What would the meta look like if we had a smart AI with perfect mechanics?

[Jaeger]

I don't think it would be very difficult to beat consistently to be honest.

How would you beat it? It has PR35+ strategy and perfect mechanics.

Like I said before, it has imperfect information. So there are a few things you can do.

One that comes to mind is just moving units in and out of LOS on different parts of the map. You can either keep them there or leave. A computer should respond the instant a unit appears on the minimap: either by moving units to intercept or moving villagers etc. Because you're also working outside of LOS a lot of the time when doing this it won't necessarily be able to predict the best course of action (moving vills or sending units to defend).

I think you should be able to invite it to make actions that can be punished in this sense (e.g. ambushing units that go to catch your raid).

Or at the same time, just hide your whole army and wait for it to attack your base and try to trap it. As long as it hasn't scouted you, it has no way of knowing what your unit composition is or where it is.


Basically, unless you gave it maphack I don't think it would do the best strategy all the time, or even the majority of the time. There are probably a lot of cheese builds you can do vs it tbh. Or even standard builds and just doing small things like cutting certain unit types out of your composition for a short period of time to counter something they're doing. It would be mostly about how well you can deny it from scouting you.[/quote]

Thats interesting, I will have to read it again later. But its a smart AI so it could learn from its mistakes, and one definition of PR35 level is that if a human PR 35 wouldnt make that mistake, the AI wont either (its not unrealistic)

[HeatitUP_]

It's possible to come up with the desired result but that would require a lot of work. Rushing was the strongest but the AI will have to factor in lag as well to make it realistic. People like Kynesie will still be lagging even in the the year fucking 3010

[farran34]

It's possible to come up with the desired result but that would require a lot of work. Rushing was the strongest but the AI will have to factor in lag as well to make it realistic. People like Kynesie will still be lagging even in the the year fucking 3010

Why do you hate kynesie so much lol

[HeatitUP_]

It's possible to come up with the desired result but that would require a lot of work. Rushing was the strongest but the AI will have to factor in lag as well to make it realistic. People like Kynesie will still be lagging even in the the year fucking 3010

Why do you hate kynesie so much lol

He goes on smurfs and wastes real time over one game. I hate anyone who fits that description and with his wall mod it makes it even worse because you can still lay out ur wallz in lag. Lag like he creates fries my brain because it's simply not normal at all so fuck kynesie can go to hell

Nobody has figured out he's one of the biggest trolls to come onto AOE3. Just check out his tourney games vs Umeu(he was smart enough to say fuck it) and Acergame on the stream if you don't believe me.

[Mitoe]

It's possible to come up with the desired result but that would require a lot of work. Rushing was the strongest but the AI will have to factor in lag as well to make it realistic.

Playing vs an AI would never lag, and even if it did it wouldn't affect the AI...

[HeatitUP_]

It's possible to come up with the desired result but that would require a lot of work. Rushing was the strongest but the AI will have to factor in lag as well to make it realistic.

Playing vs an AI would never lag, and even if it did it wouldn't affect the AI...

Lag does change the way some units behave.

[yemshi]

Playing vs the AI would never lag...

[HeatitUP_]

Playing vs the AI would never lag...

This is AI vs AI we're talking about. If we can't make lag a factor then I would throw the OP's idea out the window.

[[Armag] diarouga]

Playing vs the AI would never lag...

This is AI vs AI we're talking about. If we can't make lag a factor then I would throw the OP's idea out the window.

Eh?

[musketjr]

basically phish

[mGravitus]

With perfect micro, I think this game would devolve into HI+LI colonial battles. Heavy Cav would not be used.

Think about it, what beats pike +xbow(LI) with perfect micro?

I'v watched pro games where someone would put bunch of pike into cover just as a volley went off, and thought "Damn!, I will never have that APM",

So imagine doing this on an individual unit basis. Like playing aoe3 at 0.1x speed. Like seeing each arrow fire slowly at a unit at toggling it to cover.
You could literally have a perfect unit formation also of a pike on top of each LI.

Heavy Cav would be too easily countered to be cost effective.

Also if AI vs AI, going up to Fort would be impossible. The resources sunk would be soo noticeable, and easily punished. The defenders advantage becomes really weak with perfect Mirco because all you can do is mirco TC fire. Your opponent will be weaving weakened units just out of TC range picking off houses etc. They wont have to siege TC, just contain you long enough, to secure map control.

[_venox_]

Heavy cav can raid...

[Jaeger]

With perfect micro, I think this game would devolve into HI+LI colonial battles. Heavy Cav would not be used.

Think about it, what beats pike +xbow(LI) with perfect micro?

I'v watched pro games where someone would put bunch of pike into cover just as a volley went off, and thought "Damn!, I will never have that APM",

So imagine doing this on an individual unit basis. Like playing aoe3 at 0.1x speed. Like seeing each arrow fire slowly at a unit at toggling it to cover.
You could literally have a perfect unit formation also of a pike on top of each LI.

Heavy Cav would be too easily countered to be cost effective.

Also if AI vs AI, going up to Fort would be impossible. The resources sunk would be soo noticeable, and easily punished. The defenders advantage becomes really weak with perfect Mirco because all you can do is mirco TC fire. Your opponent will be weaving weakened units just out of TC range picking off houses etc. They wont have to siege TC, just contain you long enough, to secure map control.

Idk about the aging thing, you could build all your houses very close to TC. But yeah the computer could always keep the units just out of range of TC if it wanted

[Mitoe]

I don't think that could ever be true just because of the imperfect information again. You need to scout and if you don't you can easily get countered by something like full cav or full LI. Your comp is only good if your opponent's comp doesn't counter you.

Scouting is super important in this game, and even perfect micro isn't going to exempt you from getting cheesed by someone cutting a unit from their comp or hiding a lot of cav to counter you. This is why RTS games aren't all about mechanics, but also how you adapt to your opponent.

[howlingwolfpaw]

this would be interesting to apply to your units as well, so that skirms target HI and goons first unless attacked by cav or you tell it to. And cav seek out Skirms... etc... this could be applied to a commander unit that gives an aura of high caliber AI to your units.

[Jaeger]

I think Mitoe is really underestimating the computer a lot here, I would put my money on it anytime vs anyone in BO7.

Macro:
Even if you find a way to cheese it, like move units in and out of LOS (which first of all won't work because it's a smart AI), it won't matter that much. The computer would have absolutely perfect macro and optimized villager pathing. Nobody knows how many more units it would have, because even at 13 minutes all of it's herds will be herded as close to the TC as possible, it will only shoot an animal with 2 or 3 vills instead of 10, etc. Nobody is even close to that good at macro.

Micro:
When a good player has 30 skirms, they might be able to split their skirms on like 3 musks (unless they focus 100% only on skirms, when let's say they could do 4). Meanwhile, the computer will split the 30 skirms on 5-6 musks, at the same time as microing their 20 cav vs your 20 cav while you don't micro them almost at all.
It would also always win or at least be neutral in the falc war.

[pecelot]

Can somebody tell me the outcome of AI v AI clash in chess? I'm fairly certain it would be played in a very similar way...

[Jaeger]

Can somebody tell me the outcome of AI v AI clash in chess? I'm fairly certain it would be played in a very similar way...

You can do this yourself, just open the same game of chess on 2 devices (or maybe the same one), play against each of them, and copy each other's moves exactly

[Papist]

A computer would have perfect mechanics, but wouldn't be able to respond to every situation (weird starts).

[Imperial Noob]

Both "PR 35-40 strategy level" and "perfect strategy" require making an educated guess once in a while...
The best words to describe an AI you here speculate about, had it an ability to guess, would be "downright depressing".
It would steamroll us all nearly every game, and even if it guessed poorly, it would still have its perfect mechanics and learning ability to fix the situation and make a comeback better, than any player would.

Such AIs would probably scout a lot to gain information (resque native scouts, native warriors, scout with sheep..., then build cavalry... they would predict how much of any resource their opponent has - they would count villagers gathering it, etc.).

Edit: 5 cossacks / 5 uhlans /4 axe riders... those would probably be pretty high on the list of priorities in age II.

[princeofcarthage]

If you would have 2 perfect AI comps fighting against each other, it would be a never ending game.

[_H2O]

Both "PR 35-40 strategy level" and "perfect strategy" require making an educated guess once in a while...
The best words to describe an AI you here speculate about, had it an ability to guess, would be "downright depressing".
It would steamroll us all nearly every game, and even if it guessed poorly, it would still have its perfect mechanics and learning ability to fix the situation and make a comeback better, than any player would.

Such AIs would probably scout a lot to gain information (resque native scouts, native warriors, scout with sheep..., then build cavalry... they would predict how much of any resource their opponent has - they would count villagers gathering it, etc.).

Edit: 5 cossacks / 5 uhlans /4 axe riders... those would probably be pretty high on the list of priorities in age II.

This is something I am interested in doing. Something like a running tab in the computer of how much is possible vs what is known and making a calculated decision on the risk. It's quite hard as this changes second by second. The longer and more that is hidden the more speculation would be needed and the more likely a big tech switch would be.

[farran34]

If you would have 2 perfect AI comps fighting against each other, it would be a never ending game.

nah, whoever got better treasures or better map spawn would win probably

[farran34]

Can somebody tell me the outcome of AI v AI clash in chess? I'm fairly certain it would be played in a very similar way...

AI vs AI games are often quite strange from my experience because they will follow the opening book lines very very far, often not choosing the best moves but just followed programmed lines (I do not think they are programmed generally to divert much from standard lines in openings), and usually I think it ends in a draw.

The way AIs play chess (against people) would be quite different from aoe3 I imagine since the games are quite different.

[Jaeger]

Even on a perfectly even map tge games wouldnt be infinite if its not a mirror