Interesting info from the gamespot preview

[redwings1340]

I just watched gamespot's interview with Ed Beach, and while most of it was rehashing what we've been discussing, he said two things that were particularly interesting and new.

First, Civ VI is going to have a better implementation of a true earth map, probably with various civs in places they originally were founded in, with a plan to balance the true earth setup. This is pretty cool.

Secondly, the way they tested the game and the AI seems really cool. They set up a flat screen TV near the lead engineer's office, and had this TV playing AI vs AI games of Civ VI 24/7, for months. Whenever anyone updated the game and created a bug, the AI games caught it, and this means that they've gotten a pretty good idea of how the AI plays against each other, and a lot of data on how to improve it. While knowing what's wrong and being able to fix it are dramatically different things, I thought this was a pretty clever way to test out the game.

He discusses all of this starting at the 11:45 mark: http://www.gamespot.com/videos/civilization-vi-developer-gameplay-preview-with-ed/2300-6432860/

 

[Acken]

Yes but lets not forget that since this is a strategy game, a lot of the ai performance is linked to "playing well". And what is "playing well" is something humans will come up with.

Ai performance will then be linked to whether or not they improve the ai overtime to make it use good moves discovered by players.

A simple example is expanding. If the 1.0 ai is programmed to expand slowly because that is the developer guess at a good strategy. Then if player find that rex is by far a better strategy always, the ai needs to be improved to be able to compete.
Or rexing should be made less stronger or a combination of both.

What made civbe too easy was this combination of imbalances and ai ignorance allowing the player to just crush the hardest difficulty from day 1.

 

[stealth_nsk]

Both were already discussed. Developers told us about AI testing nearly right after announcement

 

[cpm4001]

The AI testing thing certainly sounds like it has potential, and agreed that it's an innovative way to test the game, but in all honesty I'm not sure if it'll actually improve gameplay, simply because AI vs. AI play will be inherently different from Human vs. AI play. I guess we'll see, though. The TSL thing we've known about for quite a while; though again it's promising my concern is that it'll lead to the inclusion of bland Civs chosen because they happen to have originated in a specific location, rather than because of interesting traits about the Civ.

Also, didn't know you were back, @Redwings, but good to see you again!

 

[stealth_nsk]

The AI testing thing certainly sounds like it has potential, and agreed that it's an innovative way to test the game, but in all honesty I'm not sure if it'll actually improve gameplay, simply because AI vs. AI play will be inherently different from Human vs. AI play. I guess we'll see, though. The TSL thing we've known about for quite a while; though again it's promising my concern is that it'll lead to the inclusion of bland Civs chosen because they happen to have originated in a specific location, rather than because of interesting traits about the Civ.

For AI the biggest task is to have tactics look not stupid. That's the sort of thing you could mostly notice in AI vs. AI battles.

 

[Aheadatime]

Well Beach stated he was the head AI guy for vanilla Civ5, and that he did AI vs AI tests all the time back then. That didn't seem to help Civ5's AI all that much, so I'm not sure why it would stand to reason that it'll help Civ6's AI this time around.

I think Firaxis needs 'pro' civ players to help hash out early strategies while the game is in development. This way, they can spot out imbalances and tweak the AI to mimick these pro players do a degree.

 

[stealth_nsk]

Well Beach stated he was the head AI guy for vanilla Civ5, and that he did AI vs AI tests all the time back then. That didn't seem to help Civ5's AI all that much, so I'm not sure why it would stand to reason that it'll help Civ6's AI this time around.

The goal was different. In Civ5 they were trying to have AI so the game is playable and it was achieved. Having AI look smart is completely different and much more difficult goal. Civ5 developers didn't have resources for such luxury, for Civ6 we could hope they have.

I think Firaxis needs 'pro' civ players to help hash out early strategies while the game is in development. This way, they can spot out imbalances and tweak the AI to mimick these pro players do a degree.

Have you ever heard of chess grandmaster making a chess game? Game skills have nothing to do with making game, AI included.

 

[SupremacyKing2]

Yes but lets not forget that since this is a strategy game, a lot of the ai performance is linked to "playing well". And what is "playing well" is something humans will come up with.

Ai performance will then be linked to whether or not they improve the ai overtime to make it use good moves discovered by players.

A simple example is expanding. If the 1.0 ai is programmed to expand slowly because that is the developer guess at a good strategy. Then if player find that rex is by far a better strategy always, the ai needs to be improved to be able to compete.
Or rexing should be made less stronger or a combination of both.

What made civbe too easy was this combination of imbalances and ai ignorance allowing the player to just crush the hardest difficulty from day 1.

Agree completely but I do think that AI v AI can be a good way of testing certain things about the AI.

1) It allows A/B testing to see if new AI code is doing what it is supposed to do. The devs can give one AI the old code and the other AI the improved AI code and see what happens. The devs can also use the same saved game and run the same with differen code and see if the AI makes a better move. The review mentions the devs running AI v AI games after updates so it seems they did this a lot.

2) The other thing is that it allows the devs to watch for dumb AI mistakes. For example, if the devs see an AI build a navy and have lots of units on the shore but never use them, they can see that the AI was not able to do a naval invasion.

So hopefully, it will help the devs at least identity the obvious AI weaknesses.

 

[redwings1340]

Also, didn't know you were back, @Redwings, but good to see you again!

Thanks! I've been lurking around here from time to time, making the occasional comment. I'm really excited about civ VI, it looks to me like about 90% of the changes they're making are just generally great ideas, and the other 10% I'm willing to wait on to see how it goes. Nice to see you around here too!

 

[Callonia]

I am fairly confident that firaxis won't fix blatant flaws like AI's inability to move and shoot in same turn. I don't understand why firaxis won't patch that for civ5.

This gives a worrisome precedent for civ6.

 

[cazaderonus]

The goal was different. In Civ5 they were trying to have AI so the game is playable and it was achieved. Having AI look smart is completely different and much more difficult goal. Civ5 developers didn't have resources for such luxury, for Civ6 we could hope they have.



Have you ever heard of chess grandmaster making a chess game? Game skills have nothing to do with making game, AI included.

Your comparison with chess is actually a bad one. With the recent victory of an AI vs the best Go player in the world, dozens of articles were published about AIs. And what came out of it is that AIs are great at games that are made of a single layer of gameplay. Combinations are then extremely easy for them to use and calculate.

On the other hand, games with multiple systems linked to each other are still a huge roadbump for AIs and simply cannot compete with players on an equal setting (prince difficulty for example).

So yeah for a game like civ, players skill is what should drive the construction of the AI. At least so that they have interesting patterns on various system ( bpstarting build order and objectives, composing an army, positionning units, selecting policies etc). That s why some smart ai mods were a huge improvement on civ v, because they were based on player tactics.

And also, i d be very surprised if the people who programmed the chess ai in the first place didnt use the moves from the best chess players as a base ground.

 

[Acken]

Agree completely but I do think that AI v AI can be a good way of testing certain things about the AI.

1) It allows A/B testing to see if new AI code is doing what it is supposed to do. The devs can give one AI the old code and the other AI the improved AI code and see what happens. The devs can also use the same saved game and run the same with differen code and see if the AI makes a better move. The review mentions the devs running AI v AI games after updates so it seems they did this a lot.

2) The other thing is that it allows the devs to watch for dumb AI mistakes. For example, if the devs see an AI build a navy and have lots of units on the shore but never use them, they can see that the AI was not able to do a naval invasion.

So hopefully, it will help the devs at least identity the obvious AI weaknesses.

Oh yeah I'm not arguing it's bad. It's necessary even. I do it all the time when working on my mod.

But it's not the whole story. Good players have to chime in and points issues.

Have you ever heard of chess grandmaster making a chess game? Game skills have nothing to do with making game, AI included.

There are no good AI without good players. Someone has to know enough about the game to "teach" the AI that. Especially in a computer game where a "learning AI" is out of the picture. That doesn't mean the programmer himself has to be a master of the game. But maybe if a master says "in X situation do Y" the programmer should probably take that into account.

 

[Loaf Warden]

The TSL thing we've known about for quite a while; though again it's promising my concern is that it'll lead to the inclusion of bland Civs chosen because they happen to have originated in a specific location, rather than because of interesting traits about the Civ.

My concern is the opposite of this. I worry that they'll overlook interesting civs because they already have a civ in that region and so they'll figure that region is already "represented". I assume that's the basic reason why Mali and the Khmer were left out of V after already being in IV; in V they put in Songhai and Siam and then never went back and revisited those areas. With the TSL consideration being such a strong influence on civ choice, I'm concerned that will only amplify the problem.

I'm not sure there's such a thing as an inherently "bland" civ. In theory I think any civ could be implemented in either an interesting or a bland way. If Greece turns out boring, it'll be because it was implemented poorly, not because the Greeks themselves are boring. If [insert obscure tribe you've never heard of] gets announced as a civ, and you look at the tribe and decide it's not interesting, that doesn't mean that the implementation in-game won't end up really cool.

Personally, I want to see Mali and Songhai in the same game . . . along with the Benin and the Ashanti. I want to see the Khmer and Siam in the same game . . . along with Lan Xang and Vietnam. I don't want to see them just pick one from each of these groups and say, "Welp, that region is covered now. Now which Western European civs have we not done yet?"

 

[cpm4001]

My concern is the opposite of this. I worry that they'll overlook interesting civs because they already have a civ in that region and so they'll figure that region is already "represented". I assume that's the basic reason why Mali and the Khmer were left out of V after already being in IV; in V they put in Songhai and Siam and then never went back and revisited those areas. With the TSL consideration being such a strong influence on civ choice, I'm concerned that will only amplify the problem.

Though I can get the concern, I hope this won't happen and somewhat doubt it will, at least not to the extent you're worried about. In my view, a greater TSL focus will probably mean that, though Western Europe will undoubtedly be overrepresented, it will also be considered 'represented' rather faster than other areas, so they'll put more effort towards spreading balances of civs into those other areas.

I'm not sure there's such a thing as an inherently "bland" civ. In theory I think any civ could be implemented in either an interesting or a bland way. If Greece turns out boring, it'll be because it was implemented poorly, not because the Greeks themselves are boring. If [insert obscure tribe you've never heard of] gets announced as a civ, and you look at the tribe and decide it's not interesting, that doesn't mean that the implementation in-game won't end up really cool.

When I say a civ might be 'bland' I'm mostly referring to gameplay implementation, not necessarily anything related to the civ's history. I agree that just about any civ (regardless of where it originated) could be implemented in an interesting way; my concern is that they'll pick a civ because of its geographic origin and then just tack on nonsensical or lame Uniques/Agendas/etc., either out of laziness or because the selected civ doesn't have good potential for interesting gameplay mechanics.

Note though that I'm not suggesting RotW civs would be the only ones (or even the main ones) susceptible to this sort of thing; the best example of this happening that I can think of off the top of my head is Austria in Civ V - a relatively dull civ chosen because of its location and so taking the place of a civ (Kilwa? The Mapuche? The Chinook?) that might have not only been geographically more unique but also had some really neat possibilities for interesting mechanics.

Personally, I want to see Mali and Songhai in the same game . . . along with the Benin and the Ashanti. I want to see the Khmer and Siam in the same game . . . along with Lan Xang and Vietnam. I don't want to see them just pick one from each of these groups and say, "Welp, that region is covered now. Now which Western European civs have we not done yet?"

Agreed; though Western Europe had and has plenty of interesting civs (and shouldn't be ignored by any means), that particular overemphasis in Civ V did get a bit old. Ideally of course we end up having Spain, France, Germany and the Netherlands AND Mali, Songhai, Benin and the Ashanti AND the Khmer, Siam, Lan Xang, and Vietnam, but if it can't happen due to limits on the number of civs then at least selecting a few - rather than just one - from each region would be the best possible option

 

[Loaf Warden]

When I say a civ might be 'bland' I'm mostly referring to gameplay implementation, not necessarily anything related to the civ's history. I agree that just about any civ (regardless of where it originated) could be implemented in an interesting way; my concern is that they'll pick a civ because of its geographic origin and then just tack on nonsensical or lame Uniques/Agendas/etc., either out of laziness or because the selected civ doesn't have good potential for interesting gameplay mechanics.

Fair enough. Their implementation of India has historically been very poor indeed, and that's certainly not because freaking India is an uninteresting place. To the extent that I don't want to see civs implemented poorly, with uniques built from ignorance or lazy stereotypes (for example), then I agree with you. I'd rather see a civ get left out than made badly.

Ideally of course we end up having Spain, France, Germany and the Netherlands AND Mali, Songhai, Benin and the Ashanti AND the Khmer, Siam, Lan Xang, and Vietnam, but if it can't happen due to limits on the number of civs then at least selecting a few - rather than just one - from each region would be the best possible option

I completely agree with this. I'm actually not bothered by the number of European civs, per se. It's the ratio of European to non-European, and the token way non-European civs get treated, that annoys me. I'm perfectly happy to have Germany and Austria in the game together, but I hated that they just put in Siam and apparently decided that one civ represents all of mainland Southeast Asia well enough that we don't need anyone else from there. My worry is that the TSL consideration will make that problem worse, but if you're right that they'll end up filling up Europe and then looking toward other places and doing more with them, then I'll be satisfied with that.

 

[m15a]

Indonesia is also in Southeast Asia. So, it's pretty apparent the developers didn't think that one Civ represented all of Southeast Asia. (And even if they didn't add Indonesia, the opposite claim still wouldn't be apparent.)

 

[labellavienna]

Omg that is so awesome op! Thanks for this tidbit

 

[Loaf Warden]

Indonesia is also in Southeast Asia. So, it's pretty apparent the developers didn't think that one Civ represented all of Southeast Asia.

I said mainland Southeast Asia. I didn't forget about Indonesia, and I also think that Siam and Indonesia are far enough apart to not cause each other to be overlooked on a TSL map. But it isn't as though they looked at the region and decided to bring the Khmer back as well, or include any form of Vietnam, say. And if they have Siam and Indonesia both in VI, will that mean they're less likely to add any form of Malaysia?

(And even if they didn't add Indonesia, the opposite claim still wouldn't be apparent.)

I'd be perfectly happy to see evidence that the Khmer weren't overlooked in Civ V because of regional considerations. I said they "apparently" decided Siam was enough, not that I know they definitely thought that way. But Beach has said that they're looking at TSL maps to decide which civs to include for Civ VI, which means civs will be chosen because of regional considerations this time. I'm just concerned that they'll look at a region, such as mainland Southeast Asia, see that there's a civ there, be satisfied that the region is covered, and move on to somewhere else instead. I don't say I know this will happen, just that I have some concerns that it might.

 

[redwings1340]

From what I can tell, there are just a lot of civilizations and we want representation in all of them. Every civ takes a good deal of time to balance and figure out. Maybe the choices in civ V weren't ideal, and a few could have been different, but I feel like the developers are in a no win scenario here, no matter what civs they choose, they'll annoy somebody.

 

[stealth_nsk]

There are no good AI without good players. Someone has to know enough about the game to "teach" the AI that. Especially in a computer game where a "learning AI" is out of the picture. That doesn't mean the programmer himself has to be a master of the game. But maybe if a master says "in X situation do Y" the programmer should probably take that into account.

No, that's not correct. Human definitions of X and Y would require too much programming to implement themselves. AI programming works with completely differently terms than human player.

Also, important thing to consider is the definition of "good" AI. The general requirement and the one achieved in Civ5 is to actually make AI play. You see, AI could build empire with significant number of cities, it uses buildings, improvements, military units and so on. On high difficulty levels AI could beat average player in both military and peaceful victories. This means AI works.

More complex requirement is to make AI look smart (ir can't be actually smart, because it's AI) - it wasn't goal for Civ5 design. That's part of immersion, not gameplay as it doesn't make game more playable, it just removes player frustration about "bad" AI.

I'd recommend brilliant video of Soren Johnson "Playing to Lose: AI and Civilization". There's also great article of Jon Shafer regarding AI looking smart or dumb where he analyzed why AI in Civ5 was considered bad. It's somewhere on his site https://jonshaferondesign.com/