Numbers for the AI

[Securion]

Its seems that if I bild alot of defensive units, the AI is very agressive towards me. While I bild offensive units, the AI is much more likely to agree to peace.
Can anyone else confirm this?

I think the AI is made to look at the 3 different numbers that makes units. Lets say you have 10 deffensive units with a 1-3-1 rating and 10 offensive units with a 4-1-2 rating.
In a AI brain that would look like:
50 points attack rating
40 points defensive rating
30 points range/mobility rating
I think the AI looks at his own defensive rating before saying no to a threat while he takes a look at his offensive rating compared to his victims defensive rating when he wants to attack.
It must work something like that. The question is, how big deal is it for the AI? Anyone know?
This can also explain why the AI surrenders in very strange manners even though the AI hasnt suffer very large losses in bases/cities. It just takes a look at his agressors attack rating and he sees that hes own defensive rating is way to small.
Again, how big deal is this for the AI when making a decission?

[Avenoct]

No, I don't have Deirdre's Phone number either.

Yes, I think you're on to something. Should be easy, but a PIA to set up some tests. There's another thread currently touching on this topic as well...now what was that?

-Smack

[Black Sunrise]

My thread on the computer getting intimidated by numbers:

http://apolyton.net/forums/showthread.php?s=&threadid=23546

I agree: Take a look at the military screen (Fsomething). See the 256-153-20 number for your faction? Since Fireaxis put the numbers in there so smoothly, I'll bet they let the computer look at those very same numbers.

The computer also gets more intimidated when your army is on his doorstep, though.

Indra

[johndmuller]

Re: Numbers for the AI

Originally posted by Securion
Its seems that if I build a lot of defensive units, the AI is very agressive towards me. While I build offensive units, the AI is much more likely to agree to peace. . . . .

Its easy to believe that the AI would use those numbers somehow (hopefully adjusting them for scale like dividing by the number of bases) as a back of the envelope measure of its opponents. It is also easy to believe that the AI's built in autoprobes would have those numbers for everyone. On the surface level, I can see why it would get peaceful when you build up your offense, however its not so obvious why your expanding defense would make it more aggressive (unless it is your puny offense it is looking at). In your example, one can assume that its attack odds were going down as you built up defenses, but at the same time, your attack potential was going down too and from its point of view, you weren't a threat, so maybe it was feeling free to try bullying you for free goodies.

You'd think that it should be comparing its offense to your defense and vice versa and then compare those two differences to decide how to position itself. For example:
-a- if its offense were 30% greater than your defense and your offense was 30% greater than its defense both should probably think about attacking or making peace - letting the other person attack would be bad;
-b- (this would be the the example where both sides have big defenses, but when was the last time you saw the AI with a big defense? and when did you have a big defense?);
-c- if its offense were 30% worse than your defense and your offense was 30% greater than its defense, it should be interested in peace (this is where it often demands that you give up Fusion Power in exchange for ending this vendetta);
-d- if its offense were 30% better than your defense and your offense was 30% worse than its defense, then it ought to be coming down your Main Street any moment now.

It is interesting to try to think what the best strategy is for some of the possible situations, especially if bluffing is considered. With a strong defense, one might feel like any strategy is cool, but we know that the AI can (sometimes at least) put on a decent attack if you just sit there. On athe other hand, having a strong offense, per se, doesn't guarantee your safety, so it might be programmed to use that offense rather than risk your offense being used against it first (it would also be a boring game if the AI just sat there all the time).

In the game, I think that this kind of analysis is probably done by the AI, but may be overridden by other parameters. Against a decent human player, the game score (which I presume that the AI also utilizes) will get lopsided soon enough and then it doesn't seem to matter how suicidal the odds are the AI will have a vendetta on you anyway.