Why Sid Meier gave up on stacking units?

[warpus]

At least historically, this has been too costly/impractical to be viable

It's all relative, isn't it? My initial post complains about the fact that they barely spend any time improving the computer opponent behaviour/mechanics for Civ games, and prefer to spend the time and effort on other aspects of the game, such as graphics or what have you.

Other games are able to give us somewhat realistic computer opponents because enough emphasis is placed on that sort of thing right out of the gate. Meanwhile the status quo w/ Civ is that they've already figured out their approach of how computer opponents behave years ago.. and aren't interested in making any improvements or alterations to the status quo at all.

It's fine to speculate that it's just too complicated to improve the computer opponents in Civ. This could very well be the case, but I don't believe it. I don't think they've even tried. They're happy with the status quo because it's simple and it allows them to focus on other aspects of the game and that's that. And since only experienced players notice and the vast majority of people who buy the game don't - that works for them.

 

[TheMeInTeam]

It's fine to speculate that it's just too complicated to improve the computer opponents in Civ. This could very well be the case, but I don't believe it. I don't think they've even tried. They're happy with the status quo because it's simple and it allows them to focus on other aspects of the game and that's that. And since only experienced players notice and the vast majority of people who buy the game don't - that works for them.

We're kind of saying the same thing though. It's a matter of cost proposition. If we agree it's easier to make a (financially) successful game by skimping on AI and emphasizing other game assets, and almost certainly cheaper, then that's what we're going to see the most often. Why invest in something that most players won't engage with much, when the average player is still sub-emperor (IIRC)?

I agree the AI sucks, but it seems enough of the player base is either worse than it or not better than it by enough to matter.

Other games are able to give us somewhat realistic computer opponents because enough emphasis is placed on that sort of thing right out of the gate.

Which other games? I've only observed this when the game was significantly simpler and/or more dependent on speed of processing/doing inputs than strategy. It's why the machine learning used to make strong Go/Starcraft AI was so impressive...AI playing that sort of game at that sort of level was otherwise (to my knowledge) unprecedented.

 

[uppi]

No one actually wants a good AI.

A good AI would ally all other AIs and then dogpile the player. No one wants to play against that.
A good AI would make diplomacy pointless. It would backstab you as soon as you are weak. I think it was the Civ 5 AI that would relentlessly backstab you (actually acting more like a human) and players were complaining all the time.
A good AI would play the meta. No immersion of any kind.

 

[TheMeInTeam]

A good AI would ally all other AIs and then dogpile the player. No one wants to play against that.

False. Sufficiently strong AI would likely evaluate the player as among the weakest nations/positions on the board. And it would be right.

Far less than dogpiling, it would likely leverage an advantage from better play to kill you off while trying to minimize the gains other AIs get off you. Like how good players target bad players in FFA. They don't "dogpile" them. Why share the spoils?

A good AI would make diplomacy pointless. It would backstab you as soon as you are weak.

False. This is not what observed best practices are in games where each faction has a tiny % of the total resources available in the game. It would depend on situation, to evaluate when it's time to backstab. Though in a game with one and only one winner, the notion of "backstabbing" is an awkward, fake concept unless we're talking about breaking a specific agreement, which Civ often doesn't even allow.

I think it was the Civ 5 AI that would relentlessly backstab you (actually acting more like a human)

False. This is not what good play looks like. I submit actual performance of Civ 5 AI as evidence. You don't want high investment, slow wars by attacking someone at first opportunity, if you're trying to win. You want to pick a target where you can win as decisively as possible.

A good AI would play the meta. No immersion of any kind.

A good AI would *define* the meta, based on the current game rules. If the result lacks immersion, it's an indictment of the game's design/balance. And indeed, I will call out Civ games in this regard. They add lots of "options", aka consistent false choices that outright depend on opponents not trying to win.

You otherwise seem to hold some odd conceptions about what a "good" AI does or would do. But let me give an example to demonstrate why the quoted statements are false:

In Dominions 5, diplomacy is limited. There are no formal truces/forced peace treaties, nothing to compel you to act in any way. It's a true free for all, with anybody able to attack anybody else at any time for any reason they want. And like in Civ, only one person or team can win.

The best of the best, the players that win the most consistently by far...do you really think they do not engage in diplomacy with other players in this environment, even when there can be only one winner? Do you believe these players immediately "backstab" other people, or break agreements they make? Do you think they all gang up on one player, in a game where everyone is at their own level of play (or worse)? Do you believe these great players behave diplomatically like the Civ 5 AI (lol)?

If you do not believe these things, what makes you believe a "good" AI (in the sense that it's trying to optimize its winrate and is well-trained to do so) would do these things? How often did the Civ 5 AI which rolled to engage in such behavior win the game, on average? At best, you could make the case that Civ 5 AI mimics an aggressive human player who is bad at the game in both micro and macro decisions. But using that as a model for predicting how a "good" AI would behave is naive.

 

[Gorbles]

My point is that the current way difficulty levels are built into the game is lazy and gets boring if you are an experienced enough player.

I get you. But it's one thing to point out a problem. It's another to actually address it. And I'm not saying you should, or anyone here should!

But it's a known problem. Remember, my original points only related to correcting you on the effort the developers have likely put in. You keep claiming they "haven't bothered". They've bothered a lot. They haven't succeeded, but that's a different thing entirely. Very few, if any, video games in existence have solved this problem. Nearly every single game AI or logic routine "cheats" in some way.

Maybe it'd help if you gave examples of games that you think did this well. Because I'm sure I can find players for those games that make it look like a joke. But it'd still be a good benchmark for the kind of thing Civ should be aiming for.

Technically the game makes no use of AI concepts whatsoever, so there is no AI there to improve in the first place... I threw out "AI" as an idea as to how better difficulty levels could be accomplished. There are likely other ways to get to a better game, not necessarily using any sort of AI concepts at all.

Technically, the game absolutely makes use of concepts that are popular in AI. Pathfinding, for example, is commonly based on popular algorithms like A* search (and / or other variants, but A* is ridiculously effective, so). Just like in other games. But you can't crack every nut with A* search. Or even with a combination of search algorithms.

There are definitely other ways to tackle difficulty levels and the how the game responds to these. Like I said, I'm fully on board with the devs thinking on their current 1UPT systems, on their game AI and difficulty tuning, and so on.

But rationally, a lot of this thinking will still involve the game "cheating" in some way. If you can't make a game think like a human (which is obvious, the Turing test exists for a reason), you need a way to balance the scales. If you can't make a game make intelligent decisions, you need to try for the next best thing, accounting for the development time you're allowed. Making a game is inherently about compromise, and yes, some of that compromise will harm the vision of the game as it was intended to be.

But in this specific case, I feel pretty comfortable in choosing my hill to die on as "the developers did, in fact, try".

 

[warpus]

We're kind of saying the same thing though. It's a matter of cost proposition. If we agree it's easier to make a (financially) successful game by skimping on AI and emphasizing other game assets, and almost certainly cheaper, then that's what we're going to see the most often. Why invest in something that most players won't engage with much, when the average player is still sub-emperor (IIRC)?

And that's the essence of the problem. They can do this because it's not a problem that's noticable unless you've put in a certain number of hours, at which point it begins to become annoying. So yeah, they probably figure: "We can throw $1 million more at this for the next iteration of the game.. or.. you know what, screw it, who cares, everybody's going to buy this anyway"

So we essentially agree (I think)

Which other games? I've only observed this when the game was significantly simpler and/or more dependent on speed of processing/doing inputs than strategy. It's why the machine learning used to make strong Go/Starcraft AI was so impressive...AI playing that sort of game at that sort of level was otherwise (to my knowledge) unprecedented.

We should really stop using the term "AI" in this conversation, unless it relates to artificial intelligence specifically. This is not a suggestion but more like an observation, since I don't think it's ever going to happen (people are used to the term, so they're going to continue using it in this fashion)

Are there even any consumer-level video games out there that used any sort of AI principles when programming the computer opponent? I wouldn't be surprised if they exist, but it can't be very common. I don't think this is the way to go to give us a Civ computer opponent that's more fun, even. I mean, it could be.. but I would be tempted to try more traditional approaches first.

As for which games have good computer opponents.. I am going through my gaming library trying to pick one out that's "good", but.. that's not really the point. My point is that it shouldn't be "bad". It doesn't even have to be "good". It could be "average" and I'd be happy with it. I'm just bored of the way the difficulty level is implemented in the game.. It's lazy

I get you. But it's one thing to point out a problem. It's another to actually address it. And I'm not saying you should, or anyone here should!

I agree. I'm just annoyed that they've never even tried. Every iteration of the game, they just pump out the exact same appraoch. It works "well enough" from their pov so they keep doing it.

I get it, but I don't like it.

But in this specific case, I feel pretty comfortable in choosing my hill to die on as "the developers did, in fact, try".

It seems that they weren't even given a chance to succeed. They were told what to build and they built it. The suits made those decisions for them

 

[uppi]

False. Sufficiently strong AI would likely evaluate the player as among the weakest nations/positions on the board. And it would be right.
Far less than dogpiling, it would likely leverage an advantage from better play to kill you off while trying to minimize the gains other AIs get off you. Like how good players target bad players in FFA. They don't "dogpile" them. Why share the spoils?

The player being killed increases the win chance of every AI in the game, so the AI will learn to cooperate to take down the player. And once a player is on the losing side of a war, the winning move for other AIs would be to join in to get the spoils before someone else does. Any AI that could take a city would, no matter the previous relations with the loser.

False. This is not what observed best practices are in games where each faction has a tiny % of the total resources available in the game. It would depend on situation, to evaluate when it's time to backstab. Though in a game with one and only one winner, the notion of "backstabbing" is an awkward, fake concept unless we're talking about breaking a specific agreement, which Civ often doesn't even allow.
False. This is not what good play looks like. I submit actual performance of Civ 5 AI as evidence. You don't want high investment, slow wars by attacking someone at first opportunity, if you're trying to win. You want to pick a target where you can win as decisively as possible.

Obviously, the backstab would be a competent attack, not the bumbling around of the Civ 5 AI. But the point is that players like to see opinion modifiers of your relations with another civilization. But this makes no sense if the AI is trying to win. If it sees an opportunity to quickly subjugate your empire, the +100 best friends forever modifier should be meaningless.

A good AI would *define* the meta, based on the current game rules. If the result lacks immersion, it's an indictment of the game's design/balance. And indeed, I will call out Civ games in this regard. They add lots of "options", aka consistent false choices that outright depend on opponents not trying to win.

Any strategy game will have false choices. If you cannot lose on turn 1, why have a turn 1? Trying to balance a game that there a multiple paths to victory is extremely hard. If you are good and lucky, there are two or three, anything beyond that is quite impossible to balance. So if you are trying to make a game about a complex concept like civilization, there will always be suboptimal systems. Is the answer really to strip those away?

You otherwise seem to hold some odd conceptions about what a "good" AI does or would do. But let me give an example to demonstrate why the quoted statements are false:

In Dominions 5, diplomacy is limited. There are no formal truces/forced peace treaties, nothing to compel you to act in any way. It's a true free for all, with anybody able to attack anybody else at any time for any reason they want. And like in Civ, only one person or team can win.

The best of the best, the players that win the most consistently by far...do you really think they do not engage in diplomacy with other players in this environment, even when there can be only one winner? Do you believe these players immediately "backstab" other people, or break agreements they make? Do you think they all gang up on one player, in a game where everyone is at their own level of play (or worse)? Do you believe these great players behave diplomatically like the Civ 5 AI (lol)?

If you do not believe these things, what makes you believe a "good" AI (in the sense that it's trying to optimize its winrate and is well-trained to do so) would do these things? How often did the Civ 5 AI which rolled to engage in such behavior win the game, on average? At best, you could make the case that Civ 5 AI mimics an aggressive human player who is bad at the game in both micro and macro decisions. But using that as a model for predicting how a "good" AI would behave is naive.

The point of diplomacy in a game that only has one winner is to convince other people to make unfavorable deals. You use the whole spectrum of human interaction to achieve that and hope that the other player does not notice how this will help you win. But you cannot do that with a good AI, which will always know better than you what the consequences of a deal are. So diplomacy with a good AI would be a losing move for a human player.

 

[Gori the Grey]

I'm joining this late and haven't read all of the posts, but I have two little observations:

1) how demoralizing would it be as a programmer to go to all the trouble to develop an AI that could challenge diety-level players and then have to dumb it down for all the other levels!

2) if I were lead developer on a new Civ, I at least know how I'd approach the matter. I'd have the topic of one (or several) of the early brainstorming sessions be "how can we build a combat system so that a computer will be good at it?" I think up to now they have some idea how combat ought to occur and then they build the best computer AI to conduct that kind of combat. But what if they asked "what are computers better at than humans? and how can we make combat in our game a function of those abilities?" The answer is "number-crunching." So you'd build a combat system where super-complex math is required to maximize advantages. (Of course then players would just develop computer programs to help them do that super-complex math).

Ditto, by the way, for all other elements of the game. What kind of victory conditions would computers be especially strong at achieving? Well then, make those the victory conditions in the game.

 

[Fippy]

Always attacking the player would create an unfair AI (and predictable), not a good one..
One of IVs strongest mechanics is indeed diplo. Their personalities are a big part of what makes things fun.
Starting near Hatty or Monty creates different situations. If Hatty would also constantly look to backstab you, all games would play similar.

I consider IVs AI as well programmed overall.
Sure they make silly moves here and there, but you can lose on deity no matter how good you play if things go against you.
Which is much more than can be said about V and VI, from what i heared.

 

[Sofista]

I think creating a serious AI matching player level for each difficulty setting would be tantamount to creating a handful and some distinct AIs, which would be both overly expensive and largely unappreciated by hardcore gamers, who make both the most vocal community and the one that would think it some kind of private embarrassment not to play the game at its most difficult. Handing freebies to the AI is just much easier and efficient.

 

[Angst]

Is there a kinship between Sid and George Lucas then? Both geniuses who made masterpieces and didn't even understand why those masterpieces were good?

Imo 1upt was a stupid idea and an artificial solution to a problem they invented. In that sense it was quite Lucas.

That said... people like what they like. People that authentically engage with 1upt do so. We have to respect that.

The question is more to figure out how to arrange the mechanics in such a way that the audience can engage with it.

Designer vision is damn overrated. Who cares. It's all about whether you can engage with something. I only mentioned it due to people *appealing to sid's vision* while not *recognizing what it was*.

Sid always wanted a civ6 style war. Sid didn't want doomstacks, but wanted a wide front. Whether the methods towards the mechanics were ok to him is another question.

 

[aelf]

That isn't a given in Civ 5 or 6 though, because of crowding. Player and AI alike can trivially fill up way more than the available frontage. This is part of the reason ranged is so dominant in the first place right now, and a major factor in the problems with 1UPT in general.

Trivially? Nah. Unless you're playing on a very small map or low difficulty, production is too limited to simply fill up the frontage with units. Maybe if you're simply aiming for a quick conquest victory.

 

[Ajidica]

If they were trying this, they got the opposite of what they wanted, at least in Civ 5. Changes to the underlying terrain-improvement-yield system (to say nothing of the removal of the slavery whip) made city-management and worker-turn decision far less meaningful in Civ 5 than they were in Civ 4.

Civ5 was a mess though, Jon Shaefer while a good expansion designer is not a good core game designer. Indeed, Civ5 on release was disappointing enough it almost completely killed my interest in the franchise. (That, and I discovered RPGs in college.)

 

[Hygro]

Civ 5 after all expansions is super duper fun especially my multiplayer which always was the best way to play

 

[Lemon Merchant]

Civ5 was a mess though, Jon Shaefer while a good expansion designer is not a good core game designer.

Agreed. The whole 1UPT thing is a hopeless mess (as was Civ5 and now Civ6).

Has anyone ever heard what happened to Schafer's much touted "At the Gates" game that he was designing with Kickstarter money?

 

[Hygro]

I find UPT just fine though my nostalgia is not that.

 

[Broken_Erika]

I won't touch Civ5/6 with a ten foot pole 'cause of 1UPT.

 

[Ajidica]

Agreed. The whole 1UPT thing is a hopeless mess (as was Civ5 and now Civ6).

I don't think 1UPT is a bad idea. It was just implemented really badly in Civ5 and Civ6 decided to focus on cities and terrain rather than units.

Has anyone ever heard what happened to Schafer's much touted "At the Gates" game that he was designing with Kickstarter money?

It took forever to release from its announcement and was generally regarded as 'meh'. It had some interesting ideas but never really came together as a compelling game, held back by an dull endgame and very slow gameplay.
https://www.eurogamer.net/articles/...iew-a-4x-experiment-more-fascinating-than-fun
https://www.rockpapershotgun.com/jon-shafers-at-the-gates-review