Play like a nation or play to win?

[Naokaukodem]

What I meant was that the fact we were forced to make peace were more or less simulating the nation role playing. Role playing is not a gameplay feature as we mean it usually, it is not a game mechanic, even if it can be a factor of fun. I think that the question about playing 'like a nation' of 'like a player' touches this role playing part. Do I have to attack any unit as soon as I have the opportunity of it, or should I show myself more disinterested in earning an advantage of each situation regarding the arbitrary winning conditions, simulating more human feelings/peace aspiration for example? Of course, I think that giving the possibility of the two would be perfect. Don't say me it is already the case, as it could be way better, regarding both game-play (game strategies: stacks, culture influence...) and nation-play. (war orientation of Civ3 particularly in higher difficulty levels)

If you don't have the exact rules, that is precisely because Civ is not a math tool. It would not be interesting this way. It is the same reason why battles are more or less predictables. Well, the fact is that in Civ3 we can outpass this by raising the number of units involved. To go until the end of this logic, units should have been way more expensive to maintain. That is also what make me say that in Civ3 research is not THAT MUCH limited by war, except if you practice it exclusively neglecting city growth and building (ICS?), what is valuable mostly from the very start of the game.

Civ is about 'simulation of bacterias', but at a very lower scale, with only 8 civs starting with a standard map... size, strengh and complexity don't really have an opportunity to really express, unless basically, and for that reason, does not exist that much (especially complexity). Oh, and the big does not always eat the small, as AI even in Civ3 remains smart role/nation -playing wise, not systematically attacking you or another one. Though, making so could make a good challenge: a 'bacterias simulation', with the same evolutions everywhere on the globe at the same time (except for isolated areas). This would combine intense gameplay (wars, strategy..) and nation-playing, as we would consider every nation naturally aspiring to conquests.

 

[Sirian]

It doesn't matter how the AI plays as an individual. It matters how the AIs play collectively.

Diplomacy is the overarching issue of single player.


- Sirian

 

[KrikkitTwo]

The way to resolve issue is
1. AIs should play like players (with the exception that they should play like players that are dedicated to seeing the game through, I'd hate to see the German AI hitting reload after their attack on the Polish activated my MPP with them, or have the Roman AI quit the game in ~450 AD... For that they also need goals Besides winning...points seems the most reasonable)

2. Playing to win should entail playing like a 'winning nation'. The game rules should be designed so that certain strategies that seemed to work for a while in History should probably be the strategies used on the path to victory. If the game rules encourage a strategy that seems ridiculously inconsistent with what 'feels right', the game rules need to be changed, so that the strategy is no longer encouraged, rather than making the AI ignore this good if ridiculous strategy (ie mass late Game ROP violation).

 

[HourlyDaily]

Krikkitone you started me thinking - does the AI set goals? How does it evaluate them?

You know how when they invade via sea, they drop of 2 units and leave it at that. Instead of the AI using turn to turn thought, would extended goal-setting improve it? If the AI had a list of 10 things it wanted to achieve that changed as the game unfolded (and those goals were achieved or not) then it would seem to play like a human player. At least if it carried through its decisions.

The skill would be in getting the AI code to recongnise when to make a decision and what criteria to use. I've seen some AI models for various things here, but I don't know the extent of the AI's goal setting. If long term strategy is what we want so the AI plays smart, it can also be tailored to heavily weight the Civ leaders preferences into its decision.

I know little about AI coding, but is this goal-setting idea feasible? Does it apply to Civ 3? (like when they send 12 units to raid one of your cities halfway across the map, and you declare peace very easily) and if not, what strategy games use it as part of AI coding? I'm a little short of food for my brain at the minute so I apologise if this is a stupid post.

I don't want to hijack the thread - but maybe if this is discussed here it can solve the dilemma.

 

[Darwin420]

I play as a nation. It's more immersive that way.

 

[kryszcztov]

was i? i mean, i didnt lose a single life until level 54. then i lost all 3 of them in the next 2 levels. should i really start over?

YES in that game. And many of such games have continues, or, better, checkpoints. I think that even Super Mario 1 makes you restart at the beginning of the 4-level world.

"train" would be the right word. but, IMHO, creativity should be stimulated even more. but thats just an opinion!

Both games can co-exist. I think the article's point was that there were so little, "old school" games today, and this would need to be addressed somehow.

but if you stop, it means the game stopped being fun, doesnt it? a good game will be fun forever. even if you reach the end of an arcade game, it will have an end! :-\

A fun game can be one where you're constantly asked to find the key to the next level and where it's becoming really tough. In this regard, the longer the better. The Zelda series is a nice example. This is a game that the majority of players and critics will find at least good, if not excellent. Yet there are usually moments in this game where you're completely puzzled (good word ! ) and can't find the key (metaphor or not).

 

[mastertyguy]

I play to have FUN!!!!!!!

 

[Deep_Blue]

The AI doesnt understand what FUN is .
The thread was originaly asking about how the AI should play.

 

[Yusaku Jon III]

I chose "as a player" because, frankly, I'd like to be the winning civilization at the end of the game. Hell, even real-world civilization had some spells where they saw themselves as the victor.

For instance, Rome. Did they just stay there in Italy after driving the Greeks and Carthaginians out? Nope, because they knew that if they didn't go straight to Carthage or take out the Macedonians soon, there'd be another round with either of the two rival civilizations, and this one would likely have ended up going straight to Rome instead of messing around in Italy.

If you wanted a non-warlike game, all you'd have to do is turn off the Domination and Conquest victory options and reduce the number of AI civs to a more comfortable level. That way, wars are less likely and you'd have plenty of room to expand your borders.

 

[Aussie_Lurker]

Yusaku, this has nothing at all to do with whether civ players want a non-warlike game or not. It is about whether the AI should behave like it is a true nation-guided by certain underyling principles, ideologies and agendas, or merely behave like a player who's only goal is to win 'ultimate victory'. Trust me, even an AI playing like a 'nation' will still start wars, destroy neighbours and rivals and build empires, its just that they will do it for a reason simply beyond 'winning the game'.
Hope that clarifies things for you.

Yours,
Aussie_Lurker.

 

[apatheist]

Trust me, even an AI playing like a 'nation' will still start wars, destroy neighbours and rivals and build empires, its just that they will do it for a reason simply beyond 'winning the game'.

If playing like a nation diverges from "winning the game," maybe they should redefine "winning the game."

 

[Hygro]

If playing like a nation diverges from "winning the game," maybe they should redefine "winning the game."

Brilliant! Somebody said that earlier but you made the point much, much clearer.

 

[Gr3yL3gion]

I wonder if anyone would like to do a Leeroy?

 

[warpstorm]

It doesn't matter how the AI plays as an individual. It matters how the AIs play collectively.

Diplomacy is the overarching issue of single player.

I disagree with you here. While the Collective gang up of Civ2 felt wrong and should be forgotten about in future designs, I would like the AI to individually play in their best interests. (Now does that mean like a nation or to win?)

 

[Naokaukodem]

I disagree with you here. While the Collective gang up of Civ2 felt wrong and should be forgotten about in future designs, I would like the AI to individually play in their best interests. (Now does that mean like a nation or to win?)

Yeah, btw what is a collective AI: does that mean that every AI should gang up against us at every time, for example doing war to us 100% of the time? This is silly.

 

[Sirian]

I disagree with you here. While the Collective gang up of Civ2 felt wrong and should be forgotten about in future designs, I would like the AI to individually play in their best interests.

You are making a nonexistant distinction. Playing individually to their best interests -is- the same as Civ2.

"Best interests" by definition means one algorithm: the best available. Civ2 did not design the AIs to gang up on the player. It designed the AIs to follow the same logic, which inadvertantly leads to the hive mind.

Civ3 makes the same mistake. It insulates against the gang-up, but there is still a hive mind: one the player can harness to his infinite advantage.

You only get to pick one. If you aim for individual competence when it comes to strategy, the AIs all end up cycling the same logic and chasing the same objectives. One look at the GalCiv AI will show you all the AIs chasing the same tech path, over and over in every single game. It sounds good in theory, but it doesn't pan.

If you aim for collective competence, the AIs have to fan out to cover more ground, more behaviors. Different paths will emerge as stronger or weaker on different maps, so it is not as if one path is always the winner and one is always the loser, but that there WILL BE winners and losers.

This is a scattershot approach, but it beats the sniper approach when the sniper is blind and can't hit the broad side of a barn.

Individually brilliant means collectively stupid, when it comes to free for all game AI. The answer in GalCiv is to avoid chasing the AI crowd and instead branch off in to the areas they neglect, cleaning up all the wonders and Trade Goods they ignore, and trading to all parties for the stuff they collectively pursued, knowing they can't trade with one another because they all have the same stuff.


Even "playing as a nation" and "playing to win" are uncrisp descriptions. The choice is more like "all chase the same path" or "fan out and chase different paths". Of course, this assumes the game is designed well enough to HAVE different paths.


- Sirian

 

[apatheist]

"Best interests" by definition means one algorithm: the best available. Civ2 did not design the AIs to gang up on the player. It designed the AIs to follow the same logic, which inadvertantly leads to the hive mind.

Two counter-points. One, behaving similarly is not the same thing as behaving cooperatively. Two, you cannot just focus on the algorithm; the inputs matter as well. The inputs in Civ4 will be pretty varied. Starting position, map size/shape, neighbors, and the like have been in Civilization since the first one, but the leader traits and unique units should really mix things up. Genghis Khan and Gandhi should both use the same, "best available" algorithm to make decisions, but war and expansion are cheaper for the former, and commerce and culture are cheaper for the latter. Those effects are significant and should perturb most, if not all, of the other calculations in the algorithms.

Even "playing as a nation" and "playing to win" are uncrisp descriptions. The choice is more like "all chase the same path" or "fan out and chase different paths". Of course, this assumes the game is designed well enough to HAVE different paths.

Which is the point I also have been trying to make. Personality should be a logical, derivative consequent of differing leader traits. Personality shouldn't be coded in explicitly, and the leader traits should be sufficiently distinct and different that they do lead to different paths being optimal in different cases. It's a tricky balance. If you make the traits too significant, you will always get the same strategy out of any given leader. If you make them too insignificant, you will always get the same strategy out of all AIs. Clearly, the former is less bad than the latter, but it's still not best; there's an optimum in the middle.

 

[warpstorm]

double post

 

[warpstorm]

Playing to win, means just that. The AI players are aware that there are victory conditions and are actively working towards fulfilling one or more of them to win the game and preventing the other players from doing the same.

Playing like a nation (to me) means that they are following a behavior that while plausible and hopefully varied by personality and environment does not follow the goal of actually trying to win the game or actively preventing a player from doing the same.

The big difference is that they are aware what the victory conditions are and actively factor them into their decisions.

 

[mastertyguy]

The AI doesnt understand what FUN is .
The thread was originaly asking about how the AI should play.

Doospie, my mystake!
We should mod "fun" in the AI!