AI State.

PirateRob · #1 April 24th, 2002, 07:05 PM

I noticed that many of the AI mods don't make a lot of use of the AI State flags. Looking at some of the <name>_AI_Research.txt files, I noticed that in almost every case the order of research is the same for each state. (except 'Not Connected' which focuses on WP technologies)

Wouldnt it be possible to make a smarter AI (and a bit less predictable one too) if each state had its own AI research que? For example, in the 'Explore' state I would think the AI should research engine and supply technologies as a priority.. but in 'Defend (short term)' they should research defensive / weapon techs with more frequency. I notice that these states are available for MANY of the AI files.. but only realy used in the construction que related AI files.

Is there something I don't know about these AI states? do they work? or is it just that nobody knows how they work or what conditions apply for entering and exiting each state?

Rob

Jmenschenfresser · #2 April 24th, 2002, 08:03 PM

I think it is fear of the second of your two conclusions. Not sure if anyone really knows why the AI goes into different states.

Rollo · #3 April 24th, 2002, 08:39 PM

The most common AI state is Defend (Short term). As long as there are some enemy (or none treaty ships) nearby the AI will remain in this state (over simplified). Sometimes the AI switches to Attack or Prepare to Attack, but those stages do not Last very long.

I have seen some AI files make more differences between the AI states in research, but IMHO it does not make any big difference, since the AI spends aprox. 90% of its time in Defend (Short). So it is probably not worth the extra effort.

Defend (Short term) really should be viewed as "Infrastructure" IMHO.

Rollo

Atraikius · #4 April 24th, 2002, 10:31 PM

Rollo - I've been catching some AI's in other states a lot more frequently recently (try watching the orks in a few games - I've seen them in explore, infrastructure, and attack more than anything else).

PirateRob - What you were asking about does work, but it is very slow making any progress at it, making changes based off of a single game frequently only causes problems, you almost need to use a statistical model using many games to determine problems. Whenever a new tech is added to the que, it is chosen from the branch that corresponds to the state that the AI is in for that turn. Can throw the AI off a bit because it could go through attack, explore, infrastructure, and defend states with the state changing each turn while at war and attacking thier enemy. For an example look at the Krill and Orks in the TDM-Modpack, I have them shifting research priorities based on thier state. It does make them do some odd and surprising things occasionally

Baron Munchausen · #5 April 24th, 2002, 11:17 PM

quote:
Originally posted by PirateRob:
I noticed that many of the AI mods don't make a lot of use of the AI State flags. Looking at some of the <name>_AI_Research.txt files, I noticed that in almost every case the order of research is the same for each state. (except 'Not Connected' which focuses on WP technologies)

Wouldnt it be possible to make a smarter AI (and a bit less predictable one too) if each state had its own AI research que? For example, in the 'Explore' state I would think the AI should research engine and supply technologies as a priority.. but in 'Defend (short term)' they should research defensive / weapon techs with more frequency. I notice that these states are available for MANY of the AI files.. but only realy used in the construction que related AI files.

Is there something I don't know about these AI states? do they work? or is it just that nobody knows how they work or what conditions apply for entering and exiting each state?

Rob

Actually, it's not what you 'dont know' about the AI states but what you have not thought through. What if it were modded to research 'infrastructure' technologies when not in conflict and 'weapons' technologies when in a conflict? It would then be MORE predictable, not less, and clever players could exploit it by choosing the AI which had been in fewer conflicts to attack for themselves, since it would almost certainly have less advanced weapons and more advanced infrastructure techs. 'Free pickings' for the exploiting player! Or conversely, you could cripple an AI by keeping it in low-level conflict and preventing it from researching things that it needs in the long term - infrastructure techs like higher levels of resource extraction. Finally, it's quite possible that an AI would expand peacably for 50 turns researching peaceful techs and then run up against a human player to find itself completely helpless. I don't think the 'AI states' are useful in terms of research at all.

PirateRob · #6 April 25th, 2002, 12:37 AM

I have to disagree with the Baron. A SMART modder will EMPHASIZE certain technologies based on AI State but not completely ignore ANY critical technology.

Think about it a second.. a HUMAN player will generaly start the game out researching infrastructure items... and SOME military techs.. enough to satisfy thier need to feel secure from attack. However, if a war breaks out, Military tech becomes paramount in the race to get a tech edge in combat. A smart human player will try and sneak in infrastructure items that will enhance production even in war time to try and avoid being outpaced economicaly by his enemies.

The fact that the AI's all have ONE research list makes them pretty 'middle of the road' and very predictable. What the Baron is talking about is MANIPULATION of the enemies research queues, which could be a problem but its less of a problem than the 'middle of the road' approach.

My experience tells me that middle of the road is a BAD tactic in a 4X game.. the key to victory is to balance growth with military strength.. which SOUNDS like a middle of the road approach doesnt it? Well, its not exactly. I have observed, over and over, that focusing on growth early with just enough military strength to ward off invaders is rewarded in the long run. 4X games are often won by the BEST ECONOMY.. which 'middle of the road' isnt going to get you. The player who can keep pumping out infrastructure the longest before switching to a wartime footing WILL have the stronger position in the end. The balancing act is keeping enough defenses to protect your growing economy until you need to switch to evil empire mode

and kick everyones @$$

Rob