How Should Beyond Earth Handle AI Difficulty Scaling?

[Jman5]

One of the major complaints in the Civilization series has been how the AI gets to "cheat" as you go up the difficulty. The AI will get things like free techs, reduced production costs, increased gold, reduced tech time, lower unhappiness, and more. It's often used as an excuse for why people don't like playing higher difficulties.

Here is a complete list of the bonuses the AI get at different difficulties.
http://www.civfanatics.com/civ5/difficulties

So my question to all of you, is what sort of cheats or handicaps do you feel are acceptable for the AI at higher difficulties and what are deal breakers that just cheapen the game and make it less fun for the player? Or are no cheats acceptable?

I'll start off...

AI Bonuses that are acceptable:
1. Starting off with an extra basic fighting unit
2. Starting off with a little extra money
3. Bonus to base happiness (ie player starts with 5 happiness ai starts with 8)
4. Minor GPT bonus
5. Bonus fighting barbarians/aliens
6. Tech bonuses to keep them in the game (I'm not sure how this will work in BE)

AI Bonuses are unacceptable:
1. Reduced military upkeep costs
2. Insane happiness modifiers
3. No punishment for running a deficit at 0 gold
4. No punishment for having an unhappy civilization
5. Unfair production bonuses
6. Massive gold bonuses

I think what bothered me most about civilization 5 was how the AI is basically playing a different game at higher difficulties. Instead of being constrained by things like happiness, gold, production time, and military strength, it was only constrained by its own stupidity. This was particularly evident when fighting wars. You could be pillaging his territory left and right while cutting off most territory bonuses. Yet it could often continue building new units at faster than your pristine empire while rush-buying a new unit every single turn. Meanwhile its happiness is at double digits. It also allowed the AI to build a ridiculously huge military and an equally large empire.

Beyond Earth needs to do AI advantages better than has been done in past civilization series. Players need to be able to target and cripple the AI's resource generation better.

 

[AriochIV]

Ideally, the AI should get smarter at higher difficulty levels, rather than having to cheat. But I think we all know that is not going to happen. I can count on the fingers of one hand the number of computers games I have ever played in which the AI could beat me fairly (and most of them were chess games). I'll easily settle for an AI that's not incompetent.

The worst kinds of cheating are the ones where the AI is basically omniscient and can see everything you do regardless of whether they have visibility in your territory. And that's pretty common.

 

[Lorizael]

I think there might be a built in method of difficulty scaling for Beyond Earth. They mention that your faction arrives first and other factions arrive later with bonuses so as to make up for the lost time.

On higher difficulty settings, the AI could arrive earlier and earlier (but with the same bonuses). Eventually, you might be one of the colony ships arriving later, but without any bonuses to make up for it. I'd like that--being put in progressively harder situations--a lot more than active rule-breaking by the computer.

 

[KrikkitTwo]

Actually, I'd most prefer the AI to NOT start off with initial bonuses but instead continual bonuses.

That way no era of the game is more difficult than any other... (particularly annoying in Wonder+religion races)


So

Bad.
1. Significant Initial bonuses (units, extra cities, techs) [because this means that once you get catch up you dominate..but you are forced to abandon the early game]
2. Reduced "maintenance Bonuses" (ie less unhappy per person or gold per unit) [is either OP or exploitable sign of incompetent AI]

Good.
1. Reduced cost (AIs pay less research/production/gold/etc. to GET a tech, unit, social policy, building, etc.)
2. Minor initial bonuses (ie happiness, an initial bonus that is not immediately usable)

 

[Darsnan]

So my question to all of you, is what sort of cheats or handicaps do you feel are acceptable for the AI at higher difficulties and what are deal breakers that just cheapen the game and make it less fun for the player? Or are no cheats acceptable?

So I am going to segue off-topic a little bit here before arriving at the answer, and if its too far off-base please let me know.
I designed a lot of scenarios for SMACX, and I remember a lot of conversations over the years related to the subject matter in that it was observed by various people that scenarios in general weren't balanced, and I had to agree as most of the scenarios I designed (and which most of the other CMN designed) had moderate to outrageous bonuses for the AIs. These scenarios did present new and unique challenges for players, "curveballs" is what I liked to call them as they presented non-standard worlds for the players to adapt to and conquer, but generally speaking scenarios were the "next step" after players had exhausted everything the AIs (and really the game) could throw at them in the base game.

So how to scale the game? My first observation is that gaming AIs will always play the same way given the same set of circumstances: the AIs have no choice, as they are completely governed by the same set of algorhythms every time. Given the fact that humans will learn the nuances of any game and post the game's exploitable weaknesses online, then given the fact that game development budgets are finite, the better option is to give AIs scaling bonuses to present continued challenges to developing players.

The other observation I'll make is to paraphrase the Sid Meier quote "between fun and X, fun wins". For the higher level players/ players who are looking for new and unique challenges, then how do you make a competitive AI in order to make the game continually fun? What are your options?

And the final observation I will make: Beyond Earth really represents a great opportunity here in this respect in that, because the base game is giving you so many options regarding various worlds to select (i.e. a virtual "Diaspora of Man" from what has been revealed to date), then this opens up even greater vistas in regards to mod'ing for changing the base elements for these disparate worlds. I think (and this is only my observation from what has been revealed to date) the framework is being laid to offer an extremely large and diversified set of worlds to explore in Beyond Earth. Or, to put it slightly differently, if Beyond Earth is developed correctly in regards to how the AIs play in these various environments (to include scaling bonuses), then the base game will be increasingly and variably challenging based on the environments the players choose, and it will allow CMNs to "enhance" these environments even moreso.

So, to answer your question, I think scaling cheats for the AIs are the way to go, and the main reason I think this is the best way to go is that this then allows CMN a defined and static parameter (i.e. they know what the AI cheats are) to enhance the game even moreso in the future.

D

 

[Jman5]

I think it's pretty unavoidable to have some bonuses and advantages to the AI in order to give the player a challenge. What I was hoping to do is find out which advantages in particular you can live with and which ones make the experience not fun, or overly cheap.

Giving the AI omniscience or built in maphack is a good example of an advantage that many players find unfun and overly cheap.

 

[Liex]

The AI wouldn't need such generous happiness bonuses if they improved its luxury resources faster, or if they were more pragmatical when trading redundant resources (if they didn't like how you defended yourself against a common foe, they will simply not accept any reasonable trade proposal, even if they need it more than you...)

 

[Aeson]

Actually, I'd most prefer the AI to NOT start off with initial bonuses but instead continual bonuses.

Exactly right.

Starting bonuses make balancing the rest of the game impossible. The bigger the starting bonuses, the worse it gets. It becomes a "catch up, win" type game, because the hardest part is overcoming the starting bonuses. If you can do that, then the rest is easy in comparison.

 

[sendos]

If it is possible to code a smarter AI on higher difficulties, the following features might help:

- AI only settles at really ideal spots. If there are none, or if other person/AI has that area, they become more aggressive.
- AI develops a specific focus at start of game. For example:
--Production start: AI focuses on improving production to the point where they either build a large army or shift to support growth and science. AI could also aim for early projects/wonders.
--growth start: AI focuses on growth and science, and also on some expansion and improving productive tiles.
--military start: AI places emphasis on exploring and maybe even rushing, designed to harass rather than defeat other factions early on. Have a look at SMAC, where some factions like Hive and Believers start harassing you as soon as you meet them.
--science start: depending on type of faction, AI either focuses on better weapons for attacking or beeling for techs that further improve science. Kind of like coding AI rushing to education in civ 5 and then building those universities on higher difficulties.
- Then there are affinities:
-- Purity: AI places emphasis on terraforming and spacewarp victory.
-- Harmony: AI places emphasis on taming aliens and contact victory, or maybe attack purity players that are negatively affecting the planet.
-- Supremacy: AI focuses on science and spacewarp victory, maybe even a domination victory, too.

I also agree with OP's list of good and bad buffs for AI on higher difficulties. However, strength of aliens should be independently determined, not by difficulty, kind of like raging barbs option in civ 5 but without the decreasing combat bonus against them. On easier difficulties, there can be a combat bonus against aliens but it should end at prince-like level.

 

[stealth_nsk]

Civ5 tactic is much more complex than previous SoD, so the AI is actually near limits of modern hardware and significant improvements are not possible. So, yes, AI bonuses are needed and they should be big.

I also there should be less initial bonuses and more long-lasting ones. I would even suggest some gradually increasing bonuses. However this would suit Civ more CBE, which doesn't have epochs.

 

[AriochIV]

Civ5 tactic is much more complex than previous SoD, so the AI is actually near limits of modern hardware and significant improvements are not possible. So, yes, AI bonuses are needed and they should be big.

I'm a bit tired of hearing that excuse. I don't agree that good AI is simply a function of available CPU cycles... my experience is that such a claim is a crutch of inadequate AI programmers. I know that AI programming is hard, and it's true that pathfinding is a challenging computational problem, but when AI units jump in the water rather than engaging the enemy, this isn't a computational problem. It's a problem of inadequate algorithms. Anyone can beg for more CPU cycles to throw at a problem; a clever engineer finds a way to solve the problem with the resources he's been given.

Even if it were true that Firaxis had reached the pinnacle of AI design (which they haven't) and the limiting factor in a 1UPT system truly was processing power, then designing a game that your AI can't play competently within proscribed computational limits is poor game design. Either the AI programmers are incompetent, or the game designers are. Which do you choose?

 

[Kid R]

I think the AI bonuses should above all be simple and clear so you know what you're dealing with. About 5 key factors, game-long, should be fine and be easily comprehensible while providing plenty of subtlety. Remember keeping it simple for our puny brains also keeps it simple for AI algorithms to handle too

The Civ 5 AI was deliberately designed to be opaque in general, which is a separate issue really, but I don't think it's the best way to go with difficulty bonuses.

Also, I don't think somehow "smarter" AI on higher difficulty is ideal - it detracts from the AI personalities if they get smarter, which IMO reduces fun. (The only good case would be if the AI algorithms specifically benefited from longer CPU use like in chess, but Civ games aren't suitable for that kind of deterministic number-crunching AI.)

If the game can't be balanced and kept fun with a few scaling bonuses, and the designers feel the need to obfuscate their own work, then IMO the game hasn't been designed well in the first place

 

[Catalytic]

Obviously, the AI needs to be as good as it can be. In lieu of a perfect AI, bonuses are needed.

Bad bonuses:
-Bonuses that snowball (starting worker, second starting settler, etc)
-Bonuses that affect game pacing (science)
-Bonuses that result in ability to ignore fundamental limits (happiness/anti-growth)

Good bonuses:
-Scaling - scale up with difficulty, scale linearly, not exponentially
-Responsive - shifts aid to struggling civs to keep them from falling behind, less aid to leaders

If we want the game to remain interesting, we would prefer a design where the game plays a bit like a race. You want the AIs to be able to keep up with the front-runner and not fall too far behind.

Catch-up mechanics like giving reduced research costs to AIs based on the number of other competitors who've researched it would help. For example, the first 20% of factions to research a tech pay full price. The next 20% get a 10% discount and on down til the last factions get a 50% discount to keep them in the mix.

Given the AI's trouble with managing its military, a slight, scalable production bonus could be offered for production of military units so the AI could field reinforcements quickly, but would still be constrained by unit caps and maintenance costs in a fair way.

Slight gold bonuses are a good idea as it offers flexibility to the AI to simply buy what it needs, but it needs to be carefully done or we get unintended consequences of AIs being able to subvert entire victory conditions like the Diplomatic victory in Civ 5.

I prefer an AI that's hypercompetitive within the same rules as the player rather than one that re-writes or breaks the rules.

 

[stealth_nsk]

I'm a bit tired of hearing that excuse. I don't agree that good AI is simply a function of available CPU cycles... my experience is that such a claim is a crutch of inadequate AI programmers. I know that AI programming is hard, and it's true that pathfinding is a challenging computational problem, but when AI units jump in the water rather than engaging the enemy, this isn't a computational problem. It's a problem of inadequate algorithms. Anyone can beg for more CPU cycles to throw at a problem; a clever engineer finds a way to solve the problem with the resources he's been given.

Ok, let me dive a little bit deeper. The only real solution of such problems by computer is bruteforce - checking all variants. With 1UPT and civilization tactics, modern computers are barely able to plan 1 turn this way. There are possible improvements, not calculating variants which look bad without calculation (i.e. minmax algorithm), but it wouldn't help here.

Instead, Civ (as any other game with the same problem) has AI based on different principle. Speaking roughly, there are rules, which are implemented based on conditions. But the more rules are there and more complex conditions are, the more computing power is required.

Civ AI is actually brilliant, if you look at it. It has planning. It uses synchronous attacks, etc. If you try to describe its algorithms, you'll really see a lot.

Even if it were true that Firaxis had reached the pinnacle of AI design (which they haven't) and the limiting factor in a 1UPT system truly was processing power, then designing a game that your AI can't play competently within proscribed computational limits is poor game design. Either the AI programmers are incompetent, or the game designers are. Which do you choose?

AI is not match to human. For turn-based games where reaction is not a limiting factor it can't compete. So there are 2 possibilities:
- The game's tasks are so stupid that even AI could solve them, but it's not interesting for players.
- The game provides interesting challenge, but AI have to use additional bonuses to provide challenge to a player.
I really prefer second.

Also, keep in mind, the vanilla Civ5 was designed with strong multiplayer support in mind. Shafer's design was aimed at main civilization AI playing in the same terms as human players. Not all decisions were good and many things weren't implemented at all, but with overall target for multiplayer complex tactics had really good reason. After Shafer leaving the team, development moved to different direction and multiplayer never got it's attention.

 

[Fabio1701]

I think the it is safe to assume the AI will need to cheat I'm some way. If they could programme an Ai so good that it wouldn't need to cheat we would all either need a super computer or firaxis will have invented something that The Pentagon would want to buy.

I agree that linear cheats are better, starting bonuses ar ok, it depends whichones. There won't be a race to getting a religion in this game (I think) and we don't know about wonder races. I like the idea of people coming from Earth with different resources which can be used to increase difficulty but it needs to be things which don't exponentially add benefit.

I really hate the AI settling cities too close to eqchither and screwing themselves or me if I wanted take those cities.it may be a protectiv tactic but I always settle cities with least overlap possible and assuming max expansion.

 

[Lorizael]

Actually, I'd most prefer the AI to NOT start off with initial bonuses but instead continual bonuses.

That way no era of the game is more difficult than any other... (particularly annoying in Wonder+religion races)


So

Bad.
1. Significant Initial bonuses (units, extra cities, techs) [because this means that once you get catch up you dominate..but you are forced to abandon the early game]
2. Reduced "maintenance Bonuses" (ie less unhappy per person or gold per unit) [is either OP or exploitable sign of incompetent AI]

Good.
1. Reduced cost (AIs pay less research/production/gold/etc. to GET a tech, unit, social policy, building, etc.)
2. Minor initial bonuses (ie happiness, an initial bonus that is not immediately usable)

Okay, those are good points. I'm not a fan of the AI playing by a separate set of rules, but I understand why it can be necessary.

Another thought: What about a map with dynamically placed resources? That is to say, on higher difficulty levels, you find poor land and the AI finds good land. This fits my criteria (which it's possible no one else may share) of you being placed in progressively harder situations without active cheating, but it also creates continual bonuses for the AI. Their cities will naturally have higher food, hammer, and beaker output because of the better resources present.

 

[stealth_nsk]

Okay, those are good points. I'm not a fan of the AI playing by a separate set of rules, but I understand why it can be necessary.

Another thought: What about a map with dynamically placed resources? That is to say, on higher difficulty levels, you find poor land and the AI finds good land. This fits my criteria (which it's possible no one else may share) of you being placed in progressively harder situations without active cheating, but it also creates continual bonuses for the AI. Their cities will naturally have higher food, hammer, and beaker output because of the better resources present.

I think for many people it will be even more cheating than the current situation. Also, it's difficult to define, which land belongs to whom (except for capital).

 

[GoodSarmatian]

Actually, I'd most prefer the AI to NOT start off with initial bonuses but instead continual bonuses.

I'd go one step further and say the AI should get bonuses that increase with time at higher difficulties, like for example a cumulative % bonus to science and production every time they enter a new age. This would help them be competetive without turning the game into an uphill battle from the start that just eliminates good peaceful strategies because you won't be able o get certain wonders no matter what you do.

 

[stealth_nsk]

I'd go one step further and say the AI should get bonuses that increase with time at higher difficulties, like for example a cumulative % bonus to science and production every time they enter a new age. This would help them be competetive without turning the game into an uphill battle from the start that just eliminates good peaceful strategies because you won't be able o get certain wonders no matter what you do.

Yes, that's exactly what I said. Since CBE have no ages I wouldn't expect this thing there, but for Civ6 it would be interesting.

 

[True_Candyman]

Yes, that's exactly what I said. Since CBE have no ages I wouldn't expect this thing there, but for Civ6 it would be interesting.

There's nothing preventing CBE from having an ages mechanic. They are perfectly within limits to invent some. They could even just grade the tech web, so each ring outwards is the equivalent of an age.