Why I am concern about including Blake work and Khael FFH work in BTS.

[jkp1187]

Sorry, but it is incorrect. IN average only half of money avalible for trade by AI IN Vanilla and warlords 2.08.

I say in average, because it is not constant.

But if you switch human player to AI in multiplayer only half of his money will be avalible for trade, the rest AI will reserve.

Are you referring to Warlords 2.08 or the most recently updated version of Better AI?

I don't have time to find the post, but I recall Blake saying that he did not view 2.08 as being final, which is why he was working on Better AI in the first place.

You understand that some of us are speaking about innovations made for the Better AI mod above and beyond what was available in Warlords 2.08, yes?

Personally, I would much rather have an AI that acted like a human player would to a greater extent and relied upon cheats and handicapping less.

 

[Zoolooman]

Why not wait and test out the AI yourself when BtS ships, rather than bashing the work of someone who has done a tremendous amount of AI work on this game?

Mutineer's points are excellent, and I think it's worth discussing these problems before the expansion is released.

I followed Blake's AI closely. In the beginning, it improved city placement, city specialization, and AI military behavior. These improved the AI's competitiveness while maintaining the AI's fun-oriented behavior.

But with time, Blake began to make some changes with no regard to the AI's intrinsic advantages--and only after massive complaining did he even introduce the first (and only?) altered AI handicap file. I remember playing a Better AI version where Blake introduced the AI's new military defense logic. In my first game with it, Mansa Munsa produced 52 skirmishers to defend four cities and built almost no infrastructure. Whenever he felt I was producing enough forces to endanger him, he produced more defense, and so my army was always dwarfed by his massive defensive forces.

At first I didn't mind. I expected that this huge defensive investment would cripple Mansa, but lo and behold, Mansa was still rich! He had almost no upkeep costs and could afford to build all these units while nearly matching my research speed (thanks to more AI advantages!). When he hit Feudalism, he upgraded all 50+ skirmishers to longbows in one turn, and at that point, I decided that I would rather play Warlords 2.08.

It's that type of experience that makes me wary of Blake's contributions. One should prevent the AI from doing something stupid, but one shouldn't make the AI so effective that it abuses its own cheats like a player would--if those cheats were available to the player.

 

[Jet]

Even though Civ 4 is very rich, a person who plays it a lot will notice ways in which core game dynamics lack richness. This lack makes multiplayer Civ 4 simpler in some ways than single player. It leads to delicate tradeoffs when improving the AI.

 

[sooooo]

I understand mutineer's concerns, but I think we should wait to play the game until we judge. Likely BTS will play a lot different to BetterAI mod, which did have some problems sure. But working on it full time for all this time will have surely improved it a lot.

I think the new AI is by far the most important part of BTS. It has the potential to make the game much more fun, or also much less fun. Something like corporations or espionage will probably not have that much effect, but the AI changes will be huge. The enjoyment of thousands of people who love civ is in the hands of one Kiwi - I'm crossing my fingers

 

[jkp1187]

It's that type of experience that makes me wary of Blake's contributions. One should prevent the AI from doing something stupid, but one shouldn't make the AI so effective that it abuses its own cheats like a player would--if those cheats were available to the player.

Why are you assuming that a several-months-old project done for free is going to be the same as the final product in BtS?

 

[Zoolooman]

Why are you assuming that a several-months-old project done for free is going to be the same as the final product in BtS?

I'm not making any such assumption. Could you quote the specific line which accidentally implied that conclusion?

I thought this thread was to discuss the effects of Blake's and Kael's design philosophies. Since the only playable examples of Blake's philosophy are several months old, I can't help but discuss potentially outdated information. But even if the information is outdated, that shouldn't prevent us from having the discussion. As long as our eyes are open to the possibility that things have changed, I don't see any harm--and I see potential benefits--in discussing what we know and what we have.

 

[Saluki]

Perhaps I'm missing the point here, but is this actually an argument in favor of keeping the original semi-moronic AI in the game? Blake's AI may not be perfect (at least it wasn't a few months ago), but it was far better than what was originally shipped.

 

[Roland Johansen]

Mutineer's points are excellent, and I think it's worth discussing these problems before the expansion is released.

I followed Blake's AI closely. In the beginning, it improved city placement, city specialization, and AI military behavior. These improved the AI's competitiveness while maintaining the AI's fun-oriented behavior.

But with time, Blake began to make some changes with no regard to the AI's intrinsic advantages--and only after massive complaining did he even introduce the first (and only?) altered AI handicap file. I remember playing a Better AI version where Blake introduced the AI's new military defense logic. In my first game with it, Mansa Munsa produced 52 skirmishers to defend four cities and built almost no infrastructure. Whenever he felt I was producing enough forces to endanger him, he produced more defense, and so my army was always dwarfed by his massive defensive forces.

At first I didn't mind. I expected that this huge defensive investment would cripple Mansa, but lo and behold, Mansa was still rich! He had almost no upkeep costs and could afford to build all these units while nearly matching my research speed (thanks to more AI advantages!). When he hit Feudalism, he upgraded all 50+ skirmishers to longbows in one turn, and at that point, I decided that I would rather play Warlords 2.08.

It's that type of experience that makes me wary of Blake's contributions. One should prevent the AI from doing something stupid, but one shouldn't make the AI so effective that it abuses its own cheats like a player would--if those cheats were available to the player.

And after Blake had made the AI aware of a buildup of forces by its neighbours and made them react to this, he took away some of the handicap bonuses that the AI had. After that the AI might build too many troops so that it couldn't maintain them with the new and higher upkeep costs of the new handicaps for the AI. So the AI had to be taught to only use a certain fraction of its income for unit maintenance and this thus limited the number of units that the AI would build. And that is about where the BetterAI mod development stopped and Blake started to work directly for Firaxis.

I wonder why you stopped at the first incarnation of the new AI mechanic where the AI would react to the forces of its opponents and didn't tell us about the further improvements. Might it be because otherwise you wouldn't have a point to make?

Improvement of the AI is not something that will go right every time. You will try to improve it by introducing new AI routines but they won't do the exact right thing right away. It takes some experimenting and some ingame experiences with the AI from several game testers. The whole time that the BetterAI mod development was underway, we were playing a Beta version of a mod under development. New elements were being added and some worked well right away and some others needed refinement. The 1.0 version of BetterAI has never been finished because Blake started to work for Firaxis and Iustus didn't have time for it because of his new job.

The new handicap files weren't introduced due to massive pressure from the beta testers of this mod. One of the main design philosophies of the mod was to improve the AI so that it didn't need the huge handicap bonuses anymore.

 

[winddbourne]

The biggest thing that needs to change is the definition of difficulty level.

I WANT the AI to be able to do things like Mansa did in the above post by Zoolooman. But only at higher difficulty levels. Certainly not on settler or chieftain!

I hope that in the new expansion they scrap the idea of "AI cheats' and simply have different AI's for each level. So when I go up against a deity level AI it is actually five or six times smarter and more ruthless than the equivalent Settler AI.

The settler AI should make all the MISTAKES of a novice Civ player. But when you move up towards Noble or Deity the AI should become much more ruthlessly efficient.

As Roland said the point is to get rid of the need for AI bonus' altogether. If they do that, or even have bonus' start coming into play at Monarch/Emperer level instead of Noble level . . . then I'd be impressed.

I hope that makes perfect sense to everyone. I know I tend to ramble sometimes.

 

[ds61514]

Of course, when you're playing a far inferior opponent, then there are countless ways to beat him and countless tricks that you can use to beat him. So once the far inferior opponent doesn't do the most stupid moves (like trading away all of its money just before a war), the options to beat him become less and smarter moves are needed.

You said it yourself: the options to win become less. That isn't fun. Many people already say Civ4 is too one-dimensional, with early rushes and constant war more often than not being the dominant strategy. In the above situation, what smart move could you do? In that case, the "right" strategy was to kill the GM and go to war *again.* Wow, how exciting . Given equivalent difficulty, if it came down to crazy options (bribing Shaka into pacifism) versus realism (it'd be impossible to bribe Shaka into pacifism, but his army is 30% less), I'd probably prefer the former.

I dunno, I really think the whole "dumb AI with a ton of bonuses" versus "smart AI with few bonuses" is really a wash. Does it really matter if the enemy has 12 longbows (dumb AI) or 10 longbows and 3 catapults (smart AI)? Not really. It's just an extra move to split up the stack or wait a turn to heal. Big whoop. Wars are won with strategy, not tactics (especially Civ4 ones).

 

[Zoolooman]

I wonder why you stopped at the first incarnation of the new AI mechanic where the AI would react to the forces of its opponents and didn't tell us about the further improvements. Might it be because otherwise you wouldn't have a point to make?

Oh, come now. Let's be civil, even if you think I'm wrong. ;] If I didn't have a point to make, I wouldn't be posting.

I eventually downloaded and played on all versions of Better AI except for the most recent public release (which was 2/12, I think). I will admit, the problem with my example was extreme, and it was eventually handled a lot better. But really, that's beside the point. I wouldn't have chosen that example if it didn't indicate--to me--a potential problem with Better AI's design philosophy.

As a human player, I'm interested in playing the game to have fun. In most cases, since I'm a competitive player, I have the most fun when I overcome a significant challenge against a difficult opponent. For this reason, increasing my fun and improving the AI were often the same thing.

However, there is a point when fun and AI "strategy" diverge. At the time of its implementation, Blake's new defense logic was an example of that. When I found that it was busted by the AI's advantages, I went into the XML and reduced the AI bonuses to upkeep and unit upgrading, and that did a lot to solve my complaints. However, it didn't solve the fundamental conflict I had with Blake's design philosophy--he was changing AI behavior, not because he wanted to make the AI a better player, but because he felt that human military investments paid off too easily. And to me, that was a low-fun direction.

Now, you might say: "What's the difference, Zooloo? The AI is a better player if it's harder for the human to defeat it. So wouldn't more defenders be more fun for you?" And I answer, sometimes, yes, it can be more fun. But sometimes it can't. It's a matter of implementation.

Blake's earliest Better AI versions made the AI a tougher military target because the AI was playing an overall better game. But the build I described made the AI worse, in some ways, by forcing it to adopt a highly defensive strategy as a rule. Yes, it was harder for humans to win wars after that update, but it was the result of a tedious and unsatisfying change. By increasing the number of defenders in all cities, it felt as if the AI and I were simply forced to spend more and more money on our military forces, or we wouldn't win any territory. It was an arms race which simply slowed down every other aspect of the game.

In short, Better AI was sometimes designed to stop popular human strategies. And that, to me, is a design philosophy that lends itself to creating less fun gameplay.

If you like, I can dig around in old threads and find some quotes that verify that design philosophy. But I hope you'll take my word, not as someone who thinks Blake is a bad designer (because honestly, his AI logic is very subtle and smart), but as someone who thinks Blake's design aesthetic was sometimes antithetical to player fun.

 

[Quagga]

If you like, I can dig around in old threads and find some quotes that verify that design philosophy. But I hope you'll take my word, not as someone who thinks Blake is a bad designer (because honestly, his AI logic is very subtle and smart), but as someone who thinks Blake's design aesthetic was sometimes antithetical to player fun.

The threads are there for anyone who cares to read them. Anyone who does will realize that the Better AI experience was a development effort carried out with the CFC public as playtesters. Yes, there were things that worked and things that didn't. There were many debates about what was fun and what wasn't, including whether the AI should build up a large military or be a cream puff. A lot of it was learning by doing. My point is that Better AI was and is an unfinished product.

Furthermore, Blake wasn't working for Firaxis when he worked on it. He didn't have their resources nor their guidance. He didn't have to meet their standards for playability. Because of all that, I predict that the BtS AI will be improved over Better AI and that the concerns raised in this thread will not be the concerns of the vast majority of BtS owners.

 

[Roland Johansen]

You said it yourself: the options to win become less. That isn't fun. Many people already say Civ4 is too one-dimensional, with early rushes and constant war more often than not being the dominant strategy. In the above situation, what smart move could you do? In that case, the "right" strategy was to kill the GM and go to war *again.* Wow, how exciting . Given equivalent difficulty, if it came down to crazy options (bribing Shaka into pacifism) versus realism (it'd be impossible to bribe Shaka into pacifism, but his army is 30% less), I'd probably prefer the former.

I dunno, I really think the whole "dumb AI with a ton of bonuses" versus "smart AI with few bonuses" is really a wash. Does it really matter if the enemy has 12 longbows (dumb AI) or 10 longbows and 3 catapults (smart AI)? Not really. It's just an extra move to split up the stack or wait a turn to heal. Big whoop. Wars are won with strategy, not tactics (especially Civ4 ones).

Ok, it is clear that we have different preferences. I don't like strategies based on inferior predictable moves from the AI. I would call such strategies exploits of the predictable inferior thinking process of the AI and would therefore like the AI to behave smarter so that I can't exploit the AI's stupidity that much. You don't like to lose strategic options caused by the AI not doing inferior moves anymore because you think it makes the game more shallow and reduces the options for victory.

I think we have a very different look on what we expect from the AI opponents and thus we can do nothing else then agree to disagree. Ok?

Oh, come now. Let's be civil, even if you think I'm wrong. ;] If I didn't have a point to make, I wouldn't be posting.

You were misrepresenting the BetterAI mod by only showing the shortcomings of an attempt to improve the defensive capabilities of the AI and not mentioning the later improvements on those shortcomings. People reading this thread who didn't participate in the beta testing of the BetterAI mod might think that what you were describing was the intended final version of the BetterAI mod, which is clearly not true. I used irony to make that very clear.

If you were insulted by my post, then I am truly sorry. That was never my intention.

I eventually downloaded and played on all versions of Better AI except for the most recent public release (which was 2/12, I think). I will admit, the problem with my example was extreme, and it was eventually handled a lot better. But really, that's beside the point. I wouldn't have chosen that example if it didn't indicate--to me--a potential problem with Better AI's design philosophy.

As a human player, I'm interested in playing the game to have fun. In most cases, since I'm a competitive player, I have the most fun when I overcome a significant challenge against a difficult opponent. For this reason, increasing my fun and improving the AI were often the same thing.

However, there is a point when fun and AI "strategy" diverge. At the time of its implementation, Blake's new defense logic was an example of that. When I found that it was busted by the AI's advantages, I went into the XML and reduced the AI bonuses to upkeep and unit upgrading, and that did a lot to solve my complaints. However, it didn't solve the fundamental conflict I had with Blake's design philosophy--he was changing AI behavior, not because he wanted to make the AI a better player, but because he felt that human military investments paid off too easily. And to me, that was a low-fun direction.

Now, you might say: "What's the difference, Zooloo? The AI is a better player if it's harder for the human to defeat it. So wouldn't more defenders be more fun for you?" And I answer, sometimes, yes, it can be more fun. But sometimes it can't. It's a matter of implementation.

Blake's earliest Better AI versions made the AI a tougher military target because the AI was playing an overall better game. But the build I described made the AI worse, in some ways, by forcing it to adopt a highly defensive strategy as a rule. Yes, it was harder for humans to win wars after that update, but it was the result of a tedious and unsatisfying change. By increasing the number of defenders in all cities, it felt as if the AI and I were simply forced to spend more and more money on our military forces, or we wouldn't win any territory. It was an arms race which simply slowed down every other aspect of the game.

In short, Better AI was sometimes designed to stop popular human strategies. And that, to me, is a design philosophy that lends itself to creating less fun gameplay.

If you like, I can dig around in old threads and find some quotes that verify that design philosophy. But I hope you'll take my word, not as someone who thinks Blake is a bad designer (because honestly, his AI logic is very subtle and smart), but as someone who thinks Blake's design aesthetic was sometimes antithetical to player fun.

I also was a part of the beta testing of the BetterAI mod, so I know those discussions and participated in them. First Blake improved the economical aspects of the AI, some of these were implemented in Warlords 2.08 and some were added to the BetterAI mod at a later stage. The reason Blake did this first was because it was the easiest (his words). A lot of beta testers at that point were complaining that the only (or best) way to win against the AI was by using military force because the AI just wasn't very good at that. So Blake (and Iustus who had joined the BetterAI team around that time) started the hard job to improve the military AI. He improved its offensive capabilities by letting it create large offensive stacks with many catapults/trebuchets/cannons/artillery and he improved the defensive AI by letting the AI react to the buildup of a military. Because the AI is very passive in Warlords 2.08, he also increased the aggressiveness of the AI.
At that point, it became obvious that the AI was benefiting too much from the military handicap bonuses it gets in Warlords 2.08 and a new handicap file was created. In the last month, further refinements were made to the military AI and the AI switched repeatedly between too aggressive and too passive and building too many defenders and too few defenders. And at that point, Blake was contacted by Firaxis and was suddenly gone.

The military AI is a work in progress and since no version of Civilization has ever produced a remotely decent military AI, it would be a great achievement if this were possible.
In Warlords 2.08, you build a sizable stack with plenty of trebuchets and some good stack defenders and walk towards the AI city while absorbing some minor attacks which are healed by your healing unit. You bombard, do some collateral damage and take the city with your city attackers. And the AI doesn't really pose a serious threat to your city attack stack while you do this. Rinse and repeat until the AI empire is conquered.
A good military defensive AI would not defend its cities with 12 longbowmen and have huge handicap bonuses on upkeep and upgrading. It would also not defend its cities with 6 longbowmen and 3 catapults and have moderate handicap bonuses on upkeep and upgrading. No, it should defend with 3 longbowmen and 2 catapults and have some reserve troops in other cities which rush to the defense of the threatened city. It would hammer your offensive stack with 5 catapults from this city and neighbouring cities and then use knights and macemen from neighbouring cities to finish your stack. That's what a good AI would do and then you will start thinking about real strategies like diversionary attacks, feints to lead his forces into a trap or attacking from an angle that is unexpected. This AI wouldn't need massive troops to defend its territory and thus wouldn't need the AI handicap reductions on upkeep and upgrading.
And I really do think that such a military AI would improve the gameplay immensely and add to the strategic elements of the game. Heck, it would be the first time that real strategy was needed to beat the military AI.

Will Blake have created the good AI that can fight as well as I described above? Now, to be fair, I doubt it. That would be a monumental achievement. But I do think he will have come closer to that goal.

 

[sunrise089]

Ok, first of all I admit I have not read all of the posts in this thread thoroughly. I have however read enough that were critical of Mutineer (and perhaps alarmingly critical of even the right of a non-English native speaker to offer up a criticism of someone close to Firaxis) to at least jump in with some comments.

I also echo Sooooo above and agree that we must wait for the game to ship before passing anything resembling final judgment.

That said, some thoughts:

1) Unless Mutineer significantly edited and cleaned up his original post, any native English speaker who read it and was unable to understand his argument needs to reexamine their reading comprehension skills. I've read lots of Muti's posts, and while he is often well outside conventional grammar and spelling norms, I've always had no problem at least understanding what is going on. I certainly wouldn't use an attack on his grammar as my main counterpoint to his argument.

2) Enough with the automatic defense of people due to their perceived higher status. I appreciate all the work Blake has done providing BetterAI to the community. I also appreciate the SG play that Mutineer has posted about. One is not inherently more generous that the other. Blake has chosen to work on BetterAI with his own free will, so he doesn't deserve a pass or immunity to criticism because he happened to have put in a lot of work. He does indeed deserve to be free from personal attacks, as do all of us here, but Mutineer was merely attacking the decisions Blake had made, not the man himself at least as far as I read. Additionally we have absolutely no knowledge to confirm the fact that Blake isn't being paid for BetterAI. In fact his near total disappearance around here seems a strong indication he is in fact being paid by Firaxis - and I'm not saying he doesn't deserve it, but lets withhold the charity angle.

3) One of Mutineer's central assertions is that BetterAI makes the game unbalanced and unfair (as well as less open-ended) by letting the BetterAI exploit inherent AI advantages. One of the claims offered by BetterAIs supporters has been that the AI in fact is designed to allow the computer players to no longer need significant bonuses. Indeed, if this is the direction taken by BTS, I think we can all agree it will be a positive advancement. That said, the BetterAI as it is available now most certainly does not act as though the AI bonuses are irrelevant. I would call your attention to Acid_04: The final frontier, a SG with Muti and Blake as participants. Clearly we see the AI exploiting it's ability to upgrade troops for nearly no cost. Clearly we see instances of the AI outplaying the human team not with any better strategy or intelligence in the classical sense, but with better use of its bonuses.

Basically the way I see it there are three components or levels to BetterAI.

The first level fixes things that the conventional Soren-built AI does that are stupid. Founding cities in clearly the wrong locations, building the wrong tile improvements, working the wrong tiles. Almost everyone can agree these are awesome fixes and something that should be included in the game. They only marginally affect difficulty however, and seem to be able to coexist with traditional gameplay across difficulty levels.

The second level affects the way the AI makes use of the rules of the game - things like whipping its cities more often or building more or fewer units, or larger or smaller stacks. Whether or not these changes make the AI smarter or simply more aware of it's bonuses is difficult to debate, since there is currently no way I'm aware to divorce the two. This aspect of control however is where Muti makes a very valid point. Anyone suggesting that BetterAI is designed to effectively make the AI exist without bonuses needs to be able to explain the effect these changes have on the current difficulty level. In short, they make the game harder. In fact they make the game harder on each difficulty level, which is potentially a huge problem. A noble AI will still whip out an extra archer to defend a city. A Warlord AI will still build a different mix of units in a stack. The difference in difficulty levels is still maintained through the ever-present AI bonuses. At present to me it seems difficult to say whether there is any mechanism to allow these changes to BetterAI to have any real meaning without the existing bonuses - since taking the bonuses away would seem to have the effect of severely diminishing the differences between the difficulty levels. The only way in which BetterAI could function alone would be....

The third level of (potential) BetterAI changes would involve the AI making smarter and smarter decisions at each difficulty level, so that no bonuses were needed. On Settler the AI would make many mistakes and nearly all suboptimal moves, on Prince the AI would play a fairly solid game, and on Deity the AI would make the right decisions all of the time as well as make surprising the seemingly inspired moves to keep the player constantly off balance. In other worlds the SP experience would come to mimic playing beginner, average, and world-class MP players. Perhaps this is what Blake will bring to BTS, but right now I see no indication of this in the current Warlords+BetterAI game. Instead, I see smarter play tied to existing bonuses or handicaps at all difficulty levels. In order for many posters in this thread's opinions to be correct, and Mutineer to be totally wrong, the AI will need to make tremendous progress in this area, which we all can agree will be very very difficult. Until we know things are headed conclusively towards this development, automatic criticism of Mutineer (who has played the BetterAI more or less successfully at the highest levels) is at best premature and at worst blindly dogmatic.

 

[PieceOfMind]

First, I apologise if some of what I'm about to say is reiterating what others have said (jkp1187, Quagga and Roland) but I felt I needed to chime in.

I think many people forget betterAI was only ever a work in progress. Sure some will enjoy the last version like I do, but there are definitely problems with it. But to say that these problems are because of Blake's incapacity or inability to understand the core game mechanics is ridiculous. The biggest problems were simply because it was unfinished.
I don't pretend to understand how difficult it is to make a mod, especially one involving AI changes, but I'm sure a pretty solid understanding of game mechanics etc. would be necessary. Besides, I reckon Blake would have felt like he let people down when he stopped working on betterAI. The loss of contact with Blake (presumably due to some confidentiality contract) was an unfortunate circumstance of him joining the BtS team, but I believe it is unfair, or perhaps just misguided, to criticize his product (betterAI) when he himself was not yet satisfied with it.

@Zoolooman, commenting on one of the intermediate versions of betterAI is fairly pointless. The 52 skirmisher issue was fixed AFAIK. It's unfortunate you haven't played the latest build - that's the only one I use now.

-------------

Of course the AI in BtS is not going to be the last version of the betterAI mod. I presume Blake and others have had several months to program it, and the testers are paid professionals this time (I don't mean to insult the CFC testers of betterAI). This time the AI will actually be finished ie. more polished, unlike betterAI, and if there are serious or obvious problems they'll be fixed with the (inevitable) patch.

I find it ironic, personally, that some are so worried about what effect Blake's AI changes will have on BtS, while my decision to buy BtS will probably be largely due to the fact Blake is now part of the development team.

 

[ThERat]

having been part of that deity acid04 SG, I want to make a few comments.

1. Why do some CIV fans condemn anyone who dares to question the game? Are you guys so insecure about the game? Mutineer's English might not be the best (and sometimes it is a tad difficult to grasp), but that does not warrant personal insults.

2. Muti is a very good player and his ideas and comments were always valuable in the SG's I played with him. In fact during the acid games I learned a lot about CIV. So much so, that I found the game utterly static and unfun. Human's are forced into a 'preset' pattern of play. Game mechanics nerf any sort of 'smart' strategy, Blake's unfinished AI made that even worse. I don't find it fun to be forced to play that way. It might be better for overall balance, but I damn balance, if I can't have fun.

3. We don't know the status of Blake's AI for BTS. We can only hope that there are testers good enough to really test it. And that is something I doubt simply due to past experiences. I am sure Firaxis is very busy churning out a lot of goodies etc. , less playtesting though.
Just look at what happened to vanilla and warlords. Both stopped at certain patches and nobody there bothered to maybe give us proper patches. Firaxis needs the money from BTS. It's all about business, isn't it?

 

[PieceOfMind]

Another point, partly in response to sunrise089.

The AI doesn't know anything technically (with no intention to insult, but rather so I can explain my point). It doesn't know how to exploit its bonuses. If it appears to do that, then it's a consequence of the changes that otherwise make the AI more competitive. For example, suppose you tell an AI that every city should be defended with no less than 5 units, and those units get built very quickly due to its handicaps. This is not the AI exploiting its advantage. It's simply a handicap giving the wrong sort of advantage to the AI. Would you suggest the AI somehow do something deliberately to make itself less efficient while building those units, just so it doesn't appear to be using its handicap? I doubt it. However, I completely agree that it is not fun to see an AI "appear" to exploit its handicaps so blatantly. The classic example is the ridiculously cheap upgrades. I don't think many people would argue for the AI "exploiting" its bonuses at all. But as long as there are handicaps at all, there'll be this problem to some degree. And clearly it bothers some more than others.

By the way, sunrise089, I think your argument is very sound and for the most part I agree with you. The obvious exception being your different interpretation of what I described above.

And certainly, Mutineer's command of English is irrelevant to the validity and strength of his arguments. I think the contentious discussion he has brought about is very valuable.

 

[Mutineer]

Well, lets discuss why we wont AI to improve and how define AI to be improved.

Just for example, lets look on city grow, slavery and draft.
How AI work with it now?

Well, AI very rarelly used slavery and I do not know does it used draft at all or not. AI is not veyr good in deciding which ties to work and/or improve.
Insted he has follow bonuses:

1) Bonus to population grow. Most people are not aware about it, but this bonus is huge and let AI grow coties to max happiness very fast.
2) Bonus to happiness. On levels > noble AI has more happiness then HUman, in addition AI practically does not suffer from WW.

As result AI can produce more with out using slavery and draft. Good use of slavery and draft just let human to compencate for AI bonuses.

Now, what happened if city governor become better in assigning ties, improving them, better in building improvements?

AI cities start to grow even faster and bigger. Now to compensate we reduce grow bonus a bit.

Question, did AI become better?
Answer, very little. Players mostly will not notice diference. Most important, it would not open new options for a player. Simply because it does not matter to player why AI cities grow that fast and that big. What is matter is: Did it open a new way to player to effect grow? Did it give player a new options to effect AI?

Answer is: not at all. AI will still grow to max happiness or food limit very fast, making pillage farm to have the same effect on AI production = very little. Untill you make city starving pillage fars does not effect AI production at all. You really need to pillage a lot of improvements befor city will be visible effected. WW still would not effect cities. Player still restricted to the same tactics.
in early middle game player city can be crippled production wize by pillaging 2 food resources. Player get it's production from small cities, limited by low happiness and high War wareness. AI useally does not even work food resources at that moment. It most lickly run uptimise shields and one does not need that big food excess for that. pilaging couple food resource will low AI production very little.

A lot of thinks is civ are interlinked, it is dificult to decribe. Mach of AI behavious just emulate pressure and behaviour human player has.

AI programmed to ask for peace after some time. This emulate human player need for peace because WW killing him. One of "Better AI" "improwements" were to let AI to contimue war longer. But AI has a lot of bonuses to WW, so it has no natural mechanism that could force it to seek peace.

I can continue to discuss AI behavior as attempt to do it like human player why do not have problems human player forced on him for ever.

My main point is, I am trying to define, what is a better AI.

AI is better if it give players more options and they preferably are the same as when he is playing competent human opponent.
If AI reducing player options, by hiding money, producing so many units that player simple can not hope to compeate with AI maintance bonuses, et.
AI is actially worse, even if it beat human player more easy and with less bonuses.

Lets see what happened if we teach AI to use slavery, for example. WE open a whole can of worms.
First, slavery multiply AI bonuses and AI city grow bonuses are huge. I do not remember execly, but they are easy hit 50%.
then there are AI production bonuses, and whith slavery they now multiply.

so, if human can get 2 shields/food, AI get 2* 2*1.2 ~5 shields/food on some middle monarch/emperor level. OK we should reduce grow bonuses. but thinks do not finish here.
Now AI can react on invasion the same way as competent human player would. It can whipe defence/turn untill his city is size 1 or he produced enoght defenders. So, now human has to bring
useal amount of units + 2* number of units AI will whipe + 2* number of units AI can whipe in near by cities.
So, should we reduce player upkeep cost so he can have that many units? and then we will run in other problems, like is it fun?
Is now a decisive war posible at all? Is it fun when borders do not move most of the game?


There is other aspect fo this: How mach AI should emulate human player from point of view of game enjoyment?
When playing human player task often is not to beat player and take all his cities. That tend to be tooo expencive.
I single players war are profitable, in multiplayer they are not. In multiplayer again competent opponent war is a way to slow him down, not to profit from it.
Point is generally to make other players to see that he is lost, that he only wasting time now. AI does not care about wasting time.


I want to add, some people put strange questions, like if I propouse to keep old AI, stale development. I AM NOT.
I am only trying to discuss what is a meaning of "BETTER AI", what we wont from it.

And yes, I do want to point out that in my expirience Blake did not consider this questions when he work on it as an open project.
Attempt to raise this questions in his development treat meat with very defencive and agressive responce, mach like many posts there.

 

[Jet]

AI is better if it give players more options and they preferably are the same as when he is playing competent human opponent.

Are you saying that when someone plays multiplayer against competent opponents, he has more options than in SP?

If so, can you give a couple examples?

 

[Sansevero]

I understand everything you wrote. You clearly know the game very well, and your concerns seem valid to me. But Blake has professionals on his side now, and much more time and resources. The AI may well fall short of your expectations, but I don't expect it to be worse than 2.08 or Blake's last build.