AoE's AI is pretty limited really. You can fine tune it a little, but without modding the gameplay itself it's very difficult to make it very strong. Few people have done AI scripting from a purely AI point of view, most of us just modify it for use in scenarios. In a scenario you can make the AI very effective as it's not a level playing field. In a random map game I doubt any AI could really compete with a human. Even in AoK people haven't really made an AI that can compete much with a human without cheating, though they have got close at times. I remember one AoK DM AI ran me over like a steam train - it totally molested me. But when I replayed and thought a bit more about how it worked and what it was going to do I managed to beat it. BBalazs Egypt AI was a recent attempt at 'pure' AI work. You may want to talk to him about it and search the forums/downloads to check it out.

The speed editor is a must for AI/Per file testing. Who wants to watch an AI slowly advancing through the ages over and over?

http://www.digitization.org/wiki/index.php?title=AoESpeed

Well to make effective Ai there a number of files you need to mod. About 5 different files. I did some expermenting but I still didn't make a Ai the was good engouh to beat a human.

Ive been playing around with AIedit, and i would agree its hard to get a good AI. Besides .per and .ai files, what else can be edited?

Edit; looking at cty file articles now

[This message has been edited by Kataphraktoi (edited 06-30-2006 @ 10:37 PM).]

The first and foremost thing is the scenario itself - it needs to fit the AI's needs perfectly. If we're talking about a random map AI, this crosses out, because terrain effects, forests and other stuff vary from game to game.

To make a nicely working AI for Random Map, you need to combine the effectiveness of AI and Per files, and add some CTY tweaking as well, if needed.

I made an Egypt AI quite some time ago, as Rich said. It was meant to fine-tune the gameplay of Egyptians in a random map game. It worked out from that point of view, but it is still no match against a human player.

In Age of Empires, and Rise of Rome (that is AoE1), the computer is meant to play at a population limit of 50; so running it at higher popcaps brings problems. One is that it will never make more than 50 units - this was an older problem, I'm not sure if the new way of setting the popcap overrides this. If it does, and the computer does make more than 50 units, it is still not meant to be played that way... I mean, if you play at a popcap of 200, and the computer does make 200 units, the only difference is probably a multiplier of 4 in the AI file (4*50=200) - which is not exactly how a human player 'works'.

It's sad we can't get AI's to work at the same level of humans. I'd like to see some sort of AI competition though, along with more resources on the subject. Because of the small number of scripters such things are hard to make a success.

Really, I have seen the computer get stuck on forrests and have 20 guys stuck at a forest trying to mine gold, when its the stone age. Ugh. I once made a very compeditive carthaginian AI that it took me longer to beat than BBalaz's Egypt AI. I pitted my Ai against his to end the game in a stalemate, damn priests *shakes fist*. I think Aoe's main problem with AI is that, while it has good features, it doesnt go past build an economy, build a military, maintain the economy, maintain the military. If it was more cause and effect like AoM or AoE3 then it would see you attacking it with cavalry and build more infantry. Instead it just builds more units and keeps attacking and defending while never really changing its style of play.

[This message has been edited by TheRomans (edited 07-03-2006 @ 01:17 AM).]

Let's shake off some dust, I'm in for an AI Contest

Someone, please begin to organise

would be great an AI contest... also i would love to see a comp with good skills...

btw about pop200... inst there in download section a AI which makes persian DM get pop 200?

I think for AI contest we need to get some good rules to make the contest fun. We shouldn't leave it to drawn out stuff like conquest, instead something like wonder building, time limits etc to get a real result. We know how useless AIs can be at finishing each other off, let alone a human

Besides some interesting rules, we also need competitors. To have a contest we would have to advertise well (e.g. of designer/design team websites, AoEH homepage and AI scripters.com). Finally, I think a good effort to get some useful AI info up, e.g. on the wiki, that would make it easy for all new designers to get into AI scripting for AoE would make it much more feasible.

If we were to have the contest, we could take screens of the results on the host's computer to show how it went. I think play-off type format would work well. I don't know if I'd have time to take part, but if there's some interest I'll help with writing some AI/Per/Cty file editing articles on the wiki if others will help me, and I'll be happy to help organise the competition and host some games on my computer.

I think this would make a great contest idea for 2006, then later in the year or next we could have a map design or campaign competition again.

Edit: Some ideas: DM Wonder war, first to reach bronze age?

[This message has been edited by Richard Ames (edited 07-12-2006 @ 02:09 AM).]

It'd be better to make multiple goals - like wonder, okay, but also town defense, army capabilities... Perhaps, a neutral jury.

I could make a "recorded game" avi of a match, but how much do you trust a competitor?

Should we have just a small contest or a big one, I don't care, I'm in

A designer made a AI that came pretty close to human level.

I have never played with more than 50 pop limit. Is it very different from playing at lower limit?

Anyway, I would participate. Does AI also include PER and CTY files or just AI?

What will these AI files be taken for? Random maps or DM? Or maybe other competitions?

I'd like to see a DM AI/Per competition, that would be pretty neat. What do you think about a Wonder War (Death match with standard victory)?

That would definately need me to rethink the ways of an AI
Having 1 villager build that Wonder takes a lot-lot of time, but I'd say it is possible to give it a shot.

First I've got an idea as competition:

1. We could make several categories such as "standard victory", "first-in-bronze-age", "death match" or "wonders only".
2. We could do some kind of cup where AI1 plays against AI2, the winner goes on to the winner of AI3 vs. AI4 and so on.
3. I think it would be fair to play some "best of three" so that there is not too much luck involved.

Second there are a few remarks about designing an AI:

1. I tried to do a nice Yamato file by immitating my own style as Yamato player. Unfortunately this was totally wrong because the computer was incredibly slow that way. He made tool age only after more than 15 minutes! Thus I checked the predifined AI and found out that he creates only 12 villies till tool age and he can do tool in about 11:30. The problem is that the AI simply puts too many villies on wood, stone and especially gold.
2. I also discovered that the standard PER file seems to disable elephant hunting. Why this?
3. The computer has got huge problems with space. He places houses and farms such that either they block the ways or they are very far away from the town center.
4. I once told the AI to build a second town center. It did, but it built it in tool age! So there seems to be a bug which allows the AI to build town centers whenever they like.

Now, I'm going on to PER files and try to make my AI a bit more reasonnable about food collection. There must be a way to make them faster....

edit: All right, I managed to get a few minutes, not really good but ok as first try.
I discovered another strange thing: The AI places the farms pretty close to the town center. But as soon as they are gone the AI places the next farms a little bit further away. This keeps on until he reaches a certain point (city limit perhaps). Is there a way to forbid this "wandering away"? And is it somehow possible to make the AI place farms near the granary instead of the town center?

[This message has been edited by c4master (edited 07-17-2006 @ 12:15 PM).]

1: It is usually not a good idea to imitate your own methods of playing - the computer will never be able to carry out a good CA rush, for example.
2: Even if you allow the computer to hunt elephants, it won't. There is probably some reason why the programmers didn't allow this. The only way I could get the comp to hunt eles is having an ele king attack something from the computer. Than the comp's villies will go and hunt down the ele king.
3: Houses are usually placed next to a dropsite (TC, Granary, Storage pit). Farms can be controlled with a CTY file, though I didn't have time to test it too properly; what I know for sure is that the position of the 1st round of farms is controllable.
4: SSsssshhhh!!! That is my discovery!

I've seen a CPU hunt eles once or twice, but it is rare. Making the number of villies needed to hunt one lower is a start, but it will still be rare.

Seems like we have about 3 people interested in having a competition, but by ourselves it's not really so much a competition as a friendly affair between us. Is there anyone else interested in joining in?

I think in AoE (without RoR) the CPU used to hunt Elephants (or am I wrong?). If that is true there must be a reason for which they stopped this in RoR. Probably because the AI loses too many villies to elephants.

It sounds fun. I'll see if I can fit it in with the other hundreds of things I'm doing. :P But seriously I don't know if I will be able to make one soon because of some of my other projects which include helping out with some of the new Lords games, TGGA and another one for AoE and RoR which if you want to know more about you can find information on it at Tribulation Designs (yes that is ment to make you go there :P) just look on the message board.

All right I made some progress. I downloaded the Persian AI file and tried my own to fight the elephants. Well, I did not win but it was a huge battle which took nearly two hours until I had to give in. I think my AI had not enough food so I'll be planting a few more farms around. Anyway, I was able to get to bronze age faster than the Persian did :P

edit: A little tuning..et voilá! My Yamato AI can now beat the Persian War Elephant AI from Dark Avenger.

Still, I have not found out how he makes his AI do tool age in only 12 minutes. but at least my AI is quicker to reach bronze age :P

[This message has been edited by c4master (edited 07-19-2006 @ 07:11 AM).]

One of the difficult things is countering an opponents AI. How would we arrange that - say what units you must choose from, or do it 'blind' and force people to guess what the opponent's AI will do? Also, what type of game styles would you guys like to see? I'd like to see some sort of DM AI war, that would be a good quick battle and not too hard to design. A random map type of scenario would of course also be a staple, perhaps with a goal of reaching Iron Age first, reaching 10 kills first etc? Also there's the possibility of making a pair of AIs that fight a pair of other AIs in an alliance.

I'm in for any challenge. I just brushed off some dust and made minor corrections to the Egypt Ultimate AI... Today I'm going to continue on that path.

About countering an enemy AI - it is hard, and it is sort of blind, but with just a bit of thinking it doesn't cause much trouble.

So what sort of rules would you like BB? Blind AI or some sort of set definition? What sort of maps/game types would you like to see?

I personally like the idea of a 'blind' game where the enemy players AI(s) are a secret, though the problem with this is that it introduces a large element of chance.

c4master:

Can you tell me how was the games against the Persian AI files? Especially the kind of maps you had played, etc

Just to mail me at yozz1@caramail.com

Thanks!

[This message has been edited by Dark_Avenger (edited 07-22-2006 @ 07:53 AM).]

When all PCs have the AI+Per+Cty installed, you can play a multiplayer scenario and the modded AIs will work. Now, if all humans resign at start, they may still observe.

What about some neutral jury that takes a look at how the AIs work and gives them points along the ages; like the military, economy, technology etc points.

I like RM, but this doesn't mean it has to be RM. Besides, I'm a contestant, not the one who tells how things will go