Help Soren; lets improve the AI.

FrostyBoy · November 11, 2003, 16:34

The amount of work required to code the AI must be a tremendous job.

There is a large post called "Beyond Sid" at CivFanatics, it's a thread about a game being played at a stupidly difficult setting; played by Bamspeedy (sp?). He is absolutely flooring the other Civs. At the time of typing this, he has one Civ left to destroy: the Celts.

I noticed that even at this difficult setting, the AI has really not 'understood' the impact of Bam (Bamspeedy) taking over Civ by Civ; not taking into account of the threat and power that he possesses.

Quite frankly, the AI just sat there and and let themselves die; and I felt it was just pathetic.

Now, of course, more brains are better than one, especially experienced ones. I think if we can pull together a whole bunch of ways to help Firaxis improve the AI; for example, ways for the AI to learn the players style of playing, being more unpredictable, being able to wage massive wars successfully; Firaxis might take a look and might consider a future patch for finalising the AI, and hopefully finalising Civilization III.

So if we can just start posting our thoughts on how to Improve Civ3's AI (Even if some of the features have to be tweaked/changed/added/removed).

I'm not very good at understanding the AI (Haven't played many games lately). But I do think the AI should be more aware of how dangerous other civs are to them, and what they need to do to counter-act that danger.

I also think that the AI should have the ability to carry out a massive-scaled war - by preparing a certain amount of required units inside the cities. Then once ready, load the units onto ships and send em over - and continue to send in reinforcements. And before this, I think that the AI should try to gather as much information as needed from the enemy, such as strength of cities, importance of cities, best places to attack first, etc.

Of course, I can't go into great depth and reason for my suggestions, and I think that is what is really needed to help the programmers tweak the AI.

Asmodean · November 11, 2003, 16:38

The crucial word here is funding, I'm afraid.

I have no doubt that Firaxis are commited to making Civ 3 a great game, but the real question is: Would it be possible to allocate the funds needed to do this?

I am afraid that Soren has no call whatsoever regarding this.

ATARI: Please send more money

Asmodean

WarpStorm · November 11, 2003, 16:43

I think Asmodean has it correct. The problem isn't Soren's skill. I'm sure he is plenty skillful enough. It's the fact that he (or anybody at Firaxis for that matter) has so little time to spend on improving the AI.

Remember, every minute that Soren (or any other programmer at Firaxis) is called on to rework Civ3's AI is a minute that he isn't working on "The Next Big Thing" (which, incidentally, is most likely what is paying his salary).

Stuie · November 11, 2003, 16:53

Yeah, Soren's probably improving the AI, but for Pirates not Civ3. He wasn't even directly involved in the AI programming for C3C, he acted more as a consultant and Breakaway seems to have done the grunt work.

dexters · November 11, 2003, 16:56

I'm not 100% who worked on the AI in C3C. I get the impression that it was programmer(s) at BreakAway who did the majority of the work.

Snoopy, you've raised an interesting point. There are several threads in the C3C forums also on the AI dealing with everything from improvements to n ew issues.

Solver · November 11, 2003, 16:58

Soren is an amazing AI programmer. He was the only one to work on it, at least for vanilla and PTW, and created the most competent AI yet in a civ game.

Kuciwalker · November 11, 2003, 18:40

I thought C3C had a really improved AI?

dexters · November 11, 2003, 19:58

It does. But the AI's passivity to this particular strategy is nothing new.

I've noticed though the AI can and will make a more strategic choice about when they go to war in C3C.

vmxa1 · November 11, 2003, 20:51

It is sort like boiling frogs or lobsters. If you raise the heat slow enough they do not realize the danger until it is too late.

With the new aggressive setting, will they go KAI if all are set high?

Panag · November 12, 2003, 09:27

hi ,

check this out >>> http://www-cs-students.stanford.edu/~soren/

he has some brains

but one has to take in serious consideration that in order to build a better AI , well the comp specs have to go higher , and on how many double sided dvd disks is it going to be burned , ......

have a nice day

dojoboy · November 12, 2003, 15:14

Quote:

Originally posted by Sn00py

There is a large post called "Beyond Sid" at CivFanatics, it's a thread about a game being played at a stupidly difficult setting; played by Bamspeedy (sp?). He is absolutely flooring the other Civs. At the time of typing this, he has one Civ left to destroy: the Celts.

I noticed that even at this difficult setting, the AI has really not 'understood' the impact of Bam (Bamspeedy) taking over Civ by Civ; not taking into account of the threat and power that he possesses.

Bamspeedy is a sick individual.

But Sn00py, I think your assessment of how we'd like to see an AI respond is dead-on. If 8 of us could sit down and truly play an epic set-up in MP uninterrupted, we'd see more action taken by a number of human controlled civs against a human player who was slowly, or quickly, ammassing power through systematic wars aimed at enlarging his or her industrial capacity. It would be a real thrill to have an AI representative hail me to deliver a warning that any further imperial operations will result in "a war to end all wars." Then, a decision would have to be made, one that may end up where the human player must carry on w/ out any trading partners.

pvzh · November 12, 2003, 16:05

Hmm... Beware what you are asking for! Do you want gang up on a human player? I do not think it is really good idea.

Panag · November 12, 2003, 16:14

Quote:

Originally posted by pvzh
Hmm... Beware what you are asking for! Do you want gang up on a human player? I do not think it is really good idea.

hi ,

nah , that wont happen , after a while they are occupied with themselfs , .....

have a nice day

Asmodean · November 12, 2003, 17:58

We definately don't want to see the AI ganging up on the human player. Other TBS games have done that, and it is not funny.

But a smarter AI, one that anticipates your moves, and looks for patterns in your behaviour would definately be a nice improvement.

Asmodean

vmxa1 · November 12, 2003, 18:57

What is being over looked here is that the people here are deep into the game. This means they know how to beat it. Most buyers of the game are not even winning at Chief. I have had contacts with several players that lose at this level. I have seen winning scores of less than 200, so if they do make it better they will just be cutting off more players.

I also doubt that the AI will be able to handle learning a players sytle until well into the future. It will take more horse power, more time and that means money and again limiting the people that can play the game.

Asmodean · November 12, 2003, 19:02

Quote:

Originally posted by vmxa1
I also doubt that the AI will be able to handle learning a players sytle until well into the future. It will take more horse power, more time and that means money and again limiting the people that can play the game.

Not nescesarily (sp?). The GalCiv AI does exactly what I have just described above

It's just a question of the skill of the programmer, and off course.....the funding.

Asmodean

dojoboy · November 12, 2003, 20:12

Quote:

Originally posted by Asmodean
We definately don't want to see the AI ganging up on the human player. Other TBS games have done that, and it is not funny.

But a smarter AI, one that anticipates your moves, and looks for patterns in your behaviour would definately be a nice improvement.

Asmodean

Well, the same would go for an AI that was land grabbing. It would have to be a feature that "plays" fair in this regard.

dexters · November 12, 2003, 21:35

Quote:

Originally posted by Sn00py
The amount of work required to code the AI must be a tremendous job.

There is a large post called "Beyond Sid" at CivFanatics, it's a thread about a game being played at a stupidly difficult setting; played by Bamspeedy (sp?). He is absolutely flooring the other Civs. At the time of typing this, he has one Civ left to destroy: the Celts.

I noticed that even at this difficult setting, the AI has really not 'understood' the impact of Bam (Bamspeedy) taking over Civ by Civ; not taking into account of the threat and power that he possesses.

Well Sn00p, I finally dug up the thread and read it. Bamspeedy is a sick player who I is just amazing. But you have several things.

The states quite clearly the game being played is PTW 1.27f with (i assume) modded difficulty levels to emulate Sid level production bonuses for the AI. The game is not C3C.

Another significant issue is that it's an islands game. This makes his strategy of separating AI civs possible and since PTW does not have the benefit of the AI navy improvements found in C3C, a big questionmark is raise then as to how effective the AI is in this game. Even in C3C, naval invasions is logistically more difficult and playing in island settings tends to negate large killer AIs.

I agree with the gist of your request. We all want more competitive AIs. I alone have started several threads on this, but I think in this case, Bam's game may not be that applicable to your request. If it's a pangea C3C game that's being beaten in this fashion, then I give mad props to whoever has done it. It will truly be an interesting game to watch.

alva · November 12, 2003, 21:46

Dexter, could you post the link please.

-
I agree with vmxa here. I'm sure that without these boards, most people here (and certainly myself included) would be playing one or two levels lower if not more.
Overall the AI is good but not good enough to handle several hundreds of inteligent, dedicated ~~lunatics~~ eugh gamers....

dexters · November 12, 2003, 23:33

Here you go.

http://forums.civfanatics.com/showth...threadid=66169

Pyrodrew · November 13, 2003, 06:17

First, I agree an Islands game is a poor example to use.

Quote:

the AI has really not 'understood' the impact of Bam (Bamspeedy) taking over Civ by Civ; not taking into account of the threat and power that he possesses.

That's exactly why an AI must understand the concept of a balance of power. The AI instead is always short-sighted looking for the best deal TODAY rather than the long term consequences. AI Civilizations must not only realize (as humans do) that there are goals to winning the game, but strategically play for those goals offensively & defensively. Offensively to try to obtain those goals AND defensively try to prevent other leading civilizations from obtaining them (non violent ways to do this too). For now it's like watching a soccer game & you see 1 team that doesn't have any defensive players on the field. Likewise, the AI doesn't try to stop you from winning (making a goal) as it is only interested in getting Ivory (or some tech) at a good price TODAY. Right now once the human player gets to a certain size there is little challenge as it is simply a steam roll & usually predictable.

In the end, there IS a winner & the AI doesn't know that. The AI should not know you are a human player, but if you have half the cities in the world it should realize how you are a threat to making it lose. And please don't tell me that's not realistic for the AI to play like the game doesn't end at a certain time. For those looking for realistic simulations go play "The Sims".

WarpStorm · November 13, 2003, 06:54

Of course you realize that balance of power (i.e. taking out the leader) was one of the things Firaxis intentionally removed as so many players of Civ2 and SMAC thought that feature was too 'gamey' and complained loudly out it.

Panag · November 13, 2003, 11:00

Quote:

Originally posted by alva
Dexter, could you post the link please.

-
I agree with vmxa here. I'm sure that without these boards, most people here (and certainly myself included) would be playing one or two levels lower if not more.
Overall the AI is good but not good enough to handle several hundreds of inteligent, dedicated ~~lunatics~~ eugh gamers....

hi ,

thats the power of a group of people , one person is good at this , the other at that , etc , ......

and it would be more then just one or two levels , many bugs would not have been seen , therefore no solutions for them , and a great deal of time would be wasted cause the individual would spend hours trying to find something that now has allready been found , ......

not to mention of the countless ideas that came from apolyton and made there way in several games

have a nice day

Pyrodrew · November 13, 2003, 17:43

Quote:

Of course you realize that balance of power (i.e. taking out the leader) was one of the things Firaxis intentionally removed as so many players of Civ2 and SMAC thought that feature was too 'gamey' and complained loudly out it.

And now the same people (and others) are complaining about how the game is too boring & predictable at the middle & the end.

Instead of delaying or reducing the balance of power's impact (to please those vocal people) Firaxis made the mistake of removing it all together (which left strategy gamers screwed & any attempts to help the AI with the editor VERY limited). The pendulum has swung way too far to the other extreme & it's an ugly scene.

If a human player owns 1/2 the world & the AI is improved all those other 9 Civs will play intelligently against each other & destroy themselves for short-term gain. And the leader (usually human) can easily manipulate (and humans can do this better) the AI for an easy win. A good strategist tries to win the game & if the AI never tries to stop another civ from winning (defensive strategy) it isn't a good strategic AI.

So even if Soren improves the AI in other ways you will see little or no difference, except for the beginning. But the beginning by almost all threads I've read is the best part & of little problem. Unless the balance of power is added back in to at least some degree Civ3 will always be lacking a good strategic AI & instead have this "The Sims" AI.

Quote:

I'm sure that without these boards, most people here (and certainly myself included) would be playing one or two levels lower

True, but the higher levels don't improve the AI as it only gives them more resources, troops, etc. And I'm sure I'm not the only one who wanted to win some games before reading spoilers here. I like the AI for first part of the game, but near the end it's boring. Excessive corruption (which doesn't have many fans) as the only end game challenge is tedious.

FrostyBoy · November 14, 2003, 20:13

Right, so what we need here is an AI that can:

Evaluate Threat
How to Counter-act Threat
and How to find a way to beat the game

Would that be it?

Kuciwalker · November 14, 2003, 20:17

Also, an AI that has know idea whether or not another civ is human or AI controlled.

vmxa1 · November 14, 2003, 21:40

Quote:

Originally posted by skywalker
Also, an AI that has know idea whether or not another civ is human or AI controlled.

Well in the PTW guide they claim the AI does not know AI from Human. This is in the developers notes, not the authors, for what ever it is worth.

Kuciwalker · November 14, 2003, 21:44

Well, the .bix file directly disproves this.

JimMac · November 15, 2003, 16:00

I am pretty sure that I read a comment from Soren (dont ask me for the link) that he could make an AI which was harder to beat, but the game wouldn't be as much fun!

Bamspeedy · November 15, 2003, 21:57

No, it's not fun if you lose all the time.

Unless you have a 'big blue' type of AI/computer (the computer that can sometimes beat the world's best chess players), then an AI can't compete against human intelligence.

I've played this game and surfed the Civ forums 40-60 hours a week for 2 years. If I spent that much time playing and researching the game, imagine how much time it would take them to program the AI to counter all the 'tricks of the trade' I've learned during all that time.

You can make a patch to adjust the AI strategy to counter an exploit or strategy, but then the human just learns from it and changes his strategy a bit and then finds a new way to exploit/abuse/manipulate the AI.