Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Neruz]

It makes sense if you think about it; in tight formation if the archers at the back tried to use a low arc they'd shoot the guys in front of them in the back of the head.

[umiman]

Well, Aqizzar, the way archers work is kinda weird. In tight formation, only those in front can use direct fire. Use spread formation when in no fear of melee combat.

Also, archers automatically use ballistic fire on things behind cover, such as battlements.

Not true. Archers will shoot straight even in tight formation if there's nothing in the way (say... parapets / hill / other units located higher than the archers). While the game engine essentially makes it impossible to fire straight at units on walls, if you want to force your archers to shoot straight just clear the way or position them on slightly higher ground. Also, when units are behind the parapets, any unit that has a clear view through the slits will automatically shoot straight. If there are any units behind the walls where there are no parapets in the way, the entire unit will shoot straight too.

Not to mention that archers on elevated positions get massive bonuses to range and damage.

Jackrabbit: Assassins are good. They'll get rid of your briber troubles.

Aqizzar: You may think you'd be at a disadvantage when you turn off AI moves but trust me. Your life will be a lot more fun when you do. Anything the AI does that is important is normally told to you anyway and on the higher difficulties, the AI just automagically spawns units all over the bloody place, so it's pointless to watch their troop movements as they essentially have unlimited units. I remember on the hardest difficulty, I would trounce a 5000 unit army completely, then have the same general in ONE turn fight me with another full stack, then again after that with a brand new full stack. Pretty damn unfair at that level but it's great for a challenge.

[Jackrabbit]

Yeah, I think I'm turning it off.

[Aqizzar]

Aqizzar: You may think you'd be at a disadvantage when you turn off AI moves but trust me. Your life will be a lot more fun when you do. Anything the AI does that is important is normally told to you anyway and on the higher difficulties, the AI just automagically spawns units all over the bloody place, so it's pointless to watch their troop movements as they essentially have unlimited units. I remember on the hardest difficulty, I would trounce a 5000 unit army completely, then have the same general in ONE turn fight me with another full stack, then again after that with a brand new full stack. Pretty damn unfair at that level but it's great for a challenge.

Has this new century of ours really not come far enough that the AI in 4X games can't put up a decent fight without creating its armies from thin air?  I could accept that from the 1990s, it was a pretty small industry.  But dammit, there are enough programmers now that the computer should be able to make a reasonable stand without getting its shit for free.

[Neruz]

Aqizzar: You may think you'd be at a disadvantage when you turn off AI moves but trust me. Your life will be a lot more fun when you do. Anything the AI does that is important is normally told to you anyway and on the higher difficulties, the AI just automagically spawns units all over the bloody place, so it's pointless to watch their troop movements as they essentially have unlimited units. I remember on the hardest difficulty, I would trounce a 5000 unit army completely, then have the same general in ONE turn fight me with another full stack, then again after that with a brand new full stack. Pretty damn unfair at that level but it's great for a challenge.

Has this new century of ours really not come far enough that the AI in 4X games can't put up a decent fight without creating its armies from thin air?  I could accept that from the 1990s, it was a pretty small industry.  But dammit, there are enough programmers now that the computer should be able to make a reasonable stand without getting its shit for free.

AI has come almost nowhere in the last few decades. There have been improvements most certainly, but those improvements are mostly just new ways of brute forcing things and more subtle ways of cheating. Inevitably, against a seasoned player the AI must cheat for it to have any hope.

The problem AI's run into is they're not actually Intelligent at all, and thus they have no imagination and no way to predict things, this causes them to be purely reactionary and highly exploitable.

[Aqizzar]

The problem AI's run into is they're not actually Intelligent at all, and thus they have no imagination and no way to predict things, this causes them to be purely reactionary and highly exploitable.

I don't know if the author was reading too much into things, but supposedly Galactic Civilizations 2 had a pretty amazing gaming brain, if this legendary tale is to be believed.

[umiman]

The problem AI's run into is they're not actually Intelligent at all, and thus they have no imagination and no way to predict things, this causes them to be purely reactionary and highly exploitable.

I don't know if the author was reading too much into things, but supposedly Galactic Civilizations 2 had a pretty amazing gaming brain, if this legendary tale is to be believed.

Oh yeah. Don't frigging mess with the Galciv 2 AI. It's really scary. It will give you nightmares on the highest difficulties and it does it without cheating.

I just can't stand the combat in Galciv 2.

[Neruz]

The problem AI's run into is they're not actually Intelligent at all, and thus they have no imagination and no way to predict things, this causes them to be purely reactionary and highly exploitable.

I don't know if the author was reading too much into things, but supposedly Galactic Civilizations 2 had a pretty amazing gaming brain, if this legendary tale is to be believed.

Oh yeah. Don't frigging mess with the Galciv 2 AI. It's really scary. It will give you nightmares on the highest difficulties and it does it without cheating.

I just can't stand the combat in Galciv 2.

It still cheats. It doesn't cheat in the 'free resources\units' manner, but it does cheat it's pants off in the intel manner. The AI uses the fact that at any given time it knows exactly what you have done, what resources are at your disposal etc and abuses the hell out of it. The only reason the AI doesn't completely dominate you at all times is because there are hardcoded restrictions in place to make it pretend that it's not cheating it's electronic pants off.

Ultimately, even if you did have an AI that didn't cheat in the strictest sense, it would still have access to orders of magnitude more computing power than you could ever have, and thus would still be 'cheating' when compared to a human.

[umiman]

I sincerely doubt that Neruz. Please give some evidence. I know the AI gets minor economic bonuses at the highest difficulty levels (even though it doesn't need it) but I don't particularly consider that to be cheating. However, to claim the AI has omniscience is something I really don't believe.

[Aqizzar]

That idea probably came from the developer's blog, which talked at length about complaints people had about the AI in GalCiv 1, which worked exactly like Neruz said.  And while I don't know if it's true or not that GalCiv 2 works the same way, I'd be surprised if it wasn't true.  Think about it - the AI is part of the program that you're playing with.  By definition, it's plans and strategies have to draw from some kind of information.  The info fed to it is the what's going on in the game.  For the AI to not be omniscient would require making a double-blind filter, where the AI can only use information it's logged as encountering.

Entirely and obviously possible, but probably a Hell of a lot harder to make work than just letting the AI see everything in the game's memory.

[Neruz]

That idea probably came from the developer's blog, which talked at length about complaints people had about the AI in GalCiv 1, which worked exactly like Neruz said.  And while I don't know if it's true or not that GalCiv 2 works the same way, I'd be surprised if it wasn't true.  Think about it - the AI is part of the program that you're playing with.  By definition, it's plans and strategies have to draw from some kind of information.  The info fed to it is the what's going on in the game.  For the AI to not be omniscient would require making a double-blind filter, where the AI can only use information it's logged as encountering.

Entirely and obviously possible, but probably a Hell of a lot harder to make work than just letting the AI see everything in the game's memory.

It's not harder, but it does gimp your AI, since as i said, ultimately it has no imagination, which is a serious flaw in complex games like GalCiv.


What most games do is point at the AI and say 'you can't use this information'. On lower difficulty levels, they're not allowed use lots of information, in fact many 'easy' AI's are restricted to the same amount of information that the player has. Their lack of imagination then gimps them badly, which is why easy AI's tend to suck.

As the difficulty level increases, the AI is allowed to play with more and more information, until often at higher levels all the restrictions are lifted entirely.



I'm not sure about at lower levels in GalCiv2, but at higher levels the AI most certainly does cheat it's pants off. Max level AI's take a look at what you're building and researching and then build and research the counters to those things, you can actually see it if you cheat yourself and check out what the AI is doing, the AI will often deliberately switch tech-trees just to get the best possible counter to whatever it is your doing.


Ironically, the AI also does this against itself, if you get lucky and avoid it's wrath, at higher levels it will get into a rediculous progressive arms race with itself that looks like something straight out of the Lennsman series. Of course it's behaviour is randomised and it's given species-specific restrictions to try and avoid these sort of self-perpetuating loops.



The other thing you might notice against the AI on harder difficulties is that it has a remarkable knack for putting ships in exactly the wrong place at exactly the wrong time, this isn't luck, it knows exactly what it is doing.





And of course, finally, the AI is perfect at mathematics. It cannot accidentally stall it's economy and when many of the lower difficulty behaviour restrictions are removed it can use it's economy to a far greater efficiency than any human ever could. The AI has no limited attention span and is able to pay full attention to everything it owns and mesh the entirity together as a whole. If it wasn't for the fact that the AI is not actually Intelligent, it would soundly trounce you every single time from this advantage alone.


Another example of this kind of 'cheating' is Aimbots in FPS's. The AI has absolutely 100% perfect aim, because it can do complex ballistics calculations far faster and to a much higher degree of accuracy than you ever could. The game has to deliberately randomise the AI's shooting to prevent it from always winning. This isn't the AI using information it shouldn't have (although it often does as well) this is just the innate advantage an electronic computer has over the human brain.

[Aqizzar]

Another example of this kind of 'cheating' is Aimbots in FPS's. The AI has absolutely 100% perfect aim, because it can do complex ballistics calculations far faster and to a much higher degree of accuracy than you ever could. The game has to deliberately randomise the AI's shooting to prevent it from always winning. This isn't the AI using information it shouldn't have (although it often does as well) this is just the innate advantage an electronic computer has over the human brain.

You greatly underestimate my obsession with micromanagement.  Point taken all the same, but I'm curious about this "arms race with itself" bit.  You mean the different AI nations?  Because they are supposed to compete with each other after all.

[Neruz]

Another example of this kind of 'cheating' is Aimbots in FPS's. The AI has absolutely 100% perfect aim, because it can do complex ballistics calculations far faster and to a much higher degree of accuracy than you ever could. The game has to deliberately randomise the AI's shooting to prevent it from always winning. This isn't the AI using information it shouldn't have (although it often does as well) this is just the innate advantage an electronic computer has over the human brain.

You greatly underestimate my obsession with micromanagement.  Point taken all the same, but I'm curious about this "arms race with itself" bit.  You mean the different AI nations?  Because they are supposed to compete with each other after all.

It doesn't happen often, but every now and then you'll see two races get into a massive arms race.

Each race will begin researching the counters for the other race's techs, they'll then notice that their enemy just researched the counter for what they just finished researching, so they'll start researching the counter to that, ad infinitum.

I actually had one game where the two factions never actually built any warships, they just spent the entire game time researching stuff to counter the stuff that their opponent had just researched to counter the stuff that they had just researched to counter the stuff that their opponent had just researched to counter the stuff that they had just researched to counter the stuff that their opponent had just researched to counter the stuff that they had just researched to counter the stuff that their opponent had just researched to counter the stuff that they had just researched to counter the stuff that their opponent had just researched to counter the stuff that they had just researched to counter the stuff that their opponent had just researched to counter the stuff that they had just researched and so on, ad infinitum.


It's rare, because both races need to have about the same research ability and it only really happens at the max level AI, as lower AI levels seem to get 'randomized' to occasionally do something stupid and create a weakness the player exploits. But it's always amusing when it happens.


Presumably if the game lasted long enough they'd max out their research trees and get a tech victory by virtue of having researched everything.




Also, Chin, it doesn't matter how obsessive you are at micromanaging. You will never be as efficient and infallible as a computer. That's why we have computers in the first place.

[Aqizzar]

Also, Chin, it doesn't matter how obsessive you are at micromanaging. You will never be as efficient and infallible as a computer. That's why we have computers in the first place.

Like I said, point taken.  Bear in mind, GalCiv is vastly less complicated that anything like simulating a real economy, hence the fact that an idle non-professional gamer can win against the AI at all.  There's only a handful of things to build and a couple sliders to hammer into finely honed position.  Most often, my problem isn't mistakes in micromanagement; I just get so bored of trimming perfect efficiency that I abandon the game and start another one.

[Neruz]

Also, Chin, it doesn't matter how obsessive you are at micromanaging. You will never be as efficient and infallible as a computer. That's why we have computers in the first place.

Like I said, point taken.  Bear in mind, GalCiv is vastly less complicated that anything like simulating a real economy, hence the fact that an idle non-professional gamer can win against the AI at all.  There's only a handful of things to build and a couple sliders to hammer into finely honed position.  Most often, my problem isn't mistakes in micromanagement; I just get so bored of trimming perfect efficiency that I abandon the game and start another one.

Go is a sufficiently complicated game that the best Go masters can reliably beat the computer at it.

Ironically, this is no longer the case with Chess, i believe there is at least one learning Chess computer that has become effectively undefeatable. This is because Chess is actually a much simpler game to compute than Go is.


Computers are extremely good at a very specific and restricted set of tasks. Once outside those tasks, they rapidly fall to pieces. Tasks involving meta knowledge; that of knowing the mind of the player you're playing against rather than the actions that player is making are a particular weakness for compies, which is why they're not very good at Go, a game of Go at a high level has very little to do with the board game and everything to do with the mind game.