Interesting info from the gamespot preview

[Acken]

I'm aware AI programming is not as simple as that.
Still a competent AI require a programmer that knows how to play the game (or someone in the team). This is obvious.
If you program an AI and have absolutely no idea what is a good or bad strategy in a strategy game your AI will be easy to beat. Guaranteed in a TBS game as complex as civ.

Also of course there is a difference between AI works and good AI. That's not the topic. I'm only criticizing the conception that you don't need to be good at the game (or have information) to make a competent AI. Whatever the goal was for civ5 is irrelevant for that argument and we all saw how well civ5 AI has been received when stopping at "it works".

Whether or not having a competent AI is fun is also an entirely other debate and will vary from player to player.

 

[stealth_nsk]

Also of course there is a difference between AI works and good AI. That's not the topic. I'm only criticizing the conception that you don't need to be good at the game (or have information) to make a competent AI. Whatever the goal was for civ5 is irrelevant and we all saw how well civ5 AI has been received when stopping at "it works".

That wasn't my point. Of course developers need to be familiar with the game and know how to play it. And it was already the case with vanilla Civ5 AI being designed by Ed.

I disagree with people who thought inviting pro players would improve the AI development. That's surely not - Ed who is both civ player and competent AI programmer is by far better than any pro civ player without AI development experience.

 

[Acken]

At AI programming ? Of course. Nobody's arguing that.

Hell, I'm not even arguing that they need to invite people as they don't seem to have a structure for that kind of testing. However, once the game is out players will find issues. Will find AI weaknesses and have ideas on what the AI should do to compete.
That's where I suggest to look into it and implement improvements based on players experience.

There's also a balancing job along the way that works the same. With the help of player feedback to solve balance issues. This will also help the AI at the same time.

At least IF you want your AI to be competitive.

 

[stealth_nsk]

At least IF you want your AI to be competitive.

That's the key point. You don't need AI to be competitive for the game to be playable. Actually it's nearly impossible to make AI for a game like Civilization competitive. Even without 1UPT try playing Civ1-4 on difficulty level which don't give any bonuses to AI.

What the game really needs is:

1. The game needs to be competitive. A game as a whole, not only AI.

2. A game should avoid exploits. It doesn't matter whether you exploit game by tricking AI armies to move back and forth or by beelining to cheesy tech, exploit is exploit.

3. Finally, for immersion reasons AI should avoid looking too dumb. It's not necessary for game to be playable, but it really helps the game to be enjoyable.

 

[dexters]

I like that they'll likely have AI v AI option out of the box (hopefully) since it's kind of blown up on Youtube with Civ5.

I hadn't realized it was this popular since I was doing it way back with Civ3 with the Debug games I was running. It's a lot of fun to learn AI behavior and what the devs programmed for.

 

[Acken]

That's the key point. You don't need AI to be competitive for the game to be playable. Actually it's nearly impossible to make AI for a game like Civilization competitive. Even without 1UPT try playing Civ1-4 on difficulty level which don't give any bonuses to AI.

Of course you don't need it. You also don't need your game to be balanced to be playable. But unless the goal is to settle for mediocrity I'm not sure what is the point.

When I say competitive I'm not saying it should be pro level on Prince. That would be unreasonable.

But let me take a simple example. In civ5 sometimes the prince AI just doesn't expand beyond 2 cities for 100 turns. This could have been easily fixed in the 5 years we have had civ5. Or the fact that the AI neglects Aqueducts. These are just poor play.
Things like that that I'm suggesting should be improved over time. And as you know economy is the easy part of AI programming compared to tactical.

 

[OnceAKing]

While I agree AI vs AI testing isnt the end all be all. I do think that testing alongside some data could help make for AI capable civs. Perhaps if the saw data that showed America or Venice never won in CIV 5 in AI only battles, maybe they'd say let's tweak their attributes or programming. So I like if for that aspect

Also it's a hope against hope but maybe they'd notice weird AI habits like that shuffle of soldiers they like to do right next to your border. If they can find those before they hit the market great!


And I do value his emphasis on TSL and I don't think it's zero sum. I could see your fear for "quota" per se but I think Khmer not being in is as much about having 2 versions of Italian culture civs and several moderate central European civs in the game as much as its about Siam.


I just assume by focusing on TSL they'll be mindful that once we account for the must have civs like Britain or France for example, let's not throw Austria or Sweden in the game for shiggles, at least until we can explore gameplay from a Southeast Asian civ or a North American civ. I really think it's about that more so than only doing one non European civ per territory.

 

[stealth_nsk]

Of course you don't need it. You also don't need your game to be balanced to be playable.

That was point 2 - exploits.

But let me take a simple example. In civ5 sometimes the prince AI just doesn't expand beyond 2 cities for 100 turns. This could have been easily fixed in the 5 years we have had civ5. Or the fact that the AI neglects Aqueducts.

Why it bothers you? You think Prince is too easy for you? You could always increase the difficulty level. The AI not building Aqueducts - how it affects your gameplay?

What I'm trying to say is what this part is pure psychology. You could still play the game, it still gives you challenge, but once you see things done by AI which you consider obvious mistakes, you feel disappointment - you wanted to win over strong opponents, not dumb ones.

Of course, it's important thing to do, but has lower priority than game stability or balance.

 

[Acken]

Dude i have no clue where you got that it should come before balance and stability...

I feel this whole conversation to be pointless since you keep forgetting the point being argued. Im talking about how ai could be improved and your whole point is basically that it does not need to be. I argue the how while you focus on why.

As to the why. Im one of many that prefers competence over bonuses. Go figure. And while bonuses wont disapear the degree of inaptitude with or without them is staggering.

Also balance issues and exploits are 2 different things.

Ill remember playing above Prince. Thanks for this advice rofl.

 

[stealth_nsk]

Dude i have no clue where you got that it should come before balance and stability...

I feel this whole conversation to be pointless since you keep forgetting the point being argued. Im talking about how ai could be improved and your whole point is basically that it does not need to be. I argue the how while you focus on why.

If you try to explain me how to kill myself, I will ask you why too. If the goal is wrong, I don't see reasons to discuss how to achieve it. The things you're suggesting to improve AI seem to have nothing to do with how good the game is.

 

[Acken]

I wrote down why.

 

[Felis Renidens]

I don't want or expect the AI to give real competition. If I wanted that I would have played against humans. I expect the UI to punish me for neglecting an aspect of play (attacking if I'm weak, building cities if I failed to claim the territory, stealing wonders and city states etc.). Also it gives me a peaceful alternative to expansion. It's an element of the game that I play with.

I guess that just like me not all players would actually enjoy a competent AI. Then at easier levels it will be giving them bonuses which isn't as good.

 

[stealth_nsk]

I guess that just like me not all players would actually enjoy a competent AI. Then at easier levels it will be giving them bonuses which isn't as good.

From gameplay perspective, there's no difference between giving AI "skills" or just increasing its bonuses. "Competent" (which actually means doing things which mimic human behavior to human observer) AI is just an immersion thing, which most (although not all as you correctly pointed out) players enjoy.

Personally, I like "competent" AI myself, but the amount of importance some players put for it is funny to me.

 

[Felis Renidens]

From gameplay perspective, there's no difference between giving AI "skills" or just increasing its bonuses. "Competent" (which actually means doing things which mimic human behavior to human observer) AI is just an immersion thing, which most (although not all as you correctly pointed out) players enjoy.

Personally, I like "competent" AI myself, but the amount of importance some players put for it is funny to me.

I think it's easier to have different levels with bonuses than creating different AI behaviors.

 

[KrikkitTwo]

From gameplay perspective, there's no difference between giving AI "skills" or just increasing its bonuses. "Competent" (which actually means doing things which mimic human behavior to human observer) AI is just an immersion thing, which most (although not all as you correctly pointed out) players enjoy.

Personally, I like "competent" AI myself, but the amount of importance some players put for it is funny to me.

Its the immersion,
An incompetent AI gets exploited in ways that break immersion.

This especially damages immersion in play at the highest levels where exploiting the AI is more vital for winning.

That is why you want pro(advanced) players to provide feedback to the AI team. Those players will figure out the exploits rapidly.
Now the AI team then needs to work with the designer to figure out
1. designer-Should this exploit be avoided (do we actually want the AIs acting like this?)
2. AI team-Can this exploit be avoided without adding new, worse exploits.

 

[Acken]

I think it's easier to have different levels with bonuses than creating different AI behaviors.

No need for different behavior. Although that is already the case in a minor way in civ5.

 

[Felis Renidens]

You should be able to win most of the time when playing at 'prince' (same level as AI) even if you are not a 'pro' as long as you don't make serious mistakes. That's how I see it. Victory at higher levels should also be possible if you know what you are doing.

If the AI is playing like a pro player that won't happen.

Some things like unit shuffling shouldn't happen (they should fortify instead) but you don't need a pro player to notice such things.

 

[rschissler]

I don't want or expect the AI to give real competition. If I wanted that I would have played against humans. I expect the UI to punish me for neglecting an aspect of play (attacking if I'm weak, building cities if I failed to claim the territory, stealing wonders and city states etc.). Also it gives me a peaceful alternative to expansion. It's an element of the game that I play with.

I guess that just like me not all players would actually enjoy a competent AI. Then at easier levels it will be giving them bonuses which isn't as good.

A lot of truth to this. A player's goal should be to win the game and have fun doing so. Winning a game at the highest difficulty levels usually aren't that much fun playing the game, while the lowest difficulty levels can be very boring, so there should be a balance.

 

[Acken]

The challenge for "pro" should be on higher difficulties we all agree on that. But the ai performance on higher level is as much affected by the basic ai competence as the lower levels. And there is a point where bonuses dont suffice to create the illusion.

There are reasons why the number one fear people have is ai. Or why civ5 is systematicaly portrayed as being easier than civ4.

 

[stealth_nsk]

Its the immersion,
An incompetent AI gets exploited in ways that break immersion.

This especially damages immersion in play at the highest levels where exploiting the AI is more vital for winning.

That is why you want pro(advanced) players to provide feedback to the AI team. Those players will figure out the exploits rapidly.
Now the AI team then needs to work with the designer to figure out
1. designer-Should this exploit be avoided (do we actually want the AIs acting like this?)
2. AI team-Can this exploit be avoided without adding new, worse exploits.

Yes, but exploits are different thing. If you could force AI to move their units without attack (as it was in Civ3), that's an exploit. If AI requires more units to fight due to inability to use move and shoot, that's not exploit, that's just weaker AI.