Announcement

**player1** · May 16, 2003, 18:43

If game is difficult for you, play on easier level.

Some higher levels are supposed to be fun only for those who find lower levels easy.

**Spiffor** · May 16, 2003, 18:49

If you want to play in absolute fairness, play at Regent difficulty. This way, the AI will have no advantage whatsoever on you, nor the opposite. You will only do a battle of wits against the AI, and your victory or your loss will only be explained by your skill.

It seems that with civ3, there is something like more high is your player level, more small is the probability for you that most of warrior will loose against the other.

It seems so, but it isn't so. The difficulty level is not present in the combat resolution formula.
Except when it comes to Barbarians. At chieftain level, your troops have huge bonuses when fighting Barbs, and at the deity level, Barbarian troops fight as well as if they belonged to another Civ (i.e no combat bonuses).
When you are engaging a foreign unit, the calculations are exactly the same for your unit and for its.

- Cultural and science discovery as well as territory covering is always bigger for others

Did you stop playing at monarch level ? Please remember that at higher levels, even the most experienced players have trouble keeping up with the AI at the beginning of the game.

I think that if you want win with this game, you have to understand the civ 3 strategy, and then you have some chances.

I don't really understand this point... Do you complain because you can't use a real life strategy in Civ ?

or 15 min. only after having started a new game, because 2 or 3 countries has decide to do war with me despite commercial exchange

I have never, ever, been ganged up on that early, even when I was a rookie. What are you precisely doing this first 15 minutes ? Do you leave your cities wholly undefended ? (and even that won't do the trick : the AI very rarely attacks before the end of the settlement phase)

And why can I not request mutual agreement to protect each others when I have open an embassy ?

Because you have not discovered nationalism yet. Nationalism give a few new diplomatic options, Mutual Protection Pact is one of them.

Also too much time, others countries don't want take part to a conflict I have with some others countries.

When they feel it is too risky, or they have good deals to lose, or you'll betray them, I don't see the problem. If you're the weakest, be aware the AI will be against you.
Civ is about realpolitics, and there are no heroic 'white knight' to help you / save you in times of trouble. There are only crows that want to eat a bit of your corpse before the others.

As a conclusion, it is normal for the game to be 'unfair' at higher difficulty level, because the AI doesn't get smarter past regent difficulty. Regent offers you the smartest AI the game can offer, and the only way for this game to be more difficult past regent is to give bonuses to the AI.
Yes, it is sad, but an artificial intelligence is extremely hard to design properly, and the work that has been made so far for Civ3 is better than in any other game I've seen

.

**Chemical Ollie** · May 16, 2003, 18:59

Eat and die...

**jobuck** · May 16, 2003, 20:08

spiffor:

It seems so, but it isn't so.

well well well

I don't really understand this point... Do you complain because you can't use a real life strategy in Civ ?

I mean, to get a chance to win, you have to understand how civ works. That's not a natural approach for me.
I believe that the civ' ai is supposed to copy the human way of think. And it is not to the player to understand how work the civ' ai to be able to work.

You know, just after my post, I try again to play cool and good. But again it's not fair. Let say I have a army unit of the type A fighting an army unit of type B.
Let say A loose against B. Where is the logic if now I fight with B on my side against A on the other side and if I loose as well ???

I have never, ever, been ganged up on that early, even when I was a rookie.

Okay, let say 20 min. This happen each time my army get only warrior and when I'm either building very much cities against the others, or when I get some cultural impact (e.g. colosse + grand phare).

When they feel it is too risky, or they have good deals to lose, or you'll betray them, I don't see the problem.

1) Imagine their is two medium power, and one big power. If the two medium power think like you say, then they will not react if one or the other is attacked right ? So because they have a smaller power than the bifg one, they will loose ok ? Then what. After winning the war, the big power can fight the other medium one.
So in such case, it is obvious that the it is in the interest of the two medium power to fight again the big one.
Now at the begining of the game, you get often such situation (let say in the first hour). In my last game, I only know 2 others countries. One was more strong than me. But the other one were at the same level.
I ask him to fight again the big one and he refused.
And accordingly to the graph showing the power, his "power width" plus mine were more large than the big power. So what now ? After the big power will have put me out of the game, virtually, it will be the round of the other.
And this remember me again that if your country is too big (the other are limited in their expansion because you, they will start a war against you). That's what happen in the example above. And it was at the "start" of the game.

Yes, it is sad, but an artificial intelligence is extremely hard to design properly, and the work that has been made so far for Civ3 is better than in any other game I've seen .

I agree that design such ai is not an easy job.
In the other side, that doesn't mean it's a good one or a fair one. I don't feel more the level is high, more the ai is strong as with a chess game.
I have the feeling that more the level is high, more your chance too loose is high because the others nation have more power in an unfair way: they "born like that", and they have a higher probability to win if both side present the same warrior.
That's not ai for me.

**jobuck** · May 16, 2003, 20:28

I just want to add something regarding the ai' design.
If I were involved in the development of the ai, I would not do my design testing too much the parameters of the others nation *as it seems to be the case* and take some actions accordingly.
For eg NationXPower = w1*n1 + .. + wn*nN
where wX is the weight of the army of a certain type (more height is the weight, more powerfull is this unit) and nX is the number of elements of this unit,
Then doing something like:
if (myNationPower > nationXPower && nationXPower.isBad()) {
attack();
}
I mean, I would no only perform such test.
I will instead try to found some models.
Let say 5 models, corresponding to 5 levels of difficulties.
The highest level corresponding to the more complexe model. The lowest one to the easiest model.

Then I'll design a neural network and run it billion of time taking in input some parameter of the model.
The neural network weights and offset being adjusted to converge to a victory.
Then, I will use the weight found by this network to ponderate the "litteral approach" initially described.

I don't want here to show my science you know.
But if you speak ai, you could speak neural networks.
That something well known now since years, and lot of simulator already exist in the market.

Now, also, how could I know if the civ ai doesn't integrate such approach ?
I don't now. But the result is not impressive as we could be using such technologies.

**Spiffor** · May 16, 2003, 20:50

Sure, the AI could be better, but be aware it has been designed by about one person (Soren Johnson) who had limited time and limited budget to make it understand all the features of the game.
When I compare to past games I've played, whether TBS or RTS, I'm baffled by the result

And besides, I'd assume the attack formula is more complex, because I very often am in peace with the world despite having a weaker army. I get attacked at some point, but not before enjoying a long period of peace.

**Louis XXIV** · May 16, 2003, 20:51

Do you check the unit stats?

Warriors have a 1.1.1 (that is their attack, defense, and movement). That is all self explanitory. The attack is the strength of the unit when attacking. The Defense is the strength of a unit defending. And the movement is how many squares they move in a turn. The terrain a unit is on imroves the chances of defense. Then it is put through a formula. A=Attack, D=Modified Defense (uses terrain, city, walls, and fortification to improve the defense).

The formula is A/(A+D)

If you use a warrior and a warrior you have slightly less then a 50% chance of winning the battle, if they attack you, they do to. If you fortify the unit it improves greatly as well.
Attack a Spearman (1.2.1) with a warrior and you have closer to a 33% chance of winning.
However, if you attack a Spearman with a Spearman you still have the same odds (Spearmen and Warriors have the same attack)

Re: Game difficulties (anyone who wants to fill in details, please do)

In cheiften, it takes them 1.5 times as long to do everything. Happiness is really easy to deal with early in the game

In Warlord it takes them 1.2 times as long. Happiness is more fair

In Regent everything is equal. You have to know at least the strategies the AI knows to keep up. You can probably do better.

In Monarch, it takes the AI .8 times to finish things (they work faster). Happy citizens are harder to get

In Diety (I need the details, but I think) the AI starts with a TON of units (including an extra settler). Happy Citizens? You wish

The AI has an attitude level. If you do certain things, they'll like you better or hate you worse. If you are weak, they'll declare war on you (many players would do the same to the AI)

**jobuck** · May 16, 2003, 21:06

Louis XXIV :

hmm... okay
This is very interesting.
I have to confess that I didn't had really take care of these A.D.M parameters.
I have to try playing the game taking that stuff in consideration.

Thank's. You mark for me a point in favor of civ3.

**Spiffor** · May 16, 2003, 21:14

I have the feeling that more the level is high, more your chance too loose is high because the others nation have more power in an unfair way: they "born like that", and they have a higher probability to win if both side present the same warrior.

This is exactly the case. The AI gets smarter only until Regent level. Regent is its best intelligence. Since Soren couldn't develop an AI that matches human intelligence, the only thing he could do for experienced players looking for a challenge was to give unfair bonuses to the AI. Monarch, Emperor and Deity are difficulty levels where the AI plays unfair. Yet, none of this unfairness affects combat prowess of units (except Barbarians).

You know, just after my post, I try again to play cool and good. But again it's not fair. Let say I have a army unit of the type A fighting an army unit of type B.
Let say A loose against B. Where is the logic if now I fight with B on my side against A on the other side and if I loose as well ???

Good Luck or bad luck. Civ isn't Chess, and there is no such thing as a knight who will automatically win against a pikeman, or the other way around. Since tactical battles cannot be done in Civ, the only way to show the aleas of battle is trough random numbers. Of course, a stronger unit sees his chances to win much bigger, but no unit is 100% sure to win, ever.

However, if you want to reduce the importance of luck in the game, I suggest you do like me :
- save civ.bic (or .bix if you play PtW) in another folder
- go to the editor
- edit Civ3.bic / .bix
- edit the rules, go to the "combat experience" tab
- Give twice as many hitpoints to every experience level. Flukes will be much less significant from now on.

Okay, let say 20 min. This happen each time my army get only warrior and when I'm either building very much cities against the others, or when I get some cultural impact (e.g. colosse + grand phare).

Ah yes, I almost never build early wonders, because I prefer focusing on expansion at that time. Remember that unprotected cities are bait to the AI, even if they are deep within your empire, and even if the AI should not know about it (because it knows about it).
A spearman in every city, at this stage of the game, would deter any minor attack. However, you should have a good protection in your capitol and in other wonder cities. If the AI sees it can grab a wonder without effort, it'll become extremely agressive immediately. Actually, I would do the same too

Imagine their is two medium power, and one big power. If the two medium power think like you say, then they will not react if one or the other is attacked right ? So because they have a smaller power than the bifg one, they will loose ok ? Then what. After winning the war, the big power can fight the other medium one

If I remember Machiavelli's The Prince correctly, your neighbour would be well advised to attack you to gain your power while your armies are busy at the other side of your country. That's the best way to become powerful with little cost for the upcoming war against your first agressor

But if you want alliances, one tech or two would be enough. If you manage to discover a tech sooner than anyone else (a tech nobody has picked yet), you'll make several allies. Don't hesitate to ally with everybody you can : if anything, it'll prevent them to ally with your enemy

**Inverse Icarus** · May 16, 2003, 22:10

**Dominae** · May 16, 2003, 22:44

Originally posted by jobuck
I mean, I would no only perform such test.
I will instead try to found some models.
Let say 5 models, corresponding to 5 levels of difficulties.
The highest level corresponding to the more complexe model. The lowest one to the easiest model.

Then I'll design a neural network and run it billion of time taking in input some parameter of the model.
The neural network weights and offset being adjusted to converge to a victory.
Then, I will use the weight found by this network to ponderate the "litteral approach" initially described.

Please elaborate.

It's easy to say: "just apply neural nets and the AI will be better". It's quite another to actually do the work. If you know about neural nets, you know that: 1) the problem of "playing Civ3" is fiendishly difficult to set up (what are your inputs...cities, units, techs, improvements, commerce, everything?), and 2) it would take almost forever for the neural net to find a "good solution" (corresponding to a playable AI), if all you want to ship with the product is one gigantic net. And if you want to break the problem apart and send multiple nets (for instance, one for "combat", one for "research"...not that any of this makes sense), you're avoiding the problem, because those sub-problems are themselves fiendishly difficult, and there appears the problem of re-integrating the the results of each sub-problem.

It took two guys about the same time it took Civ3's AI programmer (Soren) to come up with a neural net player for Checkers, including all the months the program just sat there alone, churning out weights and more weights. Now consider how much more complex Civ3 is compared to Checkers.

Hand-waiving and muttering "neural nets" is not valid argument against Civ3's AI.

Dominae

**Dominae** · May 16, 2003, 22:51

Re: civ 3 is not a fair and good game

Originally posted by jobuck
I play the best I can with this game.
For me a good strategy game is a game you can predict.

Many games that involve strategy also involve a degree of luck/randomness. Take almost any card game, for instance.

This are tooooooo much easy programing technics that will not convince peoples who like faire battles and want to build real strategy, and not just play with probabilities.

So, the people who are good at Civ3 are just "lucky" then, since it's all just "playing with probabilies"?

I don't want to have to be "civ3 certified" to get a chance to win !!!

You want to win at Civ3 without being good at Civ3? Sounds to me like you want the game to be more random. Are Chess Grand Masters "Chess certified"?

Dominae

**jobuck** · May 16, 2003, 23:04

Dominae, spiffor:

You made some interesting comments.
But it 5am here. Will reply tomorrow.

**Dominae** · May 16, 2003, 23:18

Originally posted by jobuck
Dominae, spiffor:

You made some interesting comments.
But it 5am here. Will reply tomorrow.

Civ3 kept you up until the wee hours of the morning?

Dominae

Announcement

civ 3 is not a fair and good game

civ 3 is not a fair and good game

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment