A software engineer´s answer to the 32+ civ issue

[Ralf]

Pyaray (a member of the CTP-2 developing team) takes the bull by the horns, by once and for all (hopefully) giving a technical answer to the unrealistic 32+ simultaneously playing AI-civs idea.
http://apolyton.net/forums/Forum35/HTML/000302.html?16

Scroll further down the thread and you can read hes response.

[UltraSonix]

cool link!

------------------
No, in Australia we don't live with kangaroos and koalas in our backyards... Despite any stupid advertisments you may see to the contrary...

[ContradictioN]

Something about it doesn't make sense

You only need to make one AI, which each civ runs off, and during game play, the computer only needs to run the AI for each civs turn, not all the civs at the same time.

[MarkG]

quote: Originally posted by ContradictioN on 11-05-2000 07:27 AM You only need to make one AI, which each civ runs off, and during game play, the computer only needs to run the AI for each civs turn, not all the civs at the same time.

yes, but
1) there are more civs there the AI will run more times (=more time added for each turn)
2) the AI will have to take into account more civs for his decisions, therefore it will need more time (=even more time added for each turn)
3) if you put(and ctp1(at least) had) a time limit for the AI to "think" for a civ it will obviously turn out of time(= weaker AI)

[Ribannah]

quote: Originally posted by Ralf on 11-05-2000 01:22 AM Pyaray (a member of the CTP-2 developing team) takes the bull by the horns, by once and for all (hopefully) giving a technical answer to the unrealistic 32+ simultaneously playing AI-civs idea.

His horn-taking is not very accurate, however (and yes, I've done some software engineering myself).
If the map is of the same size or even say 4x as large, the number of unit interactions will not grow quadractically. There will be more civs, but they will be small. Or a lot will have perished already by the time the entire world is covered.

If we modify map-exchanging to exclude unit locations, growth in this respect will in fact be non-existant since the civs will only know about units in the proximity of their own towns and units.

Diplomacy options may grow quadratically - in the extreme case that all the civs are still in the game and all have contact - but that doesn't take much computer time even for 100 civs.

------------------
If you have no feet, don't walk on fire
[This message has been edited by Ribannah (edited November 05, 2000).]

[Thue]

With a bit of proper caching it shouldn't be a problem for them.

[Thue]

Oh, and did I mention I added support for 30 concurent players in freeciv some time ago?

They don't do much diplomacy, however.
[This message has been edited by Thue (edited November 05, 2000).]

[Ralf]

quote: Originally posted by Dom Pedro II on 11-05-2000 12:48 PM It would be like having little Barbarian civilizations that stop your city progress at every chance. You'd spend so much time beating them back or dealing diplomatically with them that you'd have no time for anything else.

You hit the hammer on the nail there!

"Larger maps!" someone argue. Large enough so that at least 15+ civs can grow their indevidual big and well-developed 20+ city empires plus another 10+ smaller empires. What does that mean put into practice?
Well, lets forget about the (very real) AI-problems for a while, shall we. Lets fool ourselves by pretending that the increased AI-workload is a bypassable problem. Lets instead concentrate on what it would look like for the average casual strategy-gamer (who after all is going to be the bulk-byers from both Firaxis and Activision´s economical point of view):

It means that you could practically play a whole full game without ever get into personal meeting terms with about half of them. Its just too much *work* searching them up! (and too little *fun*).
Sure you can speed up the move-rates of the units in order to compensate for the larger maps. But, would it really be that fun to try to "explore" and then "conquer the world", with perhaps 5-10 times (five to ten times!!!) as many forreign cities to conquer, comparing with late end-games in Civ-2 or SMAC?

What about those who like to take a civilistic science race approach in order to reach Alpha Centauri? Who many Wonders can you hope to collect, if there is 15-20+ well-developed AI-empires each of them trying the exact same thing? Wasnt "wonders-collecting" one of the fun aspects in Civ-2, SMAC and CTP?

What about Diplomatic affairs? Who the heck would in the long run, bother to keep themselfes constantly updated what upto 32 civs had for ever-shifting entangled hard-to-overview foreign affairs status?
Also - Dom pedro II quote: "You'd spend so much time beating them back or dealing diplomatically with them that you'd have no time for anything else". Think about it.


A 32+ civ game (not to mention 100 civs) would grow totally out of proportion. It would just feel "to much to melt". To put the feeling in two words: INFORMATION OVERLOAD! (sorry for the capital letters).

"Well, if thats the way you feel about it - why not just choose fewer AI-civs (about 4-8) at the beginning of the game? You have a free choice, dont you?"

Here i will quote Adm.Naismith´s post in the "Re: Speeding up the game with 100+ Civs" thread. Hope you dont mind Naismith. Good post by the way!

------------------------------------------------------ Quote:

"Too many times I read about debated game features: "make it an option".
Well, often it can't work! Programming life isn't so easy: every time you consider to put anything IN game, you must build a data structure that can manage it, build functions to cope with them, adapt AI to consider that option if ON (and ignore it if OFF!).

"Option" mean the QA staff must test&debug twice (On and Off), but that's not enough, because every option could grow the tree as power of 2!

Raising the max number of Civ is not simply a matter of personal taste. AI tactics must change because game will change when the number of civ per map square change so widely, mid age diplomatic relations will be too much to be managed easily.

To keep game turns short enough with 100 civ (players will complain anyway, if the game will slow down too much at larger setting, we know ), Firaxis must spend most of the optimization about this aspect, avoiding much more important task.

Don't take things so lighty, options is not a magic word to solve problems, more a source of problems itself".

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

[zyxpsilon]

quote: Originally posted by Dom Pedro II on 11-05-2000 12:48 PM Why does anybody want 32 civs anyway? Thats alot. Do you want it for playing the normal game or so you can have more countries when doing a scenario? I know that I sometimes wish I had a ninth or tenth civ when doing a scenario, but 32? 24 civs I can see. But isn't there at least some hope of improving the AI while stretching it out over 24 civs and still come out with a somewhat smarter AI??

Ralf's nailed the key argument....
More "smart" civs is challenging under -conquest- mode; but as a scenario/building tool it is much more appealing to have the diversity available.
As for the AI's, given the proper (say... accurate) behavioral interaction to a player, the only obstacle to a game pace becomes "how well" they expand rather than "how long" they can stay competitive.
It is not -who and how many- the player can defeat... but -how fast!-
Considering, also... the variety of hardwares around.

[Ralf]

quote: Originally posted by ContradictioN on 11-05-2000 07:27 AM Something about it doesn't make sense You only need to make one AI, which each civ runs off, and during game play, the computer only needs to run the AI for each civs turn, not all the civs at the same time.

That would be correct if each of the AI-civs (including the human player) lived in their own solitude glass-jar, and never interacted with each other.

The *real* AI-problems sticks its ugly head up then more and more Civs starts to interact with each other, in one way or the other. In many ways AI-programming is like peeling an onion:

If you are looking close and cutting the onion with the knife, you easily gets tears in your eyes. The guy that stand a couple of feet away, says: Hey, whats the problem? Why are you crying?

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

[MarkG]

quote: Originally posted by Ralf on 11-05-2000 02:26 PM Sure you can speed up the move-rates of the units in order to compensate for the larger maps. But, would it really be that fun to try to "explore" and then "conquer the world", with perhaps 5-10 times (five to ten times!!!) as many forreign cities to conquer, comparing with late end-games in Civ-2 or SMAC?

a solution might be to give the ability for a civ to surrender. this was in smac, with the "be my ally or i'll kill you" proposition, and it saved lots of time spent in boring end-game wars...

it is very certain though, that with more than 8 "legal" civs, the military condition of victory might have to be changed a bit(how realistic is it to conquer all the cities of the world, anyway?)

[MarkG]

quote: Originally posted by Ralf on 11-05-2000 03:46 PM If you are looking close and cutting the onion with the knife, you easily gets tears in your eyes. The guy that stand a couple of feet away, says: Hey, whats the problem? Why are you crying?

you are describing the situation perfectly!!
it is the best though, that you come over and show the guy the onion, if you know what i mean...

[Ralf]

quote: Originally posted by MarkG on 11-05-2000 04:07 PM a solution might be to give the ability for a civ to surrender. this was in smac, with the "be my ally or i'll kill you" proposition, and it saved lots of time spent in boring end-game wars...

Thats a thought! That would of course cut out some of the grueling work of trying to conquer each and every city on that *huge* map.

But then someone comes along an says: Hey, whats not accurate with reality! Look at Nazi-germany! They sure as hell fought to the bitter end! (just kidding!).

Still, we have to consider the wellknown AI-pathfinding problems with AI-units wandering around aimlessly. These problems arent likely to diminsh when using ever more enormous maps. An Ribannah´s solution sure doesnt help:

Quote: "If we modify map-exchanging to exclude unit locations, growth in this respect will in fact be non-existant since the civs will only know about units in the proximity of their own towns and units".

Now, i realize that it is mathematically and practically impossible to take each and every unit-location and city-location (and more) into the calculation - even with Civ-2/SMAC´s limited number of factions/map-sizes. Todays computers and software developing tools are simply too primitive.
But thats why its so ridiculous to even *enlarge* the AI "blind-gap", by making bigger maps, with even more civs and units.

Most AI-related suggestions should go in the direction of *minimize* the "blind-gap", instead. Below is a thread about how to take the AI-city locations, out of the equation - and hopefully making it a little easier for the AI to work more efficiently. http://apolyton.net/forums/Forum6/HTML/001725.html?12

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

[Matthew]

Some of these posts reflect concern that too many civs will make the game harder. This is a non arguement for 2 reasons.

1. Option of fewer civs.

2. In the manual for Civ II it is implied that Diety level is barely beatable. He he he he. The game needs to be more difficult anyway.

About Doability, Well, I'm not a software engineer, but I know that people play with 32 civs in CtP by altering the rules text, although I don't know how well this works. And why does every AI civ need to compare every other civ every turn? If the AI really needs to know how the civs compare, wouldn't once every 5 or 6 turns suffice?

And timewise, what are the minimum specs for Civ II, which was released a few years ago? I think that 4 years ago a 100 MHz pentium was considered pretty fast. And If I'm not mistaken, Civ II can run on a 486?

By the time Civ III is released I imagine upgrading to 500 MHz will be fairly inexpensive, and 1 gHz fairly common. If computers will typically be 10 x as fast, and it takes 50 x as many operations for the AI's to move with 32 civs as opposed to 8, then the AI's should take 5 times as long to move in total, assuming 1 human and 31 AI's. It doesn't seem to take the AI's all that long to move now, so 5 x as long is certainly acceptible. It will still be almost as fast as SMAC. And if someone has a slower system, or doesn't want to wait? Simple. Choose to play with fewer Civs.

Again, I'm not a software engineer, but I would imagine that the AI could be a much more formidable opponent, if not necessarily smarter, simply by programming it to follow certain guidelines set by some of the better Civ II players from a variety of styles rather than following guidelines and priortities programmers might think best. Like making Monarchy a top research priority maybe? And leave it open enough to improve with a patch. This wouldn't require more time to think, just a better use of that time.

I think all this energy being spent in naysaying the more civs thing would be far better spent in finding ways to save and do honors to the Camel.

------------------
The camel is not a part of civ.

THE CAMEL IS CIV !!!!

SAVE THE CAMEL !!!!!!

[Dom Pedro II]

Ralf, I didn't know my rural population topic was talked about in that topic. Nice to see it was discussed somewhere....

Obviously, I can't agree with you on your opinion on it, but I do think that your right about a simpler AI that controls the computer players. I think that templates for city and tile improvements, and unit construction should be used. When certain things become available, the AI should automatically begin construction on them. This would relieve a lot of AI responsibility. They should be as in AOE where you can write your own scripts for that too if you want to adjust the computer players' actions.

I also think that there should be something that forces the AI to send units to certain places like the MoveUnit function in the events file that makes the AI attack certain places. I don't know if its possible to do this, but I think it should be done. The only problem would be with random maps. And I think the AI personalities should stream line AI decision-making in negotiations. An aggressive nation will have certain preset methods while pacifists have others, etc. The computer can zip right through by checking various facts.

I think that migrating populations could help the AI because it would direct them to build cities in areas with rich natural resources because that is where the populations condense. I also think that it does not limit the player because the player can still build settlers and plant cities wherever they want. What I meant when I said they would not be concerned with expansion was that they would not have to drive settlers out into the wilderness and spend to much time with that if the populations were already heading out that way.

Now, my knowledge of programming is limited, and my knowledge of AI programming is non-existent, but couldn't the AI that controls the rural population be a separate thing from the AI that controls the other civs?

------------------
"...The highest realization of warfare is to attack the enemy's plans; next is to attack their alliances; next to attack their army; and the lowest is to attack their fortified cities." - Sun Tzu

Dom Pedro II.... aka Hannibal3

Dom Pedro II - 2nd and last Emperor of the Empire of Brazil (1831 - 1889).