Sid Meyer's Civilization - Beyond Earth

[Simon_Jester]

@Simon_Jester: I do agree with the criticism of Civ V's war system vs. a real wargame, but still feel that it's leaps and bounds better than Civ IV's maddening stacks of doom. This is probably one of those personal tastes and preferences issues that cannot be resolved logically. However, while I prefer Civ V's basic approach, I do think both philosophies could be improved. Civ IV's could benefit by being more like Civ II's, with zone of control to prevent an entire doomstack army from snaking around any forts or "defenses" (quoted because there is effectively no such thing in Civ IV), and collateral damage for opposing stacks that are out in the open (although probably not so drastic as Civ II's "lose one unit, lose them all"). Ultimately, I wouldn't have such a problem with the infinite stacks approach if you could just force an attacker to actually attack instead of snaking all over your territory razing improvements and forcing you to go against their giant defensive bonuses (assuming they're smart enough to park their stack in rough terrain) to get them to stop.

Yes, I missed that feature myself. A few modifications that would allow you to force a megastack to give battle, to attrite their strength more gradually, and encourage them to disperse a bit and fight on a broader front, and you'd recover something like a realistic wargame feel, without necessarily making the game too hard for the AI to handle.

(Megastacking SHOULD become instant suicide in the Atomic Age; the equivalent of an invading stack of doom would probably be the A-number-one target for a first use of nuclear weapons in real life... but Civ IV didn't let you nuke your own territory even if there were huge armies sitting on it)

Civ V's one-unit-per-tile approach could benefit from a greater number of smaller tiles to prevent the issues that you mentioned. This would lead to more variety in weapon ranges (especially with direct vs indirect fire), and would also have to mean that larger cities would take up more and more tiles, which could be interesting in and of itself.

That strikes me as the best possible way to improve the system, and relatively easy. The biggest catch is that it would greatly increase the computational load on the graphics chip, and frankly if you're buying Civ V for its cutting edge graphics you have bigger problems.

The real trick is making sure that the optimal distance between cities increases to scale with the increased number of smaller tiles- in other words, cities are still 300 miles apart, it's just that each tile between them is 50 miles wide instead of 100. You'd need an incentive to discourage crowding cities, which would also have a number of other desirable effects on the Infinite Carpet of Buildings gameplay approach.

[Darmalus]

Civ V's one-unit-per-tile approach could benefit from a greater number of smaller tiles to prevent the issues that you mentioned. This would lead to more variety in weapon ranges (especially with direct vs indirect fire), and would also have to mean that larger cities would take up more and more tiles, which could be interesting in and of itself.

That strikes me as the best possible way to improve the system, and relatively easy. The biggest catch is that it would greatly increase the computational load on the graphics chip, and frankly if you're buying Civ V for its cutting edge graphics you have bigger problems.

The real trick is making sure that the optimal distance between cities increases to scale with the increased number of smaller tiles- in other words, cities are still 300 miles apart, it's just that each tile between them is 50 miles wide instead of 100. You'd need an incentive to discourage crowding cities, which would also have a number of other desirable effects on the Infinite Carpet of Buildings gameplay approach.

Sub tiles? Have the normal tiles count for economic improvements, cities and the like. Break each tile into sub-tiles that only matter for combat, maybe 6 triangular tiles. You can still have the upt limits, but it would be more of a uptt, unit per tactical tile. You only control of the hex tile when all sub tiles are in your control. This would allow situations where you lose control, but improvements are not yet destroyed because you have partial control.

[AniThyng]

I liked the style of the intro, and I got the vague feeling they were going for a sort of "this is the natural consequence of a civ5 game thus the wonders are randomly placed". But I may be wrong.

It felt to me more like "these are the social and ecological consequences that are causing us to abandon Earth". Especially since the narrator sounds a hell of a lot like Diedre Skye from the original ACen.

Well yeah, that too, I meant there was also a air of "this isn't our Earth as such, does that look like Cairo to you?". But on the other hand, Arabian gulf. And I suppose New York will degenerate into a urban slum

If I had to rank civ games though in terms of personal preference:

Civ IV + BTS
Civ II Gold (including fantastic worlds expansion, + all the wonderful mods created in the golden age of civ ii modding)
Civ V + G&K (it's not actually unfun to play, for all its worth. And it genuinely looks pretty. But it lacks in the end the oomph due to the poorly thought out UPT combat even though it was a nice idea in theory (increasing hit points greatly improved it, i must say)
Civ (the one and only.
Civ Rev (what? it was fun on Xbox and a nice diversion on NDS )
Civ III + ...whatever that final expansion was. Conquests? . Civ III had its moments, but the AI was a fail, and it didn't really have the same oomph as IV did in the end. The atmosphere was actually nice, as was the city views.

...Ultimately though, I can't really say any of them were <bad> games. Flawed in some ways, awesome in others, they ALL fulfiled many a sleepless weekend night when they came out.

[Purple]

Civ V's one-unit-per-tile approach could benefit from a greater number of smaller tiles to prevent the issues that you mentioned. This would lead to more variety in weapon ranges (especially with direct vs indirect fire), and would also have to mean that larger cities would take up more and more tiles, which could be interesting in and of itself.

That strikes me as the best possible way to improve the system, and relatively easy. The biggest catch is that it would greatly increase the computational load on the graphics chip, and frankly if you're buying Civ V for its cutting edge graphics you have bigger problems.

The real trick is making sure that the optimal distance between cities increases to scale with the increased number of smaller tiles- in other words, cities are still 300 miles apart, it's just that each tile between them is 50 miles wide instead of 100. You'd need an incentive to discourage crowding cities, which would also have a number of other desirable effects on the Infinite Carpet of Buildings gameplay approach.

Sub tiles? Have the normal tiles count for economic improvements, cities and the like. Break each tile into sub-tiles that only matter for combat, maybe 6 triangular tiles. You can still have the upt limits, but it would be more of a uptt, unit per tactical tile. You only control of the hex tile when all sub tiles are in your control. This would allow situations where you lose control, but improvements are not yet destroyed because you have partial control.

Suggestions like that (and I am talking to both of you) sound good on paper but have serious issues in practice because of the way they effect the AI. And by now you know it's going to be bad.


Basically, any tactical game can be boiled down to a finite state problem. That is to say a problem in which a finite number of pieces can be arranged on a finitely sized field in a finite number of combinations leading to a finite number of outcomes for each consecutive move. Tactical AI operates by iterating through each of these states thinking a certain number of turns in advance and rating each of them based on a understanding of the pieces it has and the pieces its opponent has and how they are arranged until it comes up with a series of moves that are supposed to lead to victory. Now if this sounds complicated it's because it is.
Take for example chess. It's a humble game with only 16 pieces per player and on a relatively restrictive 8x8 board. As such, this humble game has no less than 1047 legal positions that your computer has to either iterate through or somehow figure its way out of. The typical game of civilization on the other hand has an undeterminable number of pieces operating on a semi determinable sized field. Given this, it is a miracle they managed to get the AI to work as good as it did so far.

And there is the catch of the day. Civ is not a tactical game. It's an empire building game. The an AI for any Civ game needs to be capable of both tactical and strategic gameplay. That is to say it needs to not only understand how to make the best out of the pieces it has but also plan ahead and develop long term strategies so that it can have the right pieces at the right time. The end result of this is that any linear growth in the complexity of the tactical model is going to lead to an exponential growth in the complexity of the overall model. And I am not so sure any of us can afford to get our own Deep Blue to play Civ on.

[Simon_Jester]

Purple is correct and this is the big challenge in writing AI. Arguably the older games with their Stacks of Doom simplify the problem, because all the AI has to do is:

"Step 1: Implement build-an-army script." This step just consists of identifying, say, which of the AI's cities is a good jumping-off point for an invasion force, then ordering it to move military units to that city until it's got... 'enough.'

"Step 2: March army toward enemy city, attack!" Again, fairly straightforward. Granted the AI won't be making very good use of the terrain, but at least a mindless assault like this is a credible threat to the player if not countered effectively.

Incidentally, amphibious assaults would be a pretty straightforward variation on this theme.

When the AI is required to spread out its operations spatially, then it runs into the chessboard problem because it has to consider all the possible ways the units could be arranged, instead of just heaping up a big blob of them on one tile and moving them all simultaneously to the nearest target.

Humans don't have to solve a finite-state problem directly to play games with finite-state boards (which makes us good at playing them but terrible at optimizing our solutions)... which is probably why we invented such games in the first place. The elaborate terrain and hex-based maps of an Avalon Hill wargame (or Civ V) are a nightmare for a computer player to analyze and optimize unless they can "think strategically" by discarding the vast majority of possible moves.

Which is probably why the AI for real hex-based computer wargames is practically nonexistent and they just passively sit on the defensive except under very unusual circumstances.

Sub tiles? Have the normal tiles count for economic improvements, cities and the like. Break each tile into sub-tiles that only matter for combat, maybe 6 triangular tiles. You can still have the upt limits, but it would be more of a uptt, unit per tactical tile. You only control of the hex tile when all sub tiles are in your control. This would allow situations where you lose control, but improvements are not yet destroyed because you have partial control.

Personally I think the improvements should be lost faster than that- it would be rather interesting to have something like a World War I situation where the "front" is a belt of thoroughly destroyed, contested tiles.

But yes, this would be an interesting idea. There are some catches that might result from implementation, and it might also make the game "too complicated" in the eyes of the casual player, but it's a good idea in my book.

EXCEPT, as Purple reminded me, the AI would be totally unable to handle this.

[Darmalus]

Huh, I never thought the AI had to consider the whole board at once. I certainly never do that unless the board is really tiny.

Usually my only "grand strategy" was "I'm here, enemy there. Move forces that way." and when the fighting did happen, I just went down the list of units and considered what each unit could do right then and picked an option, which was usually focus on the weak or wounded one, or the dangerous one. I always assumed the AI did something similar.

I've never been amazing at these games, but I usually muddled through. Except in Civ V, where for some reason the AI seemed to just give up after a few confrontations. Kinda sucked the fun out when the AI just gave up and went AFK.

[Purple]

Huh, I never thought the AI had to consider the whole board at once. I certainly never do that unless the board is really tiny.

Usually my only "grand strategy" was "I'm here, enemy there. Move forces that way." and when the fighting did happen, I just went down the list of units and considered what each unit could do right then and picked an option, which was usually focus on the weak or wounded one, or the dangerous one. I always assumed the AI did something similar.

I've never been amazing at these games, but I usually muddled through. Except in Civ V, where for some reason the AI seemed to just give up after a few confrontations. Kinda sucked the fun out when the AI just gave up and went AFK.

That's because you are capable of making intuitive assumptions about the game based on past experience. We humans are remarkably good at that. Plus, you are fully capable of evaluating dozens of tiles at once just by looking at your screen without even thinking about it where as the AI literally has to take one tile at a time.

[Tribble]

The easiest solution would be to simply set the UPT to ~3-5. That way you avoid both the Carpet/Stacks of Doom, and have actual battle lines. I've used mods like that before when they were available, and the game play was a lot better. Plus, the Civ 5 A.I.'s ability to wage war improves quite a bit. I fail to see why they didn't just give us that option from the start.

[Purple]

The easiest solution would be to simply set the UPT to ~3-5. That way you avoid both the Carpet/Stacks of Doom, and have actual battle lines. I've used mods like that before when they were available, and the game play was a lot better. Plus, the Civ 5 A.I.'s ability to wage war improves quite a bit. I fail to see why they didn't just give us that option from the start.

Honestly I like what the HOTK team did (linked to them earlier). They added hero units that can form armies of 3-5 members of the same type (archer, mele, cavalry etc.) and due to the mechanics involved these armies have all the power of a doom stack without the size. And the AI knows how to use them well enough to the point that it only keeps as many units as it can put into these stacks and no more. In fact, it knows how to use them so well that I am yet to actually see a doom stack in that mod at all.

So really, just ramp up unit maintenance costs and give the AI some incentive to form small cohesive lumps (and teach it to do so) and you have a system that is both fun and engaging.

[phongn]

The easiest solution would be to simply set the UPT to ~3-5. That way you avoid both the Carpet/Stacks of Doom, and have actual battle lines. I've used mods like that before when they were available, and the game play was a lot better. Plus, the Civ 5 A.I.'s ability to wage war improves quite a bit. I fail to see why they didn't just give us that option from the start.

It's easy for a mod to play with that; actual full release-quality playtesting and AI optimization for 1 or N UPT makes life far more difficult for the designers.

The problem was one of implementation. Hex-based wargames are a surprisingly challenging thing to program good AI for- it's easy to make an AI that can defend on a hex-based map, that can park around a static point and secure that point. But it's relatively hard to design one that will attack efficiently and make good use of terrain on the offensive, or in a mobile defense.

Also, as noted earlier, if you're going to add a whole new tactical combat element to the game, you will probably have to rethink the role of unit movement and space on your map. Civ V tried to avoid really doing that,* because messing with the basic spatial relations of the map would mess with the grand strategy element. So there's a conflict that undermines the quality of the game, in the eyes of people who thought "hex-based tactical combat" and expected good hex-based tactical combat.

There's a postmortem interview where the Civ V AI designers discussed how they really didn't quite get it right - some of the things you brought up are things they recognized.

[Arthur_Tuxedo]

Honestly I like what the HOTK team did (linked to them earlier). They added hero units that can form armies of 3-5 members of the same type (archer, mele, cavalry etc.) and due to the mechanics involved these armies have all the power of a doom stack without the size. And the AI knows how to use them well enough to the point that it only keeps as many units as it can put into these stacks and no more. In fact, it knows how to use them so well that I am yet to actually see a doom stack in that mod at all.

So really, just ramp up unit maintenance costs and give the AI some incentive to form small cohesive lumps (and teach it to do so) and you have a system that is both fun and engaging.

Sounds a lot like armies in the Kohan series, except in those games they were made of 4-6 dissimilar units in formations that are pre-set by the player. This setup could make the sub-tile concept workable without increasing micro or further burdening the AI, since the whole stack would move at once on macro tiles and could incorporate melee, ranged, and artillery elements into one "unit" that the AI can simply hurl at the player.

[Purple]

This setup could make the sub-tile concept workable without increasing micro or further burdening the AI, since the whole stack would move at once on macro tiles and could incorporate melee, ranged, and artillery elements into one "unit" that the AI can simply hurl at the player.

Not really. More tiles = more complexity any way you put it. And adding combined armies like that would actually make things more complicated as the AI would have to "think" about how it combines its units as opposed to just going "these 4 swordsmen are now 1".

[Arthur_Tuxedo]

Not if there are pre-sets for optimum combinations, and there wouldn't be more tiles from the AI's point of view since you can only move the army as a unit into macro tiles. It's a scheme to keep it simple for the AI and still let the humans have a field day playing around with sub-tile ranges and stack composition.

[Purple]

Not if there are pre-sets for optimum combinations, and there wouldn't be more tiles from the AI's point of view since you can only move the army as a unit into macro tiles. It's a scheme to keep it simple for the AI and still let the humans have a field day playing around with sub-tile ranges and stack composition.

Honestly, unless you change the combat system something fierce from what I am familiar with in Civ IV (since the I feel the Civ V system is outright unsuitable as a starting point) the optimal preset is a stack of units of the same type. Preferably a type optimized against what ever you know the opponent has (if he has a lot of cavalry you get pike men) and a stack of catapults.

[Block]

I'm not getting what the problem with one unit per tile is, if anything it promotes battle lines and combined arms. I tend to have as many artillery and air units as I do front line units.

[AniThyng]

Yeah but the ai is hopeless at it

[Vehrec]

I'm sure if you threw Deep Blue money and programming time at the problem, and simplified the hell out of the game, the computer would be a much better strategist. Making computers good at 'strategy' is not easy, since they're not great at thinking ahead. And if you let each computer player think hard about its turn, how long is each turn gonna take?

On the other hand, I am somewhat annoyed by people who complain that they're backstabbed by the AI. Do you think they're just gonna let you run away and win the game? Because that's what staying friendly amounts to. Go to the city-states if you want a nicely passive and chill AI who will do your bidding.

[Purple]

I'm sure if you threw Deep Blue money and programming time at the problem, and simplified the hell out of the game, the computer would be a much better strategist. Making computers good at 'strategy' is not easy, since they're not great at thinking ahead. And if you let each computer player think hard about its turn, how long is each turn gonna take?

Well considering that Civ 4 is more complicated than chess and has no convenient library of preexisting games to use in order to shorten decision times... Let's just say that if you want to progress past turn one you should consider investing in reproduction.

On the other hand, I am somewhat annoyed by people who complain that they're backstabbed by the AI. Do you think they're just gonna let you run away and win the game? Because that's what staying friendly amounts to. Go to the city-states if you want a nicely passive and chill AI who will do your bidding.

There is good backstabbing and bad backstabbing. Good backstabbing is when the AI leader is one known historically for being sneaky either due to a mixed relationship or the personality he is based on and has something to gain from it. In other words, you aren't shocked by it in hindsight. Bad backstabbing is when an AI that has been your friend all game long suddenly decides they want to attack you for the hell of it.

There is a reason why the original Civ games had a victory type where the AI could vote you into being the world leader. If you can maintain a really good relationship with them than you deserve it.

[Tribble]

Also, good backstabbing involves having some kind of objective and at least some modicum of success. In all of the Civ games the AIs seems to love doing it for the Lolz, even when the odds are ridiculously against them. Like when Gandhi suddenly declares war on you and "invades" with 1 war elephant, even though you outnumber him 20-1 and your military consists of infantry and tanks. When a backstab does occur you should be impressed, rather than saying to yourself "WTF were you thinking???"

[Simon_Jester]

I'm sure if you threw Deep Blue money and programming time at the problem, and simplified the hell out of the game, the computer would be a much better strategist. Making computers good at 'strategy' is not easy, since they're not great at thinking ahead. And if you let each computer player think hard about its turn, how long is each turn gonna take?

So far, most success stories involve programming the AI to do specific things in order to, say, produce effectively- the AI automatically builds new cities at a certain rate, and has a subroutine to place them intelligently based on the available land, and so on.

That way, the massively complicated question of "so, what do I do now?" reduces to more specific and optimizable questions like "so, where is the best place to put this city" and "so, what do I do with this city next?"

On the other hand, I am somewhat annoyed by people who complain that they're backstabbed by the AI. Do you think they're just gonna let you run away and win the game? Because that's what staying friendly amounts to. Go to the city-states if you want a nicely passive and chill AI who will do your bidding.

The complaint is that the AI betrays people for unpredictable reasons, or at times that it is irrational to do so. Or that they hate you based on things that leave you no way to win their support- they hate you for settling and not settling, or if you approach them or if they approach you, or something like that.

As a result, the AI stops looking like a nation sincerely trying to win the game, and stops looking like a real-life nation with rational diplomacy. It starts looking like a trigger-happy schizophrenic, someone who may profess to be your friend this morning, then start attacking you this afternoon because the incomprehensible voice in their head told them to.

[Tribble]

The complaint is that the AI betrays people for unpredictable reasons, or at times that it is irrational to do so. Or that they hate you based on things that leave you no way to win their support- they hate you for settling and not settling, or if you approach them or if they approach you, or something like that.

As a result, the AI stops looking like a nation sincerely trying to win the game, and stops looking like a real-life nation with rational diplomacy. It starts looking like a trigger-happy schizophrenic, someone who may profess to be your friend this morning, then start attacking you this afternoon because the incomprehensible voice in their head told them to.

And it's alright to have the occasional trigger-happy schizophrenic civ (e.g. I'm not expecting Attila the Hun to discuss matters over tea and biscuits, and Montezuma is notorious for being a complete psycho in every game etc), but all of Civ 5 AIs act that way all the time.

[Vendetta]

Remaining friendly with a 4x AI usually just means "accede to whatever bullshit demands it tries to make", because it will ignore your industrial capacity in its assessment of your strength and thus believe you "weak" despite the fact that two turns from now you could have five tanks for every spearman it owns or equivalent.

So it will make some irrational demand and declare war if you refuse.

[Purple]

Remaining friendly with a 4x AI usually just means "accede to whatever bullshit demands it tries to make", because it will ignore your industrial capacity in its assessment of your strength and thus believe you "weak" despite the fact that two turns from now you could have five tanks for every spearman it owns or equivalent.

So it will make some irrational demand and declare war if you refuse.

Ironically that sort of reminds you of real life. Perl Harbor and all that.

[Simon_Jester]

Yes, but Japan was unusual in that it ignored the relative industrial capability of the US in launching its attack on the US and the European colonial powers. The only other nation I can think of that made a comparably big mistake of the same type in the modern era was Nazi Germany, which also had huge problems accurately judging its enemies' strength starting in 1941.

And frankly, both Japan and Germany were at least within shouting distance of the military potential of the enemies they attacked. It wasn't a case of one side having tanks and machine guns while the other side had muskets and pikes, or anything stupid like that.

Whereas the AI in Civ games routinely launches attacks under those conditions, despite this being a totally predictable recipe for disaster that it would be trivially easy for the AI to grasp- say, it could have a "tech coefficient" by which it multiplies your estimated military strength, or an "industrial term" added in to take into account the ability to produce masses of units very quickly. It's not like the game is too stupid to count "number of units of production your empire turned out today."

[Tribble]

Oh, I was being literal. As in, I've literally had countries declare war on me and attack with a single obsolete unit (like a horseman) against my "carpet of doom" consisting of infantry and tanks. It's such a pathetic display that I pretend the A.I. leader was punishing one of it's generals by deliberately sending him on a suicide mission.