21-12-2004 06:34									#1
	Where to now? Support Apolyton, buy Galactic Civilizations

Game AI

I have been playing strategy games for a very long time now. Ever since the original civ1 was released. I have played every TBS known to man (and woman). I can categorically say that the AI in all these games has been woeful. Admittedly there is considerable variation, ranging from just plain woeful to truly abysmal.

The complexity of the thinking required simply does not match the AI tools that are available at present. My understanding is that a number of people are going to attempt to improve the AI.

Given the difficulty experienced by all games in this area the question must be asked whether there will be a sufficient pay-off for the effort expended. I have absolutely no doubt whatever that improvements can be achieved. What concerns me is that a huge effort could be made here for very little gain. And there is no way of knowing in advance what the likely gain will be.

Alternatives

An alternative approach is to add features to the game which will draw attention away from the inadequacies of the AI. I have already alluded to some of these in earlier posts but I will recap some of them here as well as suggest others.

* Make the AI aggressive and give it enough military units. Sure, it will still be dumb but at least the player will have a situation to deal with.

* Make gold tight and give the player options to deal with this.

* Make research slow but give the player ways of dealing with it. (On a side note, getting advances through conquest is just a bit too easy at present IMHO.)

* Introduce other concepts such Nationality, Religion, Supply and Mining Resources.

Some of these things will require a significant programming effort. The point is that they are more certain to give a suitable pay-off in terms of effort expended.

The AI, of course, will have to deal with these issues as well. But there's always DiffDb to help it out.

I am not suggesting that this entirely an "either - or" situation but rather one of priorities. Please accept my comments in good faith. I am painfully aware that I am not able to make a significant contribution in terms of what needs to be done.

	21-12-2004 09:11									#2
	Support Apolyton or Terrorists Win

Theres a few people here who dont want to change or even add to the fundamentals of the game and some of the things you mention such as supply and resources would make it actually quite like civ3, which is even worse to some. I like the idea of resources but certainly not for the sake of adding another feature.

Perhaps when you think of AI improvements you think of its movement and grouping units, well this needs improving, but a lot can be done in txt files. The advance choices, the way it uses specialists, city improvements, and also what units to build can all be improved or rather, fixed in some cases, also add more strategies for different situations. Peter Triggs, Hex and a few others could probably tell you more than I can about the potential simple tweaks and improvements anyway.

I'd like to see how the AI would do with slower unit speeds, based on their type, foot 1, mounted 2, wheel 3, hover 4, since it has more time to react on defence it could help it a lot, of course it would attack slower but if things were done right the AI would have more units anyway. They had slower units in civ2 and they didnt even need them, because defending was much easier if you were organised, this made for epic games. This means defending cities needs to be made easier too and can be made easier by changing the way combat defence is handled mentioned by Locutus, this is really important, since you can only have 12 units defending a city at one time, 12 units defending a city should always beat the same or slightly higher rank 12 units attacking, and beat them comfortably, but this doesnt happen right now, with 3 stacks you can take 3 cities a turn sometimes, due to the speed of some units. Even if these units were weaker it would still be a problem for the AI though. A human with any unit that can move 4 spaces is gonna be a lot smarter using it than any AI, so why not reduce the unit to 2 spaces, the human has less options and the AI has less chance of failing, instead of realising its gone 4 spaces in the wrong direction, its gone 2.

When the AI can repel a proper attack or keep the human on the back foot even then the game lasts longer. Assuming we've fixed its problems with generating more science and continuing to grow, the AI will continue to be a threat all game.

	21-12-2004 15:28									#3
	Remove this text

AI is what I want above all in a Civ game, I will even enjoy a game with less complexity but a decent AI. I want the AI to make some good decisions, not just to have lots of military units, attack me and force me to go warmonger. I want an AI that can be peaceful enough and build its empire well in peacetime, so that it would not fall behind me technologically and economically.

Civ3 AI is fine. Yes, it uses some cheats, but it does make decent decisions often, and it is capable of full-scale invasions with sending huge amounts of troops over the border, and supplementing that by navy ships. CtP2 AI, with the mods, is also fine, biut it does do some extremely stupid things, and is uncapable of some other things. The main incapability of this AI is its inability to do any reasonable cross-water invasions. It might drop 1-3 units on another landmass, but that is never a threat. Generally, if you control a landmass of your own in CtP2, you are safe from any invasion. However CtP2 AI is unable to surrender properly often, is poor at keeping conquered cities and has some other problems.

But I believe that all effort put towards improving the AI is to be much appreciated. After all, getting a challenge in other ways is easy. You can use the text files yourself to give the AI a 50% research boost, and whatnot, but that will be, artificial challenge, challenge through difference in numbers. Ideally, the AI should be at least competitive on equal level, that is, with production/upkeep/research costs same as the human player. But as it is, any tweak to AI is a huge YAY in my book.

	22-12-2004 01:38									#4
	Support Apolyton, buy Galactic Civilizations

It's interesting that you should mention the across water invasion situation.

According to Peter Trigg some experimental code has been included in the 20/10 build to try and address this problem. This build is essentially unplayable with AI turns lasting in excess of 2 minutes. I abandoned all my attempts circa 1600AD because the waiting became intolerable. Whilst I am definitely in the mood to play I have been forced to stop. (As an aside, there was absolutely no naval activity whatever during the time I played.)

Perhaps I am being unfair but nevertheless it illustrates the potential dangers of going down this path.

	22-12-2004 03:22									#5
	Inflate your Upload Space

I wonder whether the long AI turns are a problem only for you or everyone who's played that build? It seems that others are playing fine... and no, I myself still have not played the latest playtest build, I am in the mood to do so, but don't have an evening to devote to CtP right now.

However, naval invasions are definitely possible to implement. Not a trivial task, but it's obviously doable. I also mentioned them in the first place because that's what I enjoy, cross-water invasions, by me or against me - and that also is what I liked about Civ3, there the AI could invade your landmass, whereas in CtP or SMAC, it would at best be limited to a single unprotected transport with few units.

	22-12-2004 04:58									#6
	Support Apolyton buy from Amazon

Lou Wigman

I know I am an advocate of change wanting some of the better things of ctp re-implementing, changing the end game goals as additions to the tile file but there all complementary alterations what a lot of people suggest these days just seam like it’s not the ctp I pulled out of the box .

Strategic recourses, the need for gold to survive, Nationality, Religion, Supply and Mining Resources

We have to face facts we can rip out the heart of ctp and change the game completely but when do we stop calling it ctp and start calling it civ 3.

	22-12-2004 05:06									#7
	Support Apolyton

Good point... what many of us want first is CtP2, as it was intended to be, but better than it was done. That means that same concepts , but with better AI, better game balance, etc. It's also what I want for now. The problem with vanilla CtP2 isn't that the concepts are flawed or such, but that the game has bugs, poor AI and very poor balance. That is what we are trying to fix.

	22-12-2004 05:17									#8
	Increase Your PM Length

I vote for better AI.

However, I'd like to see other features but am willing to see it in mods. Therefore I advocate that we get help adding more functyions into SLIC that make it easier for a modder to add things like supply and resources. Those with SLIC experience probably can think of a few functions and capabilities they'd like to add. As fr as I know we cant create science bonuses through tileimps or have food be sent from one city to the next or add trade options in diplomacy. If someone could add more SLIC functions that would be nice!

	22-12-2004 06:17									#9
	Help yourself to an AD-FREE life

Just to make it clear, I am not a fan of civ3. Indeed I dislike it intensely.

Top of the list is combat and lack of ZOC for units.

Not far behind is the nature of roads and the extreme micro-management of the workers.

These completely destroyed the experience for me. I have doubts about the strategeic resources as well.

My point is that perhaps an effort should be made to increase the depth and breadth of the game. This can be done without the need of a huge increase in micro-management.

For example (just off the top of my head) suppose we introduce a terrain type called "gold". If the human player is under sufficient financial pressure then this type of tile becomes very important in the game. Perhaps to the point of wanting to declare war to get it.

This is an example of adding breadth to the game without excessive micro-management.

I agree that the AI needs looking at. Clearly there are easy things that can be done to improve the AI.

	22-12-2004 15:49									#10
	Full PM-box? Change here!

we can do that with now.


All you would have to make is a good called gold ( in fact i have the spr file) and increase the revenue for it in trade.

	22-12-2004 23:23									#11
	Support Apolyton buy from Amazon

anybody ever discuss adding more SLIC functions? or even looked into it?

	23-12-2004 00:25									#12
	Increase Your PM Length

Yes i recommended an add slave function to speed up the city capture code

	23-12-2004 00:28									#13
	Support Apolyton buy from Amazon

maybe we should make a thread of wishlist for new SLIC functions?

don't we need a get current government function?


PS BigMC I tried PMing you a bunch of times but apolyton says the account is full.

	23-12-2004 01:05									#14
	Browse Apolyton AD-FREE

d'oh

a wish list would be good.

	23-12-2004 01:19									#15
	Support Apolyton, pre-order Civilization IV

Go ahead and post it .

	23-12-2004 02:16									#16
	Support Apolyton buy from Amazon

quote: Originally posted by Solver I wonder whether the long AI turns are a problem only for you or everyone who's played that build? It seems that others are playing fine... and no, I myself still have not played the latest playtest build, I am in the mood to do so, but don't have an evening to devote to CtP right now.

Of course if you play on one landmass maps only you don't have any problem. But if you play island maps you have definatly a problem if you use the built provided by calvitex.

quote: Originally posted by E maybe we should make a thread of wishlist for new SLIC functions?

Why don't you post one?

quote: Originally posted by E don't we need a get current government function?

Do we have now already three ways to figure this out, don't we.

-Martin

	23-12-2004 03:44									#17
	Support Apolyton buy from Amazon

quote: Of course if you play on one landmass maps only you don't have any problem. But if you play island maps you have definatly a problem if you use the built provided by calvitex.

So, does calvitex's build indeed cause water maps to run slow, is that confirmed? As said, I haven't played the latest build.

quote: Do we have now already three ways to figure this out, don't we.

But good point there, a simple get govt SLIC function is better than all the workarounds that exist.

	23-12-2004 03:46									#18
	Get a bigger avatar today!

Love This game!

Would like to see the AI expand and hold ground. What I mean is that if the ai goes to war with you will not forget defense , What I have done is only change some numbers in Goals file and and other files is to raise attack numbers but kept defense numbers as in 1.11
Will give you my numbers in another thread . What happens is the ai will send some stacks to attacked city or face off your units. If units come out of the city -units from inner part of the empire would replace them
I know many in here know the files systems alot more than Ido put have been enjoying testing numbers to see what would happen.
Defense is not always passive. The more land the ai controls and is able to hem in the human player the harder it will be for you to win the game
Chess is a good example of this. Weaker players will go for a closed game . Alot of defensive moves to keep control of the center . If the ai could expand and build up defense fast the affect would keep you off balance.
A attacking ai that could take out your cities quick would be geat but this would mean that many ai cites are open to counter attack by anyone on its borders.

	23-12-2004 03:59									#19
	Suffering from ads?

quote: Originally posted by Solver quote: Of course if you play on one landmass maps only you don't have any problem. But if you play island maps you have definatly a problem if you use the built provided by calvitex. So, does calvitex's build indeed cause water maps to run slow, is that confirmed? As said, I haven't played the latest build.

Afterwards I added the changes by calvitix after the last all pack I got that slow downs, I think they come from his first pack afterwards that all pack. And the no naval activity at all came with his last pack.

quote: Originally posted by Solver quote: Do we have now already three ways to figure this out, don't we. But good point there, a simple get govt SLIC function is better than all the workarounds that exist.

Actual only one of them can be considered as workaround. player[0].govttype is a string and since the StringCompare function works we have a direct workaround. Afterwards Peter added the player[0].government built-in we have a direct way to figure it out, and since I overloaded the player slic built-ins from that you expect an integer as return value, you can now figure out the current government with the player[0].govttype built-in directly as well. I don't see here the need to find some workarounds.

-Martin

	23-12-2004 04:10									#20
	Support Apolyton, buy Call to Power 2

Ahh, so the integer return values for government types are done now as well, then I can indeed agree that it's pretty much everything. A separate function might be created of course, but is indeed not required then.

	25-12-2004 08:17									#21
	Support Apolyton, buy Civilization: The Boardgame

How far can you go?

It seems so much has been done already to improve the game . I still recall when the first Diplomod came out it was a bright spot in the long line of improvements.
Cradle and other major mods were unbelieveable for a game the had a very poor AI.
I think taking some movement points away like stated above is a start . Maybe giving goverments along with units that only the ai could start and buid might work. Could spread the cheats out more using Units and Govement texts files this way.Wouldnt have to depend so much on the DiffDB file.

	25-12-2004 18:05									#22
	Lose 30 kilos (of popups)

Theoretically, we can now do anything as we have the full source code, but practically, we can't, because manpower is seriously limited. If we had a couple of people working full-time, it would have been a different story .

But yes, indeed, much has been done, with mods and with the source. I should note, though, that different mods have a different approach to the AI. Cradle makes it challenging by making it aggressive and improving the military part, while for instance GoodMod tries to improve general decision making on the AI's behalf, which is what I personally prefer. IMO, peace should always be a viable option in civ, therefore, the AI should be able to build well and not fall back during peace - or alternatively, conquest other AI civs to grow and possibly outgrow you.