.

 

[Aristos]

You can of course get away with mostly anything. But if you do stuff like found religions, play off bigger cities instead of more cities, play peacefully, it feels like you are handicapping yourself-- there is no reason to do these things other than challenge.

And you are in fact, and even then recover and win, thus proving that the opposite of what you initially said is the truth: this game does not punish you at all.