Strength of AI vs. Number of Civs

[Telamon]

I have heard several people mention that they would rather have 8 civs with a good AI than 32 civs with a bad AI. The number of civs in a game is personal preferance. I myself would prefer to play against 32. However, the precieved tradeoff in AI quality and number of civs is nonexistant. Anyone that thinks the number of civs has any bearing on how well they play doesn't know what they are talking about. All civs will be using the same game logic. Playing with 32 civs will definately take more time than with 8, but the quality of their play will be the same. Activision just didn't feel like giving the player the option of playing with greater than 8 civs. If the game was coded well, this could be changed very easily. I do not understand why CTP2 will not have an in-game menu to set the number of civs to at least 32, if not 64. If they do not allow play with greater numbers of civs by text-editting, we know they reserved this functionality to make more money on a patch. I am a programmer myself and I will tell you that this is a very easy thing to do.

As a side note, I hear that CTP2's AI is much improved. I certainly hope so. The critical question is: Doe the AI cheat? It has in all past civ games and that the AI had an advantage always bothered me. If Activision can create a computer player that does well and doesn't cheat than that will be a major accomplishment. If they "improved" the AI by allowing it to cheat more, well, I could have done that myself.

you would be right, if the ai had unlimited time
but it doesnt. in ctp1(and i guess in ctp2 too) there is a time limit for each ai to complete his "thinking" and for all the ai's together.
if you add more civs, without increasing these limits, you are practically sharing the same time to more ai's....

[Ralf]

quote: Originally posted by Telamon on 11-02-2000 05:55 PM However, the precieved tradeoff in AI quality and number of civs is nonexistant. Anyone that thinks the number of civs has any bearing on how well they play doesn't know what they are talking about. All civs will be using the same game logic. Playing with 32 civs will definately take more time than with 8, but the quality of their play will be the same.

I agree with MarkG:s objection, but i would add the following:

The argument that "all civs using the same logic" (more or less) is irrelevant. The problem is theres so many totally unknown and unforseeable factors in games like Civ/Ctp, from an AI programmings point of view. Adding more civs, more units, more diplomacy-interconnections on much bigger maps would quickly add salt to injury.

I suggest you read the following thread on the issue:
http://apolyton.net/forums/Forum6/HTML/001531.html?39

Im not sure if you really are that interested in AI-problematics, Telamon. I dont believe that players and developers who is really interested in these AI-problems, so generally and superficially can totally dismiss the objections of increased bugs-, programmings- and hardware-related problems, like you do.

[This message has been edited by Ralf (edited November 03, 2000).]

[Big Dave]

This really has a simple solution, Markos. On the little tag with the system requirements you state "System requirements for playing with 8 Civilizations"

Unfortunatly, not simple enough for a marketting weenie to figure out.

quote: Originally posted by MarkG on 11-02-2000 07:25 PM you would be right, if the ai had unlimited time but it doesnt. in ctp1(and i guess in ctp2 too) there is a time limit for each ai to complete his "thinking" and for all the ai's together. if you add more civs, without increasing these limits, you are practically sharing the same time to more ai's....

------------------
Big Dave

Fool me once, shame on you.
Fool me twice, shame on me!

[Monkey]

Ok time to put my computer science knowledge to practice here.

Because the diplomacy of CTP2 has been enchanced, each civ might have to take into account the current state of every other civ in the playing field. Clearly this problem has the complexity of O(N^2) meaning the amount of total processing time is proportional to the squared of the number of civs present in the game. Anyone who knows anything about programming will know that as N increases the amount of processing time will be increasing at a much faster rate. So for a 32 civ game to happen we could be looking at the kind of computational power that current generations of computers just dont have. So for anyone saying it is just a simple #define problem... errr.. nuff said.

[Daniel Frappier]

All that talked about CPU requirement is irrelevant.

The fact is that giving the opportunity to someone to play with 32 Civ will not make the game slower for others who prefer to play with 8 Civ.

I have played CTP I with 16 Civ on my old P166 and was very happy about it. Not everibody gets bored because the AI takes more than 5 seconds in between turns.

Beside i got to find something to do next year with my new 1.5Gh Sledghammer (new AMD 64bit much faster than Athlon processor)

quote: Originally posted by Big Dave on 11-04-2000 01:55 PM This really has a simple solution, Markos. On the little tag with the system requirements you state "System requirements for playing with 8 Civilizations" Unfortunatly, not simple enough for a marketting weenie to figure out.

somehow i tend to think that there would be players and/or reviewers who would consider it very bad that the game doesnt not play well with 16 civs on the minimum requirement, whatever disclaimer there was....

[Pyaray]

quote: Originally posted by Monkey on 11-04-2000 02:04 PM Ok time to put my computer science knowledge to practice here. Because the diplomacy of CTP2 has been enchanced, each civ might have to take into account the current state of every other civ in the playing field. Clearly this problem has the complexity of O(N^2) meaning the amount of total processing time is proportional to the squared of the number of civs present in the game. Anyone who knows anything about programming will know that as N increases the amount of processing time will be increasing at a much faster rate. So for a 32 civ game to happen we could be looking at the kind of computational power that current generations of computers just dont have. So for anyone saying it is just a simple #define problem... errr.. nuff said.

You stick with that CS thing, cause you nailed it. Also, it's really not cool to make a game where the minimum spec system can't play the "full game". And I'll add on to it what you said to explain the problem in even further detail. There is a concern with the AI comparing the other civs, and that is exponential. With 8 civs, each AI has to evaluate 7 other civs, so it's 7^2=49, with 9 civs it's 8^2=64, and 10 civs it's 9^2=81. So you can see, by increasing the number of civs by 2 we have nearly doubled the amount of work that needs to be done. Just for the record, 32 civs would be 961 comparison's each turn. Almost 20 times that of 8 civs.

Now consider units, and how the AI's look at each unit they can see. Let's assume later in the game where you can see nearly the whole map (if not the whole thing). And let's say each civ has 100 units. In this case, the AI must evaluate it's own units as well. So for each AI player they have to evaluate 8*100=800 units. Now do that for each AI and we have 6400 units being evaluated for a complete turn. Let's increase that to 10 civs, 10*100=1000, for each AI is 10,000 units to be evaluated each turn. Once again, nearly doubling for only 2 more civs.

I do understand your points of "well if we want to go ahead and let it take a long time, then we should have that choice". But, we can't release a game that we haven't tested everything in. If we were to increase the number of civs by 2, it would nearly double our testing time, and that's something we can't afford to do. There are already a ton of people (whom I fundamentally disagree with) that say we don't test enough. This would just add to the problem. Would you all really want us to add a completely untested feature?

On the other hand, that setting is still stored in an TXT file. But if you're looking for some kind of official sanction that we support that, you're not going to get it. We didn't test the game for every possible combination of things that users can set by modifying the settings files. Nor would we, or could we, there are just too many variables.

/soapbox

Pyaray

P.S. I wasn't really going to address this anymore, but this does appear to be one of the topics that just won't die. I hope this is an acceptable answer to everyone as to why we didn't do it.

thanks for the thorough explanation Pyaray. we're not little kids playing tomb raider, we are civers and we like these kind of responses

[Pyaray]

quote: Originally posted by MarkG on 11-04-2000 07:37 PM thanks for the thorough explanation Pyaray. we're not little kids playing tomb raider, we are civers and we like these kind of responses

I realize that, I just don't like to write them.

The problem with giving really long technical responses like this is that they are most definately incomplete, or inaccurate for the sake of simplicity. What will then generally happen is someone with technical knowledge (like a programmer or software engineer) comes along and starts to shoot holes in it, claiming that we're lying about this stuff. We're not "lying" as it were, but I did leave out a lot of the finer details due to brevity, trying to make the average consumer understand, and not saying what we consider "internal information".

But we shall see what happens with this response. Perhaps if it's not riddled with bullets after a few days I'll be more relaxed about giving these kinds of responses in the future. No guarantees one way or the other though.

Pyaray

[Monkey]

Thanks for the clear up Pyaray. I figured it would be something like that. What amazes me about some people is that they start shooting off their mouths like they know everything about the game and even propose technical solutions to fix it... all this without seeing or having touched the source code! Amazing! Then there are the people who complain that this feature is not in, that feature is not in... I'm boycotting the game!!!! Sigh... it's only a game, if you don't like it, don't buy it. US$40 is not a lot of money for what you are getting, plus if you are worried about buying a game that you don't like, wait for the reviews or buy it from a place you can return to.

The thing to realize here is that Activision cannot please everyone and nor should they even try to. By trying to please everyone you end up pleasing noone. I remember reading posts in the AOK Heaven forum about how AOK sux because it is missing such and such feature and Sandyman(an ES representative) would just say "if you don't like it then don't buy it"

If you do happen to buy it and was displeased but there was nothing you could do about it then just move on. No need to start a crusade discouraging people to buy it, just politely offer your views and people can make that choice for themselves. For everyone that disliked CTP, there must be at least one that liked it.

Everyone just lighten up, the game hasn't been released yet. Let the game come out first and we'll make the decision on the final verdict for CTP2. Sorry for the rant, but I had to get it off my chest.

[Daniel Frappier]

Thank you for the answer Pyaray.

Scroll bars on the diplomacy screen would have been nice since the game can run and will run with more than 8 Civ on many computers.

Pyaray & Monkey

Thank you for the information on how the computer works. I'm not a programer, so your posts help me understand what is happening a lot better.

Now the real question, when will Activision release it so I can start playing it.

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

[meriadoc]

Thank you for the explanation Pyaray. I myself am content with 8 civs in the game, but explanations like this are, in my opinion, always worth reading since they give a person a better understanding of how the game works. It is this understanding that allows us to present informed opinions. Unfortunately, some people aren't as well informed as others.

------------------
The Electronic Hobbit

[Lancer]

Hmmm, that was a great answer, but it still leaves me wanting more Civs, just very willing to pay for them. I wonder if you could release an add on...2 more civs kinda thing. That way, though it takes twice the time for the comp whiz's at Activision, we as gamers get what we want while Activision gets to expand it's product line without having to vastly alter the CTP system, yes? Then perhaps there could be a chart describing common computer systems and how many Civs can be run on each one. For example, if I have a Pentium 200, I get to run 9 Civs.

Anyway, regarding an add on, what might it cost per additional Civ?

W/ the altered diplomacy I wonder if going into the program and changing the # of Civs will still work, and still allow the great new diplomacy aspects.

[Telamon]

Yes, I recognize that more civs means computational time to complete turns. However, I think that many people out there with Ghz monsters would like the option, at least, of playing against a number of civs that would make a PII cough up blood.

In response to Ralf, the fact is that each computer civ will use the same game logic (code) as they grow and expand. Of course, the values of civ-dependant variables will cause them to function differently. I am not neglecting hardware problems, ect. I am stating that in a well crafted game, one should be able to change #define NUM_CIVS 8 to #define NUM_CIVS 32 and, assuming an adequate amount of resources is available, go on to start up a game with an arbitary number of civs.

[Mike the Nuke]

Mr Paraya (excuss my spelling please)


Thanks for the info on increaing the number of civilizations above 8. I, personally, never utilized more than 5-6 in recent games as the map became too crowded and all you did was fight one civ or another. Your explanation makes sense. Yes, games need to keep smaller systems in mind. I wouldn't want to buy a $2000 system just to play a game. On the other hand, how about some code that scans the system and allows only what the systme can handle. Those with smaller systems could then only ustilize (with out special mods) 7-8 civs. Those larger systems could then handle larger civilizations.

My greatest concern is the lack of the Hot Seat function. Without it the game is useless to me, my family and friends. We thoroughly enjoyed CTP I once the Hot Seat patch (1.2?) came out. The Hot seat option is available in other Civ type games. Why not your highly anticipated new CTPII? You can never satisfy everyone with new gadgets and widgits, but this one already existed. At least keep functions you already had or at least guarentee your continued efforts to fix whatever bugs are in, of which in over 100 Hot Seat games in CTP I, I never found. If the game has Hot SEat, I'll buy, If doesn't I will not touch it. I already have enough games to play with by myself. Skip the multiplayer over internet garbage as a way to play other human opponets. Those sites only thrive for a year or so, then die off slowly (in some cases) or quickly (in others) as the game grows older or new games come out. I have plenty of muliplayer games (over the internet) that cannot be played that way as there are no longer sites for them. Unlike some 'fly by night, short attention span gamers' I'll play a game for years, if it's good. I still play CIV I and II and even AOE I (even though it has unit limitations I hate).

thanks

P.S. don't be afraid to get technical, if necessary to explane a situation. And don't throw in the towel on ideas because they seem impossible. Often the most impossible things become the easiest to solve once the fog of urgency (ie your 'bean counters' pressuring to get the game out before Christmans) is removed. I thought CTP I was a great game, even with some unit ratings I did not agree with.

thanks

[Ralf]

I have always had the opinion that 6-8 simultaneously playing AI-civs is as much as any AI/hardware can handle - at least if one priorities a reasonable AI-quality out of the equation. To nevertheless increase the number of AI-civs limit somewhat, below could act as a nice workaround:

6-8 fully fledged independent AI-civs + another 6-8 neutral, more passive max 5 city miniatyre AI-civs.

The latter can neither build wonders, conquer cities nor have diplomatic relations between themselfes - only with the human and the fullblood AI-civs. Also they share the same AI-logistics between them (altrough different from game to game), yet acts like independant (but less active) civs. These miniature AI-civs only make sense however, if anti-BAB (bigger always better) is implemented intelligently, giving small empires an advantage that bigger empires dont have (for example; significantly less happiness problems building huge cities).

If its worth it in the end? I dont know - perhaps not.

Theres so much that talks against the 32+ (some even seriously discussing 100+) civs, both from a AI-programming, system resources, map-size, turn-waiting, minimum system req and plain game-structural common sense. I want AI-quality before quantity - and adding 32+ AI-civs definitly only going to give me quantity.

Finally, someone perhaps says that "adding more Civs puts stress on the game resources much the same like PBEM/Hotseat would" (jbs).

Thats not correct. In PBEM/Hotseat/Internet games the human players does all the initiatives and decision-making, relieving the AI from that othervise strenuous duty. All the AI has to do is to act as the go-between messenger boy, carrying out the orders.

As long as only human players does the decision-makings however, the number of player could be *much* higher then 8 (if one dont mind some awfull lot of waiting, that is).
Still, from Activisions point of view, it is perhaps NOT commercially viable to invest energy in trying to develop software that can handle 8-32 multiplay. After all: how many of the casual average strategy-gamers out there, is likely to participate in upto 32 multiplayer sessions?

Perhaps PBEM/Hotseat is hard to implement because of some other technical reasons - but thats another issue. In many ways Play by e-mail and Hotseat is ideally suited for turnbased timeconsuming strategy-gaming. It works much better then Internet multiplayer, from the players point of view.
I hope they get around to add upto 8 player PBEM/Hotseat support further down the road, however. Either by a downloadable patch, or together with an commercial add-on scenario-package.

[This message has been edited by Ralf (edited November 05, 2000).]

[rremus]

As wise man say, software programming is all about managing coplexity. Not enough CPU time to compute actions for 100k units? But not each unit need to make a 'decision' individually! Organize units in armies and make the AI 'think' for the whole army rather than each individual unit in it.
Also, not enough CPU time? There's plenty in the human turn! Just use the idle time to plan possible actions and then use the computer turn time only to choose the best pre-computed action in the context of whatever the human did!

[MidKnight Lament]

I don't want anyone to take this too personally, but I find it constantly surprising how people who know so (relatively) little about the game tell the programmers (who immerse themselves in the code at least 5 days a week) what they should be doing. Suggestions are fine, but don't assume to know the answer to everything! Actigrammers are probably pretty cluey people. (I've had no reason to think otherwise thus far.)

A little bit of trust is required. I know some people will wince at that, but there it is.

------------------
- MKL
"And of course Henry The Horse dances the waltz!"
Shameless Plug: http://www.poetic-license.org .............All welcome.

[Grumbold]

The (semi-)technical explanation was very welcome and I am glad to see the unsupported setting to go beyond 8 remains available in CTP2. A soft limit of 8 is quite understandable given the constraints of practical testing limits and tight deadlines. I only wish more game designers would make their code this flexible.

[MadWoodster]

quote: Originally posted by rremus on 11-06-2000 03:20 AM As wise man say, software programming is all about managing coplexity. Not enough CPU time to compute actions for 100k units? But not each unit need to make a 'decision' individually! Organize units in armies and make the AI 'think' for the whole army rather than each individual unit in it. Also, not enough CPU time? There's plenty in the human turn! Just use the idle time to plan possible actions and then use the computer turn time only to choose the best pre-computed action in the context of whatever the human did!

I have to disagree with this. How can you compute actions when the person before you hasn't done their go, you don't know what they are going to do and how it is going to affect you.

[Gedrin]

MadWoodster: This is a fairly simple thing to do in principle really. Any good chess, checkers, othello, etc... does this. The really smart ones also compare their expectations of what the player would do with what they really did and modify their routines for guessing the opponents probable move. However in this case the practical limits of doing this with N civs are reached very quickly.

As for 32 civ support: Well why not simply play on an easier level? I mean if you control 10% of the map would you rather have the other 90% controled by 1 civ or 31? Divide and conquer is a time tested strategy... if you want to do your dividing by setting the max civs to some really high number then ok but it makes it a little easy no? Well at least I've found the 32 civ games I've played to be far easier. The opponents are always too small to mount any sort of defense.

Some may say, "yeah but you make contact sooner and so you have less room to grow." Well so do all your opponents so this is a uniform dissadvantage, not a player one. I'll grant that my scores were lower but the amount by which I was winning was always much much higher... which is why I stick to 8 these days. (Although I do like a large max civs since then the cities that revolt [happens alot with my mods] are not all barbarians.)

Meanwhile, those civs you have no contact with keep each other at a uniformly low power level.

Now it would be nice to have an arbitrarily large number of civs on the science and diplomacy screens and if you have enough hamsters in your processor then power to you. But if you really wanted a more 'difficult' game then start the AI's with more settlers.

Gedrin

[rremus]

quote: Originally posted by MadWoodster on 11-06-2000 08:02 AM I have to disagree with this. How can you compute actions when the person before you hasn't done their go, you don't know what they are going to do and how it is going to affect you.

I also have to disagree with you One easy way is this: Mark the elements that the units took into account when computing its decision (during human player turn). When the computer turn make its decisionn, if any of this elements changed then then decision has to be made again. If not, the decisions already made are fain and thus the process is skipped. Of course it depends on how many elements are took into account when a decision is made. I'm pretty sure the 'decisions' took by the computer for units on a continent on wich the human player does not have any cities/units would be in most of the time valid...

Anyway, the main idea is that I'm convinced that a medium computer of today could handle a large number of civ. I wonder if the player could handle such large number of civs. There is the problem of recognizing units nation (colors will of course be out of question!), remebering at least the dimplomatic status of the civs around you and understanting what is going on from the tens of messages each computer turn would generate (Cubans are building Pyramids, Jamaicans are at war with Romans etc...)!

[Maccabee2]

Thank you for the enlightenment, Pyaray. Now, we understand what motivated Acts judgement call. I may not agree with it, but I understand it.

Just a short thought:

"Also, it's really not cool to make a game where the minimum spec system can't play the "full game". "

While it might not be cool to the marketing gurus, it might be cool to us your customers to have the packaging actually provide useful technical information as to the software's performance at different levels of hardware (read buyer's CPU speed, etc.) Imagine! The software box providing a synopsis of its real performance rated to my CPU instead of my finding it out after I buy it.
We are your free focus group. If I'm in the minority, ignore me. If, however, the majority want more options, and those options are reasonably possible (not easy, but possible) then there's your marketing plan.
Food for thought.

[Pyaray]

quote: Originally posted by Maccabee2 on 11-06-2000 04:36 PM While it might not be cool to the marketing gurus, it might be cool to us your customers to have the packaging actually provide useful technical information as to the software's performance at different levels of hardware (read buyer's CPU speed, etc.) Imagine! The software box providing a synopsis of its real performance rated to my CPU instead of my finding it out after I buy it. We are your free focus group. If I'm in the minority, ignore me. If, however, the majority want more options, and those options are reasonably possible (not easy, but possible) then there's your marketing plan.

Well, I can't say I disagree with you, but I can't agree with you fully either. There is sort of a set standard for how specs are printed on the boxes. You'll notice that pretty much all companies print their specs in the same way ("minimum spec" and "recommended spec". As anoyingly uninformative as this way is, it still confuses a lot of people. Remember, this has to go to the lowest common denominator. But I do agree that this information could be more useful, I'm just not sure how we could accomplish that goal. In the end though, it's not my choice, and I don't have any say over it. And I have serious doubts that the people who do have the choice read fan sites.

Pyaray